Systems and methods for using multi-dimensional x-ray imaging in meat production and processing applications

Abstract

The present specification teaches an imaging system for assessing meat quality. The imaging system includes an X-ray scanning system that generates X-ray scanning data of the meat and a hyperspectral imaging system that generates hyperspectral imaging data. A computing device acquires X-ray scan data and hyperspectral imaging data, automatically determines a quality of meat by analyzing the acquired X-ray scan data in conjunction with the hyperspectral imaging data, classifies the meat as one of an acceptable quality category and an unacceptable quality category based on the determined quality; and generating data indicative of the quality of the meat.

Description

[0001] Cross-referencing

[0002] This application relies on priority to U.S. Patent Application No. 18 / 329,434, filed June 5, 2023, entitled "Systems and Methods for Using Multi-Dimensional X-Ray Imaging in Meat Production and Processing Applications," the entire contents of which are incorporated herein by reference. Technical Field

[0003] This specification generally pertains to the field of raising animals and / or livestock on farms for processing and producing meat products derived from them. More specifically, this specification relates to the use of three-dimensional (3D) fixed-bench computed tomography (CT) systems to improve farming practices, thereby enhancing the quality of farmed animal products in addition to improving the management of slaughterhouse production processes. Background Technology

[0004] Farm production involves livestock destined for consumption in the human and animal food chain, including but not limited to poultry, pigs, goats, sheep, and cattle. Unlike other industries where products can be blended to achieve a level of consistency, each animal possesses individual characteristics that guarantee consumer satisfaction. The way animals are raised or processed on farms often influences characteristics that affect customer satisfaction with animal-derived meat products such as steaks or lamb chops. Consumers are increasingly valuing the quality of the meat products they consume, food safety, and food traceability. As an example, animals raised on cattle farms are sold and processed in meat processing plants to produce a variety of meat products within the food chain. Strict quality control measures are in place to ensure that animals entering the plant undergo optimal processing to produce products that meet expected consumer satisfaction in terms of edibility, food chain traceability, and food safety.

[0005] To meet such consumer demand, in addition to regular farming activities, farmers need to demonstrate compliance with standards and practices, which places a considerable burden on farmers. Therefore, the goal of farmers is to breed the highest value animals under the farming conditions of a particular farm location (high altitude, low altitude, warm, cool, humid, dry, lush, barren) and at the lowest possible cost. This means managing the cost of food, water, veterinary needs, transportation, and maintenance to provide the greatest return. Currently, farmers use a range of information sources to plan their farming practices, including weather forecasts, satellite images for pasture and water management, animal tracking for determining the optimal location of feed and water troughs, genetic analysis for herd development, and veterinary records. Typically, this information is processed by the farmer using his own farming experience in order to optimize animal health, lean meat yield (the amount of meat compared to fat or bone), and subsequent return on investment.

[0006] Once the animals reach the meat processing plant or factory, the animals are typically first slaughtered; the head, internal organs, hide, and limbs are then removed; and the carcass is placed in a chill room for a period of time to hang while the fat solidifies. Once the carcass is rigid, it is cut into large pieces called primal pieces. Each primal is then passed to a boning area where retail-ready meat cuts are processed into bone-in or boneless cuts before being packaged and transferred to the retail supply chain. Hundreds of people stand shoulder to shoulder, each performing a specific set of actions as the carcass or primal passes in front of them, in this labor-intensive process, carcasses are typically hung on a moving rail, and prims are typically on a moving conveyor belt. Each individual in the boning area is provided with instructions about which meat cuts are needed each day to meet customer demand and to meet production targets. The result is a production process, but not one that is typically operated at peak efficiency.

[0007] Efficiency losses come from trimming excess meat from the meat of retail cuts, thereby placing valuable product into a lower grade food supply chain, for example, over-cutting valuable rib eye muscle so that it ends up as lower value ground meat. Further efficiency losses come from inaccurate production plans, where carcasses are processed into a sub-optimal set of retail cuts. This is typically because the production plan provided to the individual cutting team is not specific to each individual carcass, but rather reflects average production targets for the full set of carcasses to be processed that day.

[0008] Each individual working in a factory has an obligation to meet high standards of food safety, but in some cases, the carcass can contain invisible contamination or health defects that are not possible for the individual to determine that are hidden beneath the visible surface of the carcass. This can lead to occasional but significant food safety issues, the mitigation of which can be expensive and complex. Furthermore, when retail cuts of meat are produced and packaged, errors occasionally occur in food labeling and packaging, resulting in the shipment of incorrect products to customers. This error results in the rejection of the product by retail customers or consumers, sometimes in large quantities. In these cases, there is a negative financial impact on the processor, and often the rejected product needs to be destroyed. It should also be noted that meat processing plants or factories primarily employ individual workers who use knives to dissect the carcass into the desired consumer product in stages. Therefore, each worker in the meat processing line responsible for slaughtering an animal all the way through to the final packaging of the product must receive a high level of training in order to achieve the proper cutting technique at the processing line speed required to achieve commercially satisfactory results on a repeatable basis.

[0009] In some sectors, the use of automation to replace or augment labor is common (e.g., in poultry processing), but in other sectors, the use of automation is limited (e.g., beef processing). To a large extent, this is driven by the complexity and variation in the anatomy from one carcass to another. In poultry, this variation is relatively small, while in beef, the variation can be large depending on the breed and weight of the carcass being processed.

[0010] At the retail end, customers of meat products have specific requirements for the quality and cut of meat that they purchase from the meat factory. These can include meat grading, fat thickness, weight, and other factors that the processor must adhere to regardless of the supply of animals that enter the factory. Given that the processor only knows the actual anatomy of the carcass during the dissection process in the factory, it is difficult to plan optimal production based on the significant variation in size, weight, and quality of the animals that arrive at the factory. This can result in directing higher quality products to lower value output streams, resulting in a reduction in yield and factory efficiency.

[0011] Meat quality grading systems tend to rely on relatively subjective measurements of the carcass and can include features such as, but not limited to, the following: a) comparing the color of the meat to a standard color chart at specific locations in the carcass; b) comparing the marbling and fat content of the carcass to a set of standardized photographs; and c) the size of the force required to make an indentation at specific points on the surface of the carcass among other subjective indicators. This measurement tends to be point-based and does not measure the natural variation in meat quality that can occur within or between muscle groups.

[0012] Accordingly, there is a need for using x-ray scanning systems and methods to improve farming practices, resulting in higher valuation of the raised animals. There is also a need for using x-ray screening at various stages of the animal life cycle during development on the farm, in order to better characterize the food quality and food safety of meat products from the herd. There is also a need for improving production efficiency, reducing labor utilization, adopting a carcass-centric production approach, improving plant and food safety performance, and reducing losses due to mislabeling and mispackaging of products. Accordingly, there is a need for using x-ray scanning systems and methods to improve quality control, consumption quality, carcass valuation, and food safety in a meat processing plant or slaughterhouse. There is also a need for using x-ray screening to aid overall production planning and automation, to improve slaughterhouse management. SUMMARY

[0013] The following embodiments and aspects thereof are described and illustrated in conjunction with systems, tools, and methods which are meant to be exemplary and illustrative, and not limiting in scope. Numerous embodiments are disclosed.

[0014] The present specification discloses an imaging system configured to evaluate meat, comprising: an x-ray scanning system configured to generate x-ray scanning data of the meat; a hyperspectral imaging system configured to generate hyperspectral imaging data; a computing device in data communication with the x-ray scanning system and the hyperspectral imaging system, wherein the computing device comprises a processor and a memory storing a plurality of programming instructions that, when executed by the processor, configure the processor to: acquire the x-ray scanning data and the hyperspectral imaging data; automatically determine a quality of the meat by analyzing the acquired x-ray scanning data in conjunction with the hyperspectral imaging data; classify the meat into one of an acceptable quality category and an unacceptable quality category based on the determined quality; and generate data indicative of the quality of the meat.

[0015] Optionally, the x-ray scanning system comprises a two-dimensional projection x-ray imaging system in combination with a multi-energy x-ray (MEXA) sensor, the two-dimensional projection x-ray imaging system having at least one of a single view configuration or a dual view configuration. Optionally, the x-ray scanning system comprises an inclined conveyor such that an entry end of the conveyor is at a lower elevation position than an exit end of the conveyor. Optionally, the x-ray scanning system uses a downwardly inclined conveyor such that an entry end of the conveyor is at a higher elevation position than an exit end of the conveyor.

[0016] Optionally, the hyperspectral scanning data comprises data in a visible light wavelength range and a short wave infrared wavelength range.

[0017] Optionally, the meat comprises internal organs and viscera.

[0018] Optionally, the system further comprises at least one of an inkjet, a laser beam, an LED strip, or an augmented reality headset adapted to produce a visual indication of the quality associated with the meat.

[0019] Optionally, the processor is further configured to generate at least one graphical user interface to display at least one image corresponding to the X-ray scan data and determine the quality based on the data indicative of the thickness and / or density of the meat.

[0020] Optionally, the system further comprises a conveyor that translates the meat through the system at a speed in the range of 0.1 m / s to 1.0 m / s.

[0021] Optionally, the multi-sensor imaging system has an inspection tunnel with a length in the range of 1100 mm to 5000 mm, a width in the range of 500 mm to 1000 mm, and a height in the range of 300 mm to 1000 mm.

[0022] Optionally, the X-ray scanning system comprises a first X-ray source having 120 to 160 keV with 0.2 to 1.25 mA beam current and a second X-ray source having 120 to 160 keV with 0.2 to 1.25 mA beam current, wherein the first X-ray source is configured in an overhead shooter configuration and the second X-ray source is configured in a side shooter configuration. Optionally, the X-ray scanning system comprises a multi-energy photon counting X-ray sensor array. Optionally, the X-ray scanning system comprises 6-22 data acquisition panels corresponding to the first X-ray source and 4-20 data acquisition panels corresponding to the second X-ray source.

[0023] Optionally, the X-ray scanning system is configured to acquire data in a plurality of energy bands, wherein the plurality of energy bands ranges from 3 to 20 and each energy band ranges from 20-160 keV.

[0024] Optionally, the hyperspectral imaging system comprises a first camera sensor configured to perform visible imaging in a 200 to 1200 wavelength band and a second camera sensor configured to perform short wave infrared imaging in a 400 to 700 wavelength band. Optionally, the first camera sensor is configured to operate in the range of 400 nm to 900 nm and have a spectral resolution of at least 20 nm over the width of the conveyor with a pixel size no more than 2.0 mm. Optionally, the second camera sensor is configured to operate in the range of 900 nm to 1800 nm and have a spectral resolution of at least 20 nm over the width of the conveyor with a pixel size no more than 2.0 mm.

[0025] Optionally, the hyperspectral imaging system is configured to have an acquisition rate of 30 to 150 Hz.

[0026] Optionally, the X-ray scanning system and the hyperspectral imaging system are synchronized to an X-ray fundamental frequency ranging from 150 to 500 Hz.

[0027] Optionally, the processor is further configured to determine a type of meat based on the acquired X-ray scan data and hyperspectral imaging data.

[0028] Optionally, the processor is further configured to: generate at least one graphical user interface to display at least one image corresponding to the hyperspectral imaging data; identify a region in the at least one image indicative of an anomaly; and apply an annotation to the identified region, wherein the annotation is at least one of a shape or a color. Optionally, the processor is configured to implement at least one machine learning model, wherein the machine learning model is configured to analyze the hyperspectral imaging data in order to determine a quality of the meat and the region indicative of the anomaly. Optionally, the machine learning model is adapted to be trained using K-means clustering in order to identify the region indicative of the anomaly.

[0029] Optionally, the data indicative of the quality of the meat includes at least one of lean meat yield, ratio of intramuscular fat to tissue, amount of intermuscular fat, absolute size of individual organs, relative size of individual organs, muscle volume, number of ribs, presence or absence of disease, presence or absence of cysts, presence or absence of tumors, presence or absence of pleurisy, or presence or absence of foreign objects.

[0030] The present specification also discloses a system for generating data indicative of animal breeding practices and meat production practices, comprising: a plurality of geographically distributed meat production sites having associated therewith a multi-sensor imaging system, wherein each of the multi-sensor imaging systems comprises an X-ray scanning system and a hyperspectral imaging system; at least one server in data communication with a database and each of the multi-sensor imaging systems, wherein the at least one server includes a processor and a memory storing a plurality of programming instructions that, when executed by the processor, configure the processor to: implement at least one machine learning model; provide a plurality of data accessed from the database as input to the at least one machine learning model, wherein the at least one machine learning model is configured to analyze the plurality of data to generate the data, wherein the data is intended to maximize a plurality of positive parameters related to animal breeding and meat production and minimize a plurality of negative parameters related to animal breeding and meat production; and enable a plurality of geographically distributed computing devices to access the generated data.

[0031] Optionally, the plurality of data corresponds to a collection of a plurality of animal and meat related data from each of a plurality of geographically distributed meat production sites, and wherein the plurality of animal and meat related data comprises at least one of: animal ID, animal type, animal breed, X-ray scan data corresponding to each different age of the animal, X-ray scan data of the animal carcass and / or raw meat, hyperspectral image data of the animal meat and organs, geographical location of the farm and / or meat production site, climate, weather, season, type of feed, time of year of meat production, vaccination history, medication, disease history, age of the animal upon receiving at the meat production site, lean meat yield, ratio of intramuscular fat to tissue, amount of intermuscular fat, absolute size of individual organs, relative size of individual organs, muscle volume, number of ribs, presence or absence of disease, presence or absence of cysts, presence or absence of tumors, presence or absence of pleurisy, or presence or absence of foreign objects.

[0032] Optionally, the plurality of positive parameters comprises at least one of: reduced medication requirements, lower carbon footprint, variable cost efficiency, reputation protection, lower health risk to consumers, or improvement in lean meat yield, ratio of intramuscular fat to tissue, amount of intermuscular fat, absolute size of individual organs, relative size of individual organs, muscle volume, number of ribs, absence of disease, absence of cysts, absence of tumors, absence of pleurisy, or absence of foreign objects.

[0033] Optionally, the plurality of negative parameters comprises at least one of an increase in presence of disease, presence of cysts, presence of tumors, presence of pleurisy, or presence of foreign objects.

[0034] In some embodiments, the present specification discloses a method of evaluating meat quality, comprising: operating a multi-sensor imaging system, the multi-sensor imaging system comprising: an X-ray scanning system configured to generate X-ray scan data of meat; a hyperspectral imaging system configured to generate hyperspectral imaging data; acquiring the X-ray scan data and the hyperspectral imaging data; automatically determining a health status of the meat by analyzing the acquired X-ray scan data and / or hyperspectral imaging data; classifying the meat into one of a healthy and an unhealthy category based on the determined health status; and generating data indicative of quality of the meat.

[0035] Optionally, the X-ray scanning system uses 2D projection X-ray imaging in a single view or dual view configuration with dual-energy or multi-energy X-ray (MEXA) sensors.

[0036] Optionally, the X-ray scanning system uses a conveyor positioned on an upwardly inclined plane such that a first end of the conveyor is at a lower elevation position than a second, opposite end of the conveyor.

[0037] Optionally, the X-ray scanning system uses a conveyor positioned on a downwardly sloped surface such that a first end of the conveyor is at a higher elevation position than a second, opposite end of the conveyor.

[0038] Optionally, the hyperspectral scanning data includes visible light and short wave infrared scanning data.

[0039] Optionally, the meat includes internal organs and organs.

[0040] Optionally, the multi-sensor imaging system includes an inkjet, a laser beam, an LED light strip, or an augmented reality headset to indicate the presence of a health issue while scanning the meat.

[0041] Optionally, the method further includes generating at least one graphical user interface to display at least one image corresponding to the X-ray scanning data and determining the health status based on a threshold value indicative of a thickness and / or density of the meat.

[0042] Optionally, the multi-sensor imaging system includes a conveyor that translates the meat through the multi-sensor imaging system at a speed of about 0.2 m / s.

[0043] Optionally, the multi-sensor imaging system has an inspection tunnel that is 1360 mm long, 630 mm wide, and 400 mm high.

[0044] Optionally, the X-ray scanning system includes a first X-ray source and a second X-ray source having a 160 keV with a 1.0 mA beam current, wherein the first X-ray source is configured in a top shooter configuration and the second X-ray source is configured in a side shooter configuration.

[0045] Optionally, the X-ray scanning system includes a multi-energy photon counting X-ray sensor array.

[0046] Optionally, the X-ray scanning system includes 11 data acquisition plates corresponding to the first X-ray source and 9 data acquisition plates corresponding to the second X-ray source.

[0047] Optionally, the X-ray scanning system is configured to have an acquisition rate of 300 Hz in six energy bands, and wherein the six energy bands are in a range of 20-160 keV.

[0048] Optionally, the hyperspectral imaging system includes a first camera sensor configured for visible light imaging in a 300 wavelength band and a second camera sensor configured for short wave infrared imaging in a 512 wavelength band.

[0049] Optionally, the first camera sensor operates in a range of 400 nm to 900 nm, has a spectral resolution of at least 20 nm over a width of the conveyor, and a pixel size of no more than 2.0 mm.

[0050] Optionally, the second camera sensor operates in the range of 900 nm to 1800 nm, has a spectral resolution of at least 20 nm over the width of the conveyor, and has a pixel size of no more than 2.0 mm.

[0051] Optionally, the hyperspectral imaging system is configured to have an acquisition rate of 30 to 150 Hz.

[0052] The X-ray scanning system and the hyperspectral imaging system are synchronized with an X-ray base frequency of 300 Hz.

[0053] Optionally, the method further comprises determining a type of the meat based on the acquired X-ray scanning data and the hyperspectral imaging data.

[0054] Optionally, the method further comprises generating at least one graphical user interface to display at least one image corresponding to the hyperspectral imaging data; identifying a region in the at least one image that is indicative of an anomaly; and applying a color and / or shape annotation to the identified region, wherein the shape annotation is one of a circle or a box.

[0055] Optionally, the machine learning model is configured to analyze the hyperspectral imaging data in order to determine a health condition of the meat and to identify a region that is indicative of an anomaly. Optionally, the machine learning model is trained using K-means clustering in order to identify the region that is indicative of the anomaly.

[0056] Optionally, the data indicative of the quality of the meat comprises a plurality of post- slaughter parameters including lean meat yield, ratio of intramuscular fat to tissue, amount of intermuscular fat, absolute and relative size of individual organs, muscle volume, number of ribs, and presence or absence of diseases such as cysts, tumors, pleurisy, and foreign bodies.

[0057] The foregoing and other embodiments of the present specification will be more thoroughly described in the accompanying drawings and detailed description provided below. BRIEF DESCRIPTION OF DRAWINGS

[0058] These and other features and advantages of the present specification will be further understood by reference to the following detailed description, when considered in connection with the accompanying drawings, in which:

[0059] FIG. 1A A first cross-sectional side view of a 3D stationary gantry X-ray CT imaging system configured to scan cattle of a farm is shown in accordance with some embodiments of the present specification;

[0060] FIG. 1B A second cross-sectional side view of the 3D stationary gantry X-ray CT imaging system of FIG. 1A in accordance with some embodiments of the present specification is shown;

[0061] FIG. 1CA 3D stationary gantry X-ray CT imaging system including multiple X-ray tubes is shown in accordance with some embodiments of the present specification;

[0062] FIG. 2 A bottom view, a top view, a longitudinal side view, and an end view of a linear multi-focus X-ray source for a 3D stationary gantry X-ray CT imaging system are shown in accordance with some embodiments of the present specification;

[0063] FIG. 3A A first side view, a second side view, and a top view of a single-plane stationary gantry X-ray computed tomography system configured to scan cattle on a farm are shown in accordance with some embodiments of the present specification;

[0064] FIG. 3B A first side view of a single-plane stationary gantry X-ray computed tomography system including a radar imaging or inspection system is shown in accordance with some embodiments of the present specification; FIG. 3A

[0065] FIG. 4 An exemplary stepped frequency continuous wave radar scan sequence is shown in accordance with some embodiments of the present specification;

[0066] FIG. 5 A block diagram of a radar imaging system in accordance with some embodiments of the present specification;

[0067] FIG. 6 An exemplary arrangement of multiple transmitter (Tx) and receiver (Rx) elements of a radar imaging or inspection system is shown in accordance with some embodiments of the present specification;

[0068] FIG. 7 A block diagram of multiple exemplary information, outputs, or results derived based on processing or analysis of animal scan image data generated using a 3D stationary gantry X-ray CT imaging system in accordance with some embodiments of the present specification;

[0069] FIG. 8 A workflow showing use of multiple 3D X-ray computed tomography processes during various events related to livestock farming in accordance with some embodiments of the present specification;

[0070] FIG. 9A A top view of a 3D stationary gantry X-ray CT imaging system in a first configuration for scanning meat in a slaughterhouse is shown in accordance with some embodiments of the present specification;

[0071] FIG. 9B A top view of the 3D stationary gantry X-ray CT imaging system in a second configuration for scanning meat in a slaughterhouse is shown in accordance with some embodiments of the present specification; FIG. 1A ​​

[0072] FIG. 10A First, second and third cross-sectional views of a 3D stationary gantry X-ray CT imaging system configured for dual plane scanning of a carcass according to some embodiments of the present specification are shown;

[0073] FIG. 10B A fourth cross-sectional view of a 3D stationary gantry X-ray CT imaging system according to some embodiments of the present specification is shown;

[0074] FIG. 11 First, second and third cross-sectional views of a 3D stationary gantry X-ray CT imaging system configured for dual plane scanning of a carcass according to some embodiments of the present specification are shown;

[0075] FIG. 12 A cross-sectional view of a 3D stationary gantry X-ray CT imaging system configured for single plane scanning of a carcass according to some embodiments of the present specification is shown;

[0076] FIG. 13 Bottom, top, longitudinal side and end views of a linear multi-focal spot X-ray source for a 3D stationary gantry X-ray CT imaging system according to embodiments of the present specification are shown;

[0077] FIG. 14 A block diagram of a plurality of exemplary information, outputs or results derived based on processing of carcass scan image data generated using a dual plane 3D stationary gantry X-ray CT imaging system according to some embodiments of the present specification is shown;

[0078] FIG. 15 A workflow showing the use of a plurality of 3D X-ray computed tomography processes for improving slaughterhouse management and automation according to some embodiments of the present specification is shown;

[0079] FIG. 16A A workflow showing a semi-automated meat production process according to embodiments of the present specification is shown;

[0080] FIG. 16B A block diagram showing an augmented reality based system for cutting meat in a meat processing plant according to embodiments of the present specification is shown;

[0081] FIG. 16C A flowchart showing steps of an augmented reality based method for cutting meat in a meat processing plant according to embodiments of the present specification is shown;

[0082] FIG. 17 A flowchart showing steps of assigning a carcass ID to track the location, time and / or arrival of each carcass through a meat processing plant according to embodiments of the present specification is shown;

[0083] FIG. 18 is a flowchart illustrating steps of assigning a carcass ID for tracking location and / or time when raw meat or retail meat pieces are obtained from a carcass by a meat processing plant, according to embodiments of the present specification;

[0084] FIG. 19 is a flowchart illustrating steps of assigning a carcass ID for tracking location of a carcass / raw meat / retail meat pieces by a meat processing plant, according to embodiments of the present specification;

[0085] FIG. 20 illustrates a 3D scan image of a beef carcass providing information to enable automatic alignment, positioning, and cutting of the carcass, according to some embodiments of the present specification;

[0086] FIG. 21 illustrates a histogram analysis of first and second scan images of first and second beef samples, respectively, according to some embodiments of the present specification;

[0087] FIG. 22 illustrates a perspective view and block diagram of a multi-sensory imaging system, according to some embodiments of the present specification;

[0088] FIG. 23A illustrates a plurality of plots indicating selection of illumination sources for visible camera wavelengths as a function of scan rate (Hz), according to some embodiments of the present specification;

[0089] FIG. 23B illustrates a plurality of plots indicating selection of illumination sources for short wave infrared (SWIR) wavelengths as a function of scan rate (Hz), according to some embodiments of the present specification;

[0090] FIG. 24 illustrates an exemplary barcode scanned simultaneously using X-ray, visible light, and SWIR sensors, according to some embodiments of the present specification;

[0091] FIG. 25A illustrates MEXA X-ray image data for a center vertical overhead projector view, according to some embodiments of the present specification;

[0092] FIG. 25B illustrates MEXA X-ray image data for a side projector view, according to some embodiments of the present specification;

[0093] FIG. 26 illustrates RGB images and corresponding MEXA images of fresh and mature lamb viscera, according to some embodiments of the present specification;

[0094] FIG. 27First, second and third image data for synchronization of MEXA, visible light and SWIR sensors, respectively, are shown according to some embodiments of the present specification;

[0095] FIG. 28 A test pattern for checking synchronization of a hyperspectral camera is shown according to some embodiments of the present specification;

[0096] FIG. 29 A first image from a visible hyperspectral camera and a second image from a SWIR hyperspectral camera are shown according to some embodiments of the present specification;

[0097] FIG. 30A An RGB image of a beef liver and corresponding high and low energy X-ray scan images are shown according to some embodiments of the present specification;

[0098] FIG. 30B An RGB image and a corresponding SWIR hyperspectral image are shown according to some embodiments of the present specification;

[0099] FIG. 30C A SWIR hyperspectral image and corresponding spectral signals are shown according to some embodiments of the present specification;

[0100] FIG. 30D A visible hyperspectral image and corresponding spectral signals are shown according to some embodiments of the present specification;

[0101] FIG. 31 Steps for pre-processing a visible hyperspectral image are shown according to some embodiments of the present specification;

[0102] FIG. 32 is a block diagram of a deep learning network for disease screening and abnormal heat map generation according to some embodiments of the present specification;

[0103] FIG. 33A is a workflow for automatic anomaly detection from SWIR images using a k-clustering algorithm for anomaly detection according to some embodiments of the present specification;

[0104] FIG. 33B A plot indicating the sum of intensities from each SWIR band from a beef organ is shown according to some embodiments of the present specification;

[0105] FIG. 33C A normalization process of SWIR hyperspectral data of a beef liver is shown according to some embodiments of the present specification;

[0106] FIG. 33D PCA processing of a SWIR hyperspectral image from a beef liver is shown according to some embodiments of the present specification;

[0107] FIG. 33E is a plot indicating within sum of squared errors (WSS) used in k-means clustering method for anomaly detection in bovine liver, according to some embodiments of the present specification;

[0108] FIG. 33F shows clustering merging to detect anomaly in believers from SWIR hyperspectral data, according to some embodiments of the present specification;

[0109] FIG. 34A shows RGB and X-ray images (six images for six X-ray energies) and macroscopic findings of a kidney without macroscopic lesions, according to some embodiments of the present specification;

[0110] FIG. 34B shows RGB and X-ray images (six images for six X-ray energies) and macroscopic findings of two kidneys without macroscopic lesions, according to some embodiments of the present specification;

[0111] FIG. 34C shows RGB and X-ray images (six images for six X-ray energies) and macroscopic findings of two lungs, according to some embodiments of the present specification;

[0112] FIG. 34D shows pneumonia evidenced by discoloration and consolidation of the tissue based on postmortem examination of bovine lung;

[0113] FIG. 34E shows RGB and X-ray images (six images for six X-ray energies) and macroscopic findings of a lung, according to some embodiments of the present specification;

[0114] FIG. 34F shows no lesions as a result of postmortem examination of bovine lung, according to some embodiments of the present specification;

[0115] FIG. 34G shows RGB and X-ray images (six images for six X-ray energies) and macroscopic findings of a liver with multifocal discoloration, according to some embodiments of the present specification;

[0116] FIG. 34H shows first and second X-ray images of a liver obtained from absorption assay data, where data is obtained from low energy subtracted from high energy or simultaneously using low energy absorption assay data, according to some embodiments of the present specification;

[0117] FIG. 35A shows RGB and X-ray images (six images for six X-ray energies) of a liver sample, according to some embodiments of the present specification;

[0118] FIG. 35B Multiple images of a liver sample during postmortem examination are shown in accordance with some embodiments of the present specification;

[0119] FIG. 36A RGB and X-ray images (six images at six X-ray energies) of yet another liver sample are shown in accordance with some embodiments of the present specification;

[0120] FIG. 36B First and second images of first and second cysts during postmortem examination and histopathology of a liver are shown in accordance with some embodiments of the present specification, respectively;

[0121] FIG. 37A RGB and X-ray images (six images at six X-ray energies) of yet another liver sample are shown in accordance with some embodiments of the present specification;

[0122] FIG. 37B RGB images of a liver showing large nodules at postmortem examination are shown in accordance with some embodiments of the present specification;

[0123] FIG. 38A RGB and X-ray images (six images at six X-ray energies) of yet another liver sample are shown in accordance with some embodiments of the present specification;

[0124] FIG. 38B RGB images of a liver showing first abscess on the left lobe margin and second abscess on the right side of the bile duct at postmortem examination are shown in accordance with some embodiments of the present specification;

[0125] FIG. 39A RGB and X-ray images (six images at six X-ray energies) of yet another liver sample are shown in accordance with some embodiments of the present specification;

[0126] FIG. 39B RGB images of a liver showing discoloration, bile duct thickening, and flukes at postmortem examination are shown in accordance with some embodiments of the present specification;

[0127] FIG. 40A Result visualization of multiple bovine livers with abnormalities using deep learning algorithms is shown in accordance with some embodiments of the present specification;

[0128] FIG. 40B Multiple images of bovine livers with unsupervised PCA and k-means clustering showing pixels with dissimilar different characteristics are shown in accordance with some embodiments of the present specification;

[0129] FIG. 40C Spectral analysis of sampled pixel vectors is shown in accordance with some embodiments of the present specification;

[0130] FIG. 40D is a plot of pixel vectors in 3D space from SWIR data collected in a cow liver according to some embodiments of the present specification;

[0131] FIG. 41A shows multiple images of a cow liver from SWIR hyperspectral using PCA and k-means algorithm to identify regions with anomalies according to some embodiments of the present specification;

[0132] FIG. 41B shows spectral analysis of sampled pixel vectors from SWIR hyperspectral using PCA and k-means algorithm to identify regions with anomalies according to some embodiments of the present specification;

[0133] FIG. 41C is a plot of pixel vectors in 3D space from SWIR hyperspectral using PCA and k-means algorithm to identify regions with anomalies according to some embodiments of the present specification;

[0134] FIG. 42A shows RGB and labeled X-ray images of healthy lamb offal according to some embodiments of the present specification;

[0135] FIG. 42B shows differences in multi-energy X-ray intensity between three organ types of lamb offal, each labeled and with labeled intensity histograms of organs of interest according to some embodiments of the present specification;

[0136] FIG. 43A shows RGB images of a sheep lung showing evidence of CLA and six X-ray images taken at different energy levels according to some embodiments of the present specification;

[0137] FIG. 43B shows first, second, and third images of a sheep lung according to some embodiments of the present specification;

[0138] FIG. 43C shows RGB images of another sheep lung showing evidence of abscesses and six X-ray images taken at different energy levels according to some embodiments of the present specification;

[0139] FIG. 43D shows RGB images and corresponding X-ray images indicating multi-loculated abscesses in the right lung of a sheep according to some embodiments of the present specification;

[0140] FIG. 43E shows RGB images and corresponding X-ray images indicating multi-loculated abscesses in the right lung of a sheep according to some embodiments of the present specification; FIG. 43C and 43D intensity histograms of abscesses and healthy regions of a sheep lung according to some embodiments of the present specification;

[0141] FIG. 43F A set of intensity histograms of abscessed and healthy regions of a sheep lung are shown in accordance with some embodiments of the present specification FIG. 43C and 43D A set of intensity histograms of abscessed and healthy regions of a sheep lung are shown in accordance with some embodiments of the present specification

[0142] FIG. 44 RGB images and corresponding X-ray images of diseased sheep livers are shown in accordance with some embodiments of the present specification

[0143] FIG. 45 RGB images and corresponding X-ray images of healthy sheep livers are shown in accordance with some embodiments of the present specification

[0144] FIG. 46A RGB images and X-ray images of damaged sheep livers are shown in accordance with some embodiments of the present specification

[0145] FIG. 46B Labeled X-ray images of damaged sheep livers are shown in accordance with some embodiments of the present specification

[0146] FIG. 47 Multiple energy X-ray scans of lamb lungs at different gray scale contrasts are shown in accordance with some embodiments of the present specification

[0147] FIG. 48 Multiple images of cheese-like glands in sheep meat are shown in accordance with some embodiments of the present specification

[0148] FIG. 49 Visible and SWIR surface reflectance hyperspectral intensity spectra of mixed sheep and beef organs are shown in accordance with some embodiments of the present specification

[0149] FIG. 50 First and second plots indicating the accuracy of automatic organ classification using two classification models for visible, SWIR, and their combination are shown in accordance with some embodiments of the present specification

[0150] FIG. 51 Visible and shortwave infrared spectra of organ types for diseased (red) and healthy (blue) organs are shown in accordance with some embodiments of the present specification

[0151] FIG. 52 First and second plots indicating the accuracy of machine learning models (PLS-DA - Partial Least Squares Discriminant Analysis; and RF - Random Forest) for a hyperspectral sensor to differentiate multiple sheep organs by disease status are shown in accordance with some embodiments of the present specification

[0152] FIG. 53Visible and shortwave infrared spectra for differentiating lean beef by grass or grain fed are shown according to some embodiments of the present specification;

[0153] FIG. 54 Images of raw meat at three different energy levels (low, medium, and high) using two views (overhead and side view) and the sum signal of all six energy levels are shown according to some embodiments of the present specification;

[0154] FIG. 55 X-ray and shortwave infrared hyperspectral images of beef steaks at low, medium, and high energy and wavelengths are shown according to some embodiments of the present specification;

[0155] FIG. 56 MEXA lamb images acquired in two separate scans are shown according to some embodiments of the present specification, and each image is shown with different grayscale contrast;

[0156] FIG. 57 Average visible and shortwave infrared reflectance spectra from each of liver, heart, lung, kidney of sheep and beef are shown according to some embodiments of the present specification;

[0157] FIG. 58A Model indices for classifying organs from sheep and cattle by species and type using visible (VIS) hyperspectral sensors by partial least squares discriminant analysis (PLS-DA), linear discriminant analysis (LDA), and random forest (RF) are shown according to some embodiments of the present specification;

[0158] FIG. 58B Model indices for classifying organs from sheep and cattle by species and type using shortwave infrared (SWIR) hyperspectral sensors by partial least squares discriminant analysis (PLS-DA), linear discriminant analysis (LDA), and random forest (RF) are shown according to some embodiments of the present specification;

[0159] FIG. 58C Model indices for classifying organs from sheep and cattle by species and type using a combination of visible and shortwave infrared hyperspectral sensors (COMB) with partial least squares discriminant analysis (PLS-DA), linear discriminant analysis (LDA), and random forest (RF) are shown according to some embodiments of the present specification;

[0160] FIG. 59 Spectra after undergoing different smoothing processes are shown according to some embodiments of the present specification;

[0161] FIG. 60A photograph of an untreated sheep lung, a photograph of the same sheep lung after dissection, and a multi-energy X-ray (MEXA) image of the untreated sheep lung showing caseous lymphadenitis (CLA) lesions are shown in accordance with some embodiments of the present specification;

[0162] FIG. 61 A plurality of RGB and X-ray images of a sheep kidney are shown in accordance with some embodiments of the present specification;

[0163] FIG. 62 A fully assembled sensor module is shown in accordance with some embodiments of the present specification;

[0164] FIG. 63 A line graph showing various stages of performing a scan and subsequent image analysis to generate data indicative of the quality of meat ("meat grade") is shown in accordance with some embodiments of the present specification; and

[0165] FIG. 64 An intelligent meat production system is shown in accordance with some embodiments of the present specification. DETAILED DESCRIPTION

[0166] In one embodiment, the present specification describes the use of a three-dimensional (3D) fixed gantry X-ray computed tomography system to scan animals and / or livestock to enable improved management of animal breeding processes, functions, or events. The resulting scan information, particularly when generated or applied at different stages during the development of an animal, can be used to drive breeding practices that promote the development of individual animals and the overall development of one or more herds. When such breeding practices are driven based on scan information of animals and herds, the result is improved assessment of animals, reduced breeding costs, and at the same time, improved quality of feed or consumption of each animal, thereby improving farm economics and consumer satisfaction.

[0167] The present specification also discloses the use of a 3D fixed gantry X-ray computed tomography system for carcass screening and improved slaughterhouse production planning, execution, and automation. In various embodiments, the use of scanning technology supports high-throughput, automated meat processing lines with reduced manual labor, objectively measured product quality, and improved food safety standards.

[0168] In an embodiment, the present specification discloses the use of 3D X-ray inspection to generate images of the whole carcass and parts of the carcass during the stages of dissection, final product preparation and packaging of the carcass. The generated images are used to derive indicators about, but not limited to, eating quality, animal health, lean yield (amount of meat, fat and bone present in the carcass), carcass value and 3D carcass structure. The derived indicators also drive efficiency in the abattoir through process automation, accurate production planning, providing accurate eating quality through each muscle within the carcass, culling of unhealthy carcasses from the food chain, payment based on carcass value and not just weight, quality control measures to ensure integrity of the safe product to the consumer and providing supply chain assurance to customers to verify the supply chain of the meat they are purchasing.

[0169] In an embodiment, the present specification also discloses a method for automating and improving meat production efficiency in a meat processing plant. In an embodiment, the present specification provides the use of network connected 2D and 3D X-ray imaging modalities and visible and handheld sensors such as, but not limited to, RFID and barcode readers in a meat production plant. The networked imaging and screening modalities are used to generate data that is processed in real time by specific algorithms in conjunction with production requirement information stored in a database coupled to the network to generate individualized carcass driven optimization of the entire meat production process. In an embodiment, the present specification provides a method for automated and robotic cutting of carcasses.

[0170] In various embodiments, a computing device includes an input / output controller, at least one communication interface, and system memory. The system memory includes at least one random access memory (RAM) and at least one read-only memory (ROM). These elements are in communication with a central processing unit (CPU) to enable operation of the computing device.

[0171] In various embodiments, the computing device can be a conventional stand-alone computer, or alternatively, the functionality of the computing device can be distributed across a network of multiple computer systems and architectures. In some embodiments, execution of a plurality of programming instructions or code sequences stored in one or more non-volatile memories enables or causes the CPU of the computing device to perform various functions and processes, such as, for example, performing tomographic image reconstruction for display on a screen. In alternative embodiments, hardwired circuitry can be used in place of, or in combination with, software instructions to implement the processes of the systems and methods described in this application. Thus, the described systems and methods are not limited to any specific combination of hardware and software.

[0172] The term "pass," "passes," "passes through," "passing through," or "traverses" as used in the present disclosure encompasses all forms of active and passive animal movement, including walking, being carried in a container, being suspended on a structure, or being transported / driven using a conveyor.

[0173] The term "meat" as used in the present disclosure can mean animal meat for food. In some embodiments, "meat" can mean meat including bones and edible parts but not inedible parts. Edible parts can include prime cuts, choice cuts, edible offal (head or head meat, tongue, brains, heart, liver, spleen, stomach or tripe, and in some cases, other parts such as feet, throat, and lungs). Inedible parts can include hides and skins (except in the case of pigs), and hooves and stomach contents.

[0174] The term "K-means clustering" as used in the present disclosure can mean an unsupervised learning algorithm. Unlike supervised learning, this type of clustering does not have labeled data. K-means clustering is used to perform a division of objects into clusters that share similarities and are dissimilar to objects belonging to another cluster.

[0175] The present specification relates to a number of embodiments. The following disclosure is provided in order to enable ordinary skilled persons to practice the application. Language used in this specification should not be used to interpret the general scope of any one of the embodiments. The claims hereinafter are in no way limited by the language used in this specification. The general principles defined herein can be applied to other embodiments and applications without departing from the spirit and scope of the present application. In addition, the terminology and phraseology used should not be treated as limiting. Accordingly, the present application should be given the broadest scope consistent with the principles and features disclosed. For the sake of clarity, details relating to technical material known in the art to which the application relates are not described in detail herein to avoid obscuring the application unnecessarily.

[0176] In the specification and claims of this application, each of the words "comprise," "comprises," and "comprising," and forms of the word "have," "has," and "having," whether used in the context of a feature, structure, or process, does not necessarily denote the presence of the item or items described in an exclusive manner. It is noted herein that any feature or component described in association with a particular embodiment can be used and implemented with any other embodiment unless otherwise explicitly stated.

[0177] As used herein, the indefinite articles "a" and "an" mean "at least one" or "one or more" unless the context clearly indicates otherwise.

[0178] FIG. 1Aand 1B First and second side cross-sectional views of a 3D stationary gantry X-ray CT imaging system 100 (also referred to as a real-time tomography (RTT) system) configured to scan cattle in accordance with some embodiments of the present specification are shown. Reference is made to FIG. 1A and 1B The system 100 is deployed, for example, in an animal farm, to scan cattle in real-time as the animals pass through a scan region, area or aperture 150 of the system 100. FIG. 1A The first side cross-sectional view is in a direction perpendicular to the direction of motion of the animals as they pass through the scan region, area or aperture 150, while FIG. 1B The second side cross-sectional view is in a direction parallel to the direction of motion of the animals as they pass through the scan region, area or aperture 150.

[0179] In some embodiments, the first ramp 105 is adapted to enable the animals to pass onto a horizontal platform 106 located in the scan region, area or aperture 150, to eventually pass down using the second ramp 107. In other words, the animals enter the scan region, area or aperture 150 from the left side portion of the figure and exit the scan region, area or aperture 150 on the right side of the figure.

[0180] In some embodiments, the system 100 is enclosed within a food-safe, environmentally-protected enclosure 115 manufactured using materials such as, but not limited to, stainless steel and / or plastic. In some embodiments, the system 100 is surrounded by at least one radiation shielding enclosure. A control room is provided for one or more system operators to review the performance of the system 100 on one or more inspection workstations in data communication with the system 100. In various embodiments, the one or more inspection workstations are computing devices.

[0181] In some embodiments, the system 100 is configured for dual plane scanning and includes a first plurality of linear multi-focal X-ray sources 145a and associated first detector 155a arrays positioned or deployed around the scan region, area or aperture 150 to scan the animals in a first imaging plane 142, and a second plurality of linear multi-focal X-ray sources 145b and associated second detector 155b arrays also positioned or deployed around the scan region, area or aperture 150 to scan the animals in a second imaging plane 143. Thus, the system 100 is configured in two separate planes 142, 143, where the data is combined together at the one or more inspection workstations to create a single reconstructed volume.

[0182] In some embodiments, the scan area, region, or aperture 150 has a substantially rectangular geometry or shape. In some embodiments, a value representing the entire width of the scan area 150 is within 85% of a value representing the entire height of the scan area 150. In some embodiments, the scan area, region, or aperture 150 has dimensions of 1500 mm (width) x 1800 mm (height). In alternative embodiments, the scan area, region, or aperture 150 has a substantially square or polygonal geometry or shape. In some embodiments, the first imaging plane 142 includes, for example, four linear multi-focal spot X-ray sources 145a that are separated from one another and positioned around or along a perimeter of the scan area, region, or aperture 150. In some embodiments, the second imaging plane 143 includes, for example, four linear multi-focal spot X-ray sources 145b that are separated from one another and positioned around or along a perimeter of the scan area, region, or aperture 150.

[0183] In some embodiments, as shown in FIG. 1, the linear multi-focal spot X-ray sources 145a, 145b are arranged in a pattern that is substantially uniform around the perimeter of the scan area, region, or aperture 150. In some embodiments, the linear multi-focal spot X-ray sources 145a, 145b are arranged in a pattern that is substantially uniform around the perimeter of the scan area, region, or aperture 150. FIG. 1B In some embodiments, as shown in FIG. 1, the linear multi-focal spot X-ray sources 145b (in the second imaging plane 143) are arranged or positioned to fill the gaps separating the linear multi-focal spot X-ray sources 145a (in the first imaging plane 142). Thus, the first and second linear multi-focal spot X-ray sources 145a, 145b are dispersed in their respective first and second imaging planes 142, 143 to produce a substantially uniform sampling distribution around the perimeter of the scan area, region, or aperture 150. In embodiments, it is preferable to maintain a relatively thin X-ray window around the X-ray detector regions 155a, 155b. In some embodiments, the horizontal top as well as the first and second vertical sides use 2 mm to 5 mm thick aluminum. In the floor (horizontal platform 106), a thicker plate is required, ranging from 6 mm to 10 mm of aluminum, to prevent deformation under the load of the animal’s hooves. Although this thick window reduces the total X-ray flux in the scan area 150, this also reduces the low-energy X-ray dose, which helps to reduce the radiation dose to the animal to tolerable levels.

[0184] In some embodiments, the first imaging plane 142 and the second imaging plane 143 are arranged along a direction perpendicular to the direction of movement of the animal on the horizontal platform 106 and pass through the examination area, region, or aperture 150 during scanning. In embodiments, the first and second imaging planes 142, 143 are separated from each other by a distance “d” along the direction of movement of the animal during scanning, ranging from 100 mm to 2000 mm. Therefore, an array of a first plurality of linear multifocal X-ray sources 145a and associated first detectors 155a is deployed in the first imaging plane 142, while an array of a second plurality of linear multifocal X-ray sources 145b and associated second detectors 155b is deployed in the second imaging plane 143.

[0185] In one embodiment, a first plurality of linear multifocal X-ray sources 145a are offset or shifted by a distance d1 from the array of associated first detectors 155a in the first imaging plane 142, while a second plurality of linear multifocal X-ray sources are offset or shifted by a distance d2 from the array of associated second detectors 155b in the second imaging plane 143. In some embodiments, d1 is equal to d2. In various embodiments, the distances d1 and d2 range from 2 mm to 20 mm. It should be understood that the first and second detector arrays 155a, 155b are offset from the respective planes of the first and second X-ray sources 145a, 145b such that X-rays from a source on one side of the scanning region, region, or aperture 150 pass over the detector array adjacent to the source but interact in the detector array opposite the source on the other side of the scanning region, region, or aperture 150.

[0186] In one embodiment, the 3D fixed-stage X-ray CT imaging system 100 includes a series of X-ray tubes operating in series, rather than FIG. 1A and 1B The multifocal X-ray source shown is an X-ray source consisting of multiple X-ray tubes and not multiple source points.

[0187] In some embodiments, such as FIG. 1CAs shown, the 3D stationary gantry X-ray CT imaging system 180 includes one or more X-ray tubes 181 configured in a substantially circular arrangement about a scanner axis, with each X-ray tube 181 containing an X-ray source having one or more X-ray source points 182. In an embodiment, the emission of X-rays from each source point of each X-ray tube 181 is controlled by a switching circuit 184, with a separate switching circuit for each X-ray source point. The switching circuits of each tube 181 together form part of the control circuit 186 of that tube. A controller 188 controls the operation of all the individual switching circuits 184. In an embodiment, the controller 188 is a workstation disposed in a control room for one or more system operators to check the performance of the system 180. In an embodiment, the control switching circuits 184 to emit in a predetermined sequence such that in each of a series of activation periods, a fan-shaped X-ray beam from one or more activated source points is propagated through an animal 185 passing through the center of the arrangement of X-ray tubes 181 on a ramp 187. Thus, in an embodiment, the controller 188 is configured to control the activation and deactivation of each source point within each of the first linear multi-focus X-ray source 145a and the second linear multi-focus X-ray source 145b.

[0188] It will be appreciated that in various embodiments, the controller 188 implements a plurality of instructions or programming code to a) ensure that the switching circuits 184 are controlled to fire in a predetermined sequence, and b) perform processing steps corresponding to the various workflows and methods described in this specification.

[0189] With reference to FIG. 1A and 1BDuring a scan operation, as the animal passes through the scan region, area, or aperture 150, each X-ray source point within the respective multi-focus X-ray source (145a, 145b) is turned on in sequence and projection data through the animal is collected for that one source point as the animal passes. When the exposure is complete, a different X-ray source point is turned on, e.g., within a different multi-focus X-ray source in the system 100, to create the next X-ray projection. The scan process continues until all X-ray sources have been fired in a sequence configured to optimize the X-ray image quality of the reconstruction. In some embodiments, it is preferable to activate non-adjacent sources in the next portion of the scan sequence. In practice, it is preferable to activate sources at approximately 20 to 90 degrees away from the currently activated source point. Thus, the respective X-ray source points within the linear multi-focus X-ray sources 145a, 145b within each plane 142, 143 are activated sequentially so that generally at least one X-ray beam is active at all times. In some embodiments, each source point within a first linear multi-focus X-ray source is turned on and subsequently, after each source point within the first linear multi-focus X-ray source has passed, each source point within a second linear multi-focus X-ray source is turned on. In some embodiments, one source point within a first linear multi-focus X-ray source is turned on, then one source point within a second linear multi-focus X-ray source is turned on, thus alternating back and forth (between the first and second linear multi-focus X-ray sources) until all source points are activated.

[0190] FIG. 2 A bottom view, top view, longitudinal side view, and end view 205a, 205b, 205c, 205d of a linear multi-focus X-ray source 245 for a 3D stationary gantry X-ray CT imaging system according to some embodiments of the present specification is shown. Referring now to the views 205a, 205b, 205c, 205d simultaneously, the source 245 includes a plurality of electron guns, cathodes, or source / emission points 210 and an anode 215 housed in a vacuum tube or enclosure 220. In some embodiments, the source 245 includes 100 X-ray emission points 210 at 10 mm intervals on a 1000 mm long active anode 215.

[0191] In some embodiments, the first support 222a, the second support 222b, and the third support 222c are deployed to support the anode 215 along the longitudinal axis. The first support 222a and the second support 222b are deployed at the two ends, while the third support 222c is deployed at the center of the anode 215. In some embodiments, the first support 222a and the second support 222b also serve as coolant feedthrough units, while the third support 222c enables high voltage feedthrough. In some embodiments, the anode 215 supports operating tube voltages in the range of 100 kV to 300 kV. In some embodiments, each electron gun, cathode, or source / emission point 210 emits tube currents in the range of 1 mA to 500 mA, depending on the animal thickness and the inspection region, aperture, or size - the larger the inspection aperture and the thicker the animal, the higher the tube current required.

[0192] For scanning of livestock animals (e.g., cows and buffalos), a suitable optimization is a 225 kV tube voltage and a 20 mA beam current, for a total X-ray beam power of 4.5 kW. In combination with a minimum 3 mm aluminum tube filtration, this results in a dose to the animal in the range of 2 μSv (microsievert) to 20 μSv, and in embodiments, about 10 μSv. To put this in context, a typical individual dose to a human is 2 mSv / year (millisievert / year) due to naturally occurring background radiation. An exposure of 10 μSv corresponds to 0.5% of a year of natural background radiation or about 2 days of natural background radiation.

[0193] In some embodiments, each electron gun 210 is configured to illuminate a region or lesion on the anode 215 that is in the range of 0.5 mm to 3.0 mm diameter. The specific size of the lesion is chosen to maximize image quality and minimize heating of the anode 215 during X-ray exposure. The higher the product of tube current and tube voltage, the larger the lesion is typically designed to be.

[0194] FIG. 3A A first side view 301a, a second side view 301b, and a top view 301c of a single-plane stationary gantry X-ray computed tomography system 300 configured to scan sheep, pigs, and goats are shown in accordance with some embodiments of the present specification, while FIG. 3B The first side view 301a is also shown. Referring to FIG. 3A and 3B The system 300 is deployed, for example, in an animal farm, to scan livestock animals in real-time as the animals pass through a scan region, area, aperture, or tunnel 350 of the system 300. The scan region, area, aperture, or tunnel 350 is smaller (as compared to the scan systems of FIG. 1A , 1B The first side view 301a (as well as the second side view 301b and the top view 301c) of the system 300 is shown in accordance with some embodiments of the present specification. The system 300 is configured to scan animals such as sheep, pigs, and goats. FIG. 3A , 3BThe direction of the first side view 301a is parallel to the direction of motion of the animal as it passes through the scanning region, area, or aperture 350, while the direction of the second side view 301b is perpendicular to the direction of motion of the animal as it passes through the scanning region, area, or aperture 350.

[0195] In some embodiments, the first ramp 305 is adapted to enable the animal to pass onto a horizontal platform 306 located in the scanning region, area, or aperture 350, and eventually pass down using the second ramp 307. In other words, the animal enters the scanning region, area, or aperture 350 from the left in view 301b and exits the scanning region, area, or aperture 350 on the right in 301b.

[0196] In some embodiments, the system 300 is enclosed in a food-safe, environmentally-protected enclosure 315 made using materials such as, but not limited to, stainless steel, aluminum, and / or plastic. In some embodiments, the system 100 is surrounded by at least one radiation- shielding enclosure. In some embodiments, the system 300 has a multi-focal X-ray source 345 disposed in a plane surrounding the scanning region, area, or aperture 350. The source 345 includes a plurality of X-ray source emission points, electron guns, or cathodes 346 (also referred to as an electron gun array) surrounding an anode 347. The plurality of X-ray source emission points 346 and the anode 347 are enclosed in a vacuum envelope or tube 310. In some embodiments, the source 345 includes 200 to 500 X-ray source emission points 346 arranged around a single anode 347 held at a positive high voltage relative to the corresponding electron gun array 346. In some embodiments, the tube voltage is held in the range of 120 kV to 200 kV, and the tube current is in the range of 1 mA to 20 mA. In an embodiment, a single source 345 includes a plurality of X-ray source emission points for scanning small animals (e.g., sheep, pigs, and goats); while a plurality of linear multi-focal X-ray sources (e.g., as shown in FIG. 1A 、 1B shown) disposed around a scanning tunnel are used for scanning larger animals, such as cattle. The preferred operating point for scanning small animals (e.g., sheep, pigs, and goats) is 160 kV, 4 mA, corresponding to a total X-ray beam power of 640 W. In an embodiment, this results in a dose of about 2 μSv to 20 μSv per scan of an animal. In an embodiment, due to the smaller size of the scanning region, area, aperture, or tunnel 350 (as compared to the scanning region 150 for beef scanning), the dose per animal scan is about 10 μSv. FIG. 1A 、 1B

[0197] ​An array of detectors 355 is also positioned or disposed around the scan region, area, or aperture 350 to scan the animal as it passes through the scan region, area, or aperture 350. In some embodiments, the scan region, area, or aperture 350 has a substantially rectangular geometry or shape. In some embodiments, the scan region, area, or aperture 350 has a substantially square or polygonal geometry or shape. In some embodiments, the scan region, area, or aperture 350 has a width ranging from 400 mm to 800 mm and a height ranging from 600 mm to 1000 mm in height. In one embodiment, as shown in FIG. 3, the scan region 350 has a width of 600 mm and a height of 800 mm. In some embodiments, the array of detectors 355 is offset or displaced from the X-ray source 345 by a predetermined distance such that X-rays from the source pass over the detector array adjacent to the source but interact in the detector array on the other side of the scan region, area, or aperture 350 opposite the source. In various embodiments, the predetermined distance ranges from 2 mm to 20 mm.

[0198] A control room can be provided for one or more system operators to review the performance of the system 300 on one or more inspection workstations in data communication with the system 300. Alternatively, mobile computing devices can be used to inspect image data and control system operation. In various embodiments, the one or more inspection workstations are computing devices. At least one controller positioned within the one or more inspection workstations is configured to control activation and deactivation of each of the plurality of X-ray source emission points.

[0199] It will be appreciated that, in various embodiments, the controller implements a plurality of instructions or programming code to a) ensure that the plurality of X-ray source emission points are controlled to emit in a predetermined order, and b) perform process steps corresponding to various workflows and methods described in this specification.

[0200] During a scan operation, each X-ray source point within the multi-focus X-ray source is turned on in sequence and wherein at least a portion of the X-rays pass through the animal and the resulting projection data is collected for that one source point. When the exposure is complete, a different X-ray source point is turned on, e.g., within a different multi-focus X-ray source (in embodiments employing multiple linear multi-focus X-ray sources) to create the next X-ray projection. The scanning process continues until all X-ray sources have been fired / activated in a sequence configured to optimize X-ray image quality for reconstruction. In some embodiments, it is preferable to activate non-adjacent sources in the next portion of the scan sequence. In embodiments, it is preferable to activate sources positioned about 20 to 90 degrees away from the currently activated source point.

[0201] In embodiments employing multiple linear multi-focus x-ray sources, each source point within the first linear multi-focus x-ray source is turned on, and then (only after each source point within the first linear multi-focus x-ray source has been passed) each source point within the second linear multi-focus x-ray source is turned on. In some embodiments employing multiple linear multi-focus x-ray sources, one source point within the first linear multi-focus x-ray source is turned on, and subsequently, one source point within the second linear multi-focus x-ray source is turned on, thus alternating back and forth (between the first and second linear multi-focus x-ray sources) until all source points have been activated.

[0202] In embodiments, the system 300 includes a series of x-ray source tubes operating in series, rather than a multi-focus x-ray source 345. In other words, the x-ray source is multiple x-ray tubes and does not contain multiple source points.

[0203] When passing through the scan region, area, or aperture 350, the animal can move at an uncontrolled speed, especially in the case of walking and not ambulating, and can also move from side to side. Therefore, motion correction of the x-ray projection data is needed before the back-projection algorithm is implemented or executed. In some embodiments, this can be achieved directly from the x-ray projection data itself by analyzing each set of data and forward projecting through the partially reconstructed x-ray data to see where the new projection is most likely to come from. However, this is computationally expensive, so in some embodiments, it is advantageous to use an auxiliary sensor system to monitor the surface profile of the animal and thus directly measure motion. This information can then be used to determine where each new x-ray projection should be back-projected into the 3D reconstructed image volume.

[0204] Various types of 3D (three-dimensional) surface sensing technology can be used, including, for example, point cloud optics and radar imaging sensors. FIG. 3B A radar imaging or inspection system 360 is shown, which includes radar transceivers or transceiver modules, each including multiple receiver (Rx) elements and transmitter (Tx) elements, which together operate to form a tomographic image of the animal as it passes through the scan region, area, or aperture 350.

[0205] In some embodiments, the radar imaging or inspection system 360 operates in a stepped frequency continuous wave radar scan sequence or pattern 400, as FIG. 4As shown. In radar scan sequence or mode 400, each Tx element is held at a discrete set of frequencies (in some embodiments, in the range of 5 GHz to 50 GHz, with 10 to 500 steps, depending on the required range resolution), each step lasting a fixed time period (in some embodiments, 1 to 100 μs) to provide time for the Rx elements to calculate the phase and amplitude relative to the input signal. Each Tx element is activated individually, and all used Rx elements listen in parallel to the radar signal from the individual Tx elements to form a tomographic scan dataset, which is then reconstructed to form a full-surface image across the animal. In one embodiment, for example, with Rx and Tx elements spaced 15 mm apart over an 800 mm length, 50 Rx and Tx transceiver elements are present on each of the first side 350a and the second side 350b of the scan area, region, or aperture 350. For example, assuming an imaging frame rate of 50 Hz (20 ms), each Tx element will be active for 400 μs per ramp period. With an output frequency range of 10 GHz to 40 GHz and a step size of 0.5 GHz (60 steps in total), the dwell time at each step is, for example, 6.5 μs. In the embodiment, the selection of which Tx element to activate at any given time is based on the goal of maximizing the quality of the reconstructed image.

[0206] FIG. 5 This is a block diagram of a radar imaging system 500 for determining body shape and movement using an X-ray computed tomography system to correct animal movement, according to some embodiments of this specification. In some embodiments, system 500 is an ultra-wideband radar system. In some embodiments, system 500 includes a field-programmable gate array (FPGA) 505 to generate a basic step-frequency continuous wave signal, which is then frequency-multiplied and amplified (at power amplifier element 525) to each of a plurality of Tx transceiver elements 510. FPGA 505 is coupled to power amplifier element 525 via frequency multiplier block 526, which converts a low-frequency clock (e.g., 100 MHz) of FPGA 505 to a higher output frequency clock (e.g., 50 GHz) of power amplifier element 525. Ramp generator circuitry 535, communicating with FPGA 505, generates a linearly rising or falling output signal relative to time, thereby producing a sawtooth waveform. Waveform synthesis element 540 receives a linear rising or falling waveform signal from ramp generator circuit 535 to output a basic step-frequency continuous wave signal for application to each of the plurality of Tx transceiver elements 510. Clock generator and synchronization circuit 545 generate timing signals for synchronous operation of system 500.

[0207] In parallel, the outputs from all Rx transceiver elements 515 are mixed with the Tx frequency at Rx amplifier and mixer elements 530 to generate lower frequency signals that can be measured by an analog-to-digital converter (ADC) 520 and transferred to the internal memory of the FPGA 505. Additionally, signal processing can be done in the FPGA 505 to reduce data bandwidth, or alternatively, all data can be transferred to a host computing device over a high-speed interface for processing.

[0208] In some embodiments, the Tx and Rx transceiver elements 510, 515 employ circular polarization so that the reflected waves return with opposite polarization to the transmitted waves. This reduces cross-talk between the Tx and Rx transceiver elements 510, 515, simplifying the analog front-end design as well as the algorithmic complexity in image reconstruction.

[0209] FIG. 6 An exemplary arrangement of multiple transmitter (Tx) and receiver (Rx) elements of a radar imaging or inspection system 600 according to some embodiments of the present specification is shown, which is deployed for determining body shape and motion by an x-ray computed tomography system to correct for motion of an animal. The figure shows a view 601 along a direction parallel to the direction of motion of the animal through a scan region, area, or bore 650. The radar imaging or inspection system 600 includes an array 605 of radar Tx and Rx elements disposed on first and second vertical sides 650a, 650b of the scan region, area, or bore 650.

[0210] Another view 602 along a direction perpendicular to the direction of motion of the animal through the scan region, area, or bore 650 shows multiple radar transceiver or transceiver modules 610, which can also be referred to as “cards” in some embodiments. Each transceiver 610 includes multiple Tx and Rx elements (or analog circuits) 612, 614. In some embodiments, each transceiver 610 includes 8 Rx and 8 Tx elements 612, 614. In some embodiments, the Rx elements 614 are offset from the Tx elements 612 in the vertical direction by half an element spacing.

[0211] In some embodiments, the transmitter and receiver elements or analog circuits 612, 614 with ADCs (analog-to-digital circuits) are soldered to the same PCB (printed circuit board) as the antenna structure with the overall FPGA for system control and data acquisition. Each transceiver 610 also includes a data transfer connector 616 and readout control circuit 618. Ribbon cables are used to transfer signals from one card to the next to allow flexibility in the overall system configuration.

[0212] According to some embodiments, each 3D X-ray computed tomography system of the present specification can be housed in a container located on a farm. In use, the doors of the container entry and exit ends can be opened, the X-ray system activated, and scanning performed by driving the animals from the container entry side to the container exit side. In some embodiments, quantitative information from the X-ray scans is tied back to individual animals by coordinating RFID (radio frequency identification) tags or other animal specific IDs with the X-ray image data to assist in the overall farm process as well as food supply chain integrity processes. In embodiments, the containerized 3D X-ray computed tomography systems can be permanently installed at a farm or can be transported from one location to another as needed using a truck or trailer to service multiple farms.

[0213] According to some embodiments, the 3D X-ray computed tomography systems of the present specification can be supported on a mobile, road suitable scanning platform such as a truck, van and / or trailer. This enables the system to be transported on public and private roads to a desired farm scanning location, perform the necessary scanning, and then drive away to another farm where the scanning process can be repeated.

[0214] It should be noted that in alternative embodiments, 3D high resolution imaging methods such as, for example, magnetic resonance imaging can be used in place of X-ray computed tomography. Additionally, in various alternative embodiments, rotating gantry and / or single plane, biplane and multiplane fixed gantry X-ray computed tomography methods can be used interchangeably.

[0215] FIG. 7 is a block diagram of a number of exemplary information, outputs or results derived from processing or analysis of animal scan image data generated using a 3D fixed gantry X-ray CT imaging system according to some embodiments of the present specification. In embodiments, a controller in data communication with the 3D fixed gantry X-ray CT imaging system implements a number of instructions or programming code to receive the 3D scan image data, process or analyze the scan image data and generate various outputs or results such as effective Z, density information, 3D structure of the animal, calculated lean yield, intramuscular fat analysis, intermuscular fat volume, ratio of intramuscular fat (marbling) to tissue, absolute and relative size of individual organs, muscle volume, number of ribs, and presence or absence of cysts, tumors, pleurisy and foreign objects.

[0216] According to aspects of this specification, 3D scan image data of animals provides effective Z (atomic number) and density information (box 705), resulting in insights related to the animal's 3D structure, including bone structure, the size of each muscle, and the location and amount of fat (box 706). This enables farmers to optimize multiple breeding processes (box 708), such as calculating lean meat yield, thereby determining how best to optimize herd progression plans, including how much exercise, feed, feed supplements, and water to include in the animal program.

[0217] In some embodiments, 3D scan image data of the animal is analyzed to deliver objective indicators or measurements of all muscle groups within the animal on an individual basis to determine feed quality (box 710). In embodiments, indicators or measurements are determined by analyzing intramuscular fat (marbled fat) and intermuscular fat. Intermuscular fat is known to be the fat surrounding the muscle and is typically located between the animal's muscle and skin. In embodiments, indicators or measurements are determined by analyzing the ratio of intramuscular fat (marbled fat) to tissue. Farmers can use this data to plan improvements in overall feed quality and / or to the quality of selected muscle groups in the animal's highest-value parts – resulting in a higher selling price or valuation for the animal (box 712).

[0218] In some embodiments, further analysis of the 3D image data provides indicators, measurements, or information about animal health (box 715), such as, for example, the absolute and relative size of individual organs (such as kidneys, liver, heart, and lungs), the presence or absence of cysts and tumors, the presence of chronic conditions (such as pleurisy), and the presence of foreign objects that could lead to infection (such as barbed wire and needles). This shared information about animal health leads to improvements in overall quality control in food safety (box 717).

[0219] FIG. 8 This describes a workflow 800 illustrating the use of multiple 3D X-ray computed tomography (CT) scans during various events related to livestock farming, according to some embodiments of this specification. According to aspects of this specification, workflow 800 illustrates the entire lifecycle of an animal and the entire herd, from genetic selection to customer delivery, and multiple scan points throughout the lifecycle where 3D X-ray computed tomography scans on the farm may be beneficial. In embodiments, a controller communicating with the 3D X-ray computed tomography scanner executes multiple instructions or programming code to receive 3D scan image data, process or analyze the scan image data, and generate various outputs or results.

[0220] Blocks 802, 804, 806, and 808 represent functions / events related to genetic selection, input semen, conception, and birth of animals raised on a farm, respectively. At step 810, the animal is scanned initially / early in life, such as within 0-36 hours of birth, and in some cases longer, or before the animal reaches 6 months of age, using a 3D X-ray computed tomography system such as those described with reference to FIG. 1A , FIG. 1B , FIG. 3A , FIG. 3B and FIG. 6 The data from this initial / early scan is used to identify abnormalities and also to check for predetermined genetic characteristics. For example, most lambs are born with 8 ribs, but some are born with 7 ribs, and some are born with 9 ribs. It is helpful to know how many ribs a lamb has at an early stage, as the ultimate value of the animal can depend on such information. The identified abnormalities and predetermined genetic characteristics are recorded in at least one database.

[0221] In embodiments, the animal is 3D X-ray computed tomography scanned during various stages of development. For example, in embodiments, at a first stage of development, the animal can be in a first age range, starting at a first start date and ending at a first end date. In embodiments, the first stage of development corresponds to an early stage. In embodiments, at a second stage of development, the animal can be in a second age range, starting at a second start date and ending at a second end date. In embodiments, the second stage of development corresponds to a middle stage. In embodiments, at a third stage of development, the animal can be in a third age range, starting at a third start date and ending at a third end date. In embodiments, the third stage of development corresponds to a late stage. In embodiments, at a fourth stage of development, the animal can be in a fourth age range, starting at a fourth start date and ending at a fourth end date.

[0222] In embodiments, the first start date corresponds to a birth date of the animal, and precedes each of the first end date, the second start date, the second end date, the third start date, the third end date, the fourth start date, and the fourth end date.

[0223] In embodiments, the first end date succeeds each of the first start date and precedes each of the second start date, the second end date, the third start date, the third end date, the fourth start date, and the fourth end date.

[0224] In embodiments, the second start date succeeds each of the first start date and the first end date and precedes each of the second end date, the third start date, the third end date, the fourth start date, and the fourth end date.

[0225] In embodiments, the second end date is after each of the first start date, the first end date, and the second start date and before each of the third start date, the third end date, the fourth start date, and the fourth end date.

[0226] In embodiments, the third start date is after each of the first start date, the first end date, the second start date, and the second end date and before each of the third end date, the fourth start date, and the fourth end date.

[0227] In embodiments, the third end date is after each of the first start date, the first end date, the second start date, the second end date, and the third start date and before each of the fourth start date and the fourth end date.

[0228] In embodiments, the fourth start date is after each of the first start date, the first end date, the second start date, the second end date, the third start date, and the third end date and before each of the fourth end date.

[0229] In embodiments, the fourth end date is after each of the first start date, the first end date, the second start date, the second end date, the third start date, the third end date, and the fourth start date.

[0230] In embodiments, there can be n developmental stages, where the nth start date and the nth end date occur in a chronological order as described above. In embodiments, the first end date can be the same day as the second start date or the day before the second start date. In embodiments, the second end date can be the same day as the third start date or the day before the third start date. In embodiments, the third end date can be the same day as the fourth start date or the day before the fourth start date. In embodiments, the fourth end date can be the same day as the Nth start date or the day before the Nth start date. It should be noted that the various age ranges of development depend on the animal species.

[0231] At step 814, after the animal completes the first stage of development (block 812), i.e., when the animal is in the first age range, a 3D X-ray computed tomography scan of the animal is acquired. The scan at step 814 involves determining any abnormalities or health conditions (e.g., presence or absence of cysts, tumors, pleurisy, and foreign bodies) that can affect the final value of the animal.

[0232] At step 818, after the animal has completed the mid-development period (block 816), i.e., when the animal is in the second age range, another 3D X-ray computed tomography scan of the animal is taken. The quality control scan at step 818 enables driving optimization of the animal and herd as a whole. It is at this stage that a significant shift in valuation of the animal and herd can be realized.

[0233] At step 822, once the animal has been raised through the late development period (block 820) and is ready to leave the farm, i.e., when the animal is in the third age range, yet another scan of the animal is taken. At step 822, the 3D X-ray computed tomography scan is used to generate a complete analysis of the animal (to generate metrics or measurement data, such as lean meat yield, local diet quality, and health conditions) that together fully describe the animal to showcase in an auction, thereby enabling a final purchase price. In various embodiments, the data from the scan steps 818 and 822 are evaluated / analyzed by a plurality of programmed codes or instructions to determine data indicative of the value of the animal based on at least one of a plurality of pre-sale parameters, including lean meat yield, ratio of intramuscular fat to tissue, intermuscular fat amount, absolute and relative size of individual organs, muscle volume, number of ribs, and presence or absence of cysts, tumors, pleurisy, and foreign objects. In various embodiments, the plurality of programmed codes or instructions generate data indicative of lean meat yield, the lean meat yield being associated with a first range of values; generate data indicative of a ratio of intramuscular fat to tissue, the ratio being associated with a second range of values; generate data indicative of an intermuscular fat amount, the intermuscular fat amount being associated with a third range of values; generate data indicative of absolute and relative size of individual organs, the data being associated with a fourth range of values; generate data indicative of muscle volume, the data being associated with a fifth range of values; generate data indicative of a number of ribs, the data being associated with a sixth range of values; and generate data indicative of presence or absence of cysts, tumors, pleurisy, and foreign objects.

[0234] It is well known that the transition of animals from a farm to a sales venue is stressful for the animals and expensive for the farmer. Thus, the ability to conduct a virtual auction using electronic data, including data from 3D X-ray computed tomography scan data, is beneficial.

[0235] After the animals are sold and transported from the farm (blocks 824, 826, respectively), at step 828, a 3D X-ray computed tomography scan is acquired at the feedlot. The scan at step 828 is intended to perform an incoming inspection of the animals after the auction to verify the electronic data provided at the auction and to inspect the health of the animals when combined from multiple herds. Thus, the data from the scan at step 828 is used to determine one or more of a plurality of post-sale parameters. In embodiments, the verification of the electronic data involves comparing at least a portion of the plurality of pre-sale parameters to at least a portion of the plurality of post-sale parameters. In embodiments, the plurality of post-sale parameters includes lean meat yield, the ratio of intramuscular fat to tissue, the amount of intermuscular fat, the absolute and relative size of individual organs, muscle volume, the number of ribs, and the presence or absence of cysts, tumors, pleurisy, and foreign objects.

[0236] At step 832, a final scan of the animals is performed at the end of the feedlot process (block 830), where the animals have typically been finished prior to slaughter. This final scan provides initial data to enable planning of production schedules / processes (block 834), and thus, to optimize the plant processes, and thereafter, final assignment to customers (block 836).

[0237] It will be appreciated by those of ordinary skill in the art that, in some embodiments, the scan information generated for animals at a particular stage of development is aggregated with information from other animals at similar and different stages of development to determine, using methods such as, for example, artificial intelligence and big data analysis, the predicted outcomes for the animals and the impact on the overall development of the herds within a particular farm and between different farms.

[0238] In embodiments, for the purposes of this specification, multi-energy computed tomography and transmission X-ray screening can be employed. In embodiments, the use of multi-energy transmission X-ray screening enables improved Z eff reconstruction, resulting in improved chemical barrenness accuracy and improved bone structure location, especially in high attenuation regions. Further, the use of multi-energy transmission X-ray screening enables improved Z eff reconstruction for foreign object detection and final product quality control.

[0239] In embodiments, the above-described techniques can be combined with meat processing and plant safety practices. In embodiments, this specification employs software to link three-dimensional imaging and multi-energy meat processing techniques with plant operations. In embodiments, this specification employs software to link three-dimensional imaging and multi-energy meat processing techniques with agricultural practices. In embodiments, this specification employs modified safety techniques, such as personnel and baggage screening systems, so that these techniques can be used across multiple applications in the meat industry.

[0240] FIG. 9A and 9B Top views of a 3D stationary gantry X-ray CT imaging system 900 (also referred to as a real-time tomography (RTT) system) in first and second configurations, respectively, to scan meat in a slaughterhouse 901 are shown in accordance with some embodiments of the present specification. Referring now to FIG. 9A and 9B The system 900 is deployed in the slaughterhouse 901 to scan carcasses 905 that are hung on hooks of a conveyor or conveyor rail 910 and moved by the conveyor rail 910 through the system 900. In some embodiments, the carcasses 905 are moved by the conveyor rail 910 through the system 900 at a speed ranging from 0.05 m / s to 0.5 m / s.

[0241] In some embodiments, the system 900 is enclosed within a food-safe, environmentally-protected enclosure 915 that is manufactured using materials such as, but not limited to, stainless steel and / or plastic. In some embodiments, the system 900 is surrounded by at least one radiation-shielded enclosure or tunnel 920. A control room 925 is provided for one or more system operators to inspect the performance of the system 900 on one or more inspection workstations 927. Service access 930 is also provided to the system 900. In various embodiments, the one or more inspection workstations 927 are computing devices.

[0242] In some embodiments, the system 900 is configured to perform dual-plane scanning of the carcasses and includes a first plurality of linear multi-focal spot X-ray sources and associated first detector arrays positioned or deployed around an inspection region, area, or bore to scan the carcasses in a first imaging plane 942, and a second plurality of linear multi-focal spot X-ray sources and associated second detector arrays also positioned or deployed around the inspection region, area, or bore to scan the carcasses in a second imaging plane 943. In some embodiments, the first and second imaging planes 942, 943 are along a direction parallel to the direction of motion of the carcasses along the conveyor rail 910. In embodiments, the first plurality of linear multi-focal spot X-ray sources are offset from the associated first detector arrays in the first imaging plane 942 by a distance dl, while the second plurality of linear multi-focal spot X-ray sources are offset from the associated second detector arrays in the second imaging plane 943 by a distance d2. In some embodiments, dl is equal to d2. In various embodiments, the distances dl and d2 range from 1 mm to 10 mm.

[0243] In some embodiments, as FIG. 9AAs shown, the at least one radiation shielding enclosure 920 and conveyor track 910 can have a labyrinth or maze-like layout such that there is no straight path through the at least one radiation shielding enclosure 920 and conveyor track 910, and any path through the at least one radiation shielding enclosure 920 and conveyor track 910 requires at least more than 1 turn and less than 20 turns, preferably in the range of 2 to 5 turns, with a turn radius of more than 10% per turn. Additionally, each turn is preferably in the range of 25 degrees to 80 degrees and any increment therein. In embodiments, the labyrinth layout is used to limit radiation exposure of workers in the slaughterhouse 901 to below statutory limits (typically, less than 1 μSv / hr in any one hour). In some embodiments, as shown in FIG. 9, the at least one radiation shielding enclosure 920 and conveyor track 910 have a generally linear manner of layout, but with one or more deviations, curves or turns 935 to limit radiation exposure of workers in the slaughterhouse 901 to below statutory limits. Those of ordinary skill in the art will appreciate that the configuration of FIG. 9 is merely exemplary and in no way limiting. For example, in alternative embodiments, the at least one radiation shielding enclosure 920 and conveyor track 910 can have other layout configurations, such as, but not limited to, elbow or staircases. FIG. 9B As shown, the at least one radiation shielding enclosure 920 and conveyor track 910 have a generally linear manner of layout, but with one or more deviations, curves or turns 935 to limit radiation exposure of workers in the slaughterhouse 901 to below statutory limits. Those of ordinary skill in the art will appreciate that the configuration of FIG. 9 is merely exemplary and in no way limiting. For example, in alternative embodiments, the at least one radiation shielding enclosure 920 and conveyor track 910 can have other layout configurations, such as, but not limited to, elbow or staircases. FIG. 9A As shown, the at least one radiation shielding enclosure 920 and conveyor track 910 have a generally linear manner of layout, but with one or more deviations, curves or turns 935 to limit radiation exposure of workers in the slaughterhouse 901 to below statutory limits. Those of ordinary skill in the art will appreciate that the configuration of FIG. 9 is merely exemplary and in no way limiting. For example, in alternative embodiments, the at least one radiation shielding enclosure 920 and conveyor track 910 can have other layout configurations, such as, but not limited to, elbow or staircases. 9B As shown, the at least one radiation shielding enclosure 920 and conveyor track 910 have a generally linear manner of layout, but with one or more deviations, curves or turns 935 to limit radiation exposure of workers in the slaughterhouse 901 to below statutory limits. Those of ordinary skill in the art will appreciate that the configuration of FIG. 9 is merely exemplary and in no way limiting. For example, in alternative embodiments, the at least one radiation shielding enclosure 920 and conveyor track 910 can have other layout configurations, such as, but not limited to, elbow or staircases.

[0244] FIG. 10A First, second and third cross-sectional views 1040a, 1040b, 1040c of a 3D stationary gantry X-ray CT imaging system 1000 configured for dual plane scanning of a carcass according to some embodiments of the present specification are shown. The first cross-sectional view 1040a is along a direction parallel to the direction of motion of the carcass along the conveyor track 1010 and perpendicular to the first imaging plane 1042. In embodiments, the first imaging plane 1042 includes a plurality of individual linear multi-focal spot X-ray sources 1045a arranged around the examination region 1050. In some embodiments, the first imaging plane 1042 includes, for example, five linear multi-focal spot X-ray sources 1045a separated from each other and positioned around or along the periphery of the examination region 1050.

[0245] An inspection region or aperture 1050 is defined by the food-safe environmental enclosure or housing 1015. The inspection region or aperture 1050 is positioned around an array of X-ray detectors 1055a in a first imaging plane 1042 such that the X-ray detectors 1055a are located between the linear multi-focal point X-ray source 1045a and the housing 1015. The array of detectors 1055a is offset from the plane of the X-ray source 1045a by a distance of 1 mm to 10 mm such that X-rays from the multi-focal point X-ray source on one side of the inspection aperture 1050 can pass over an adjacent X-ray detector and interact with an X-ray detector on the opposite side of the inspection region 1050, thereby forming a transmission image through the inspected carcass.

[0246] A second cross-sectional view 1040b is along a direction parallel to the motion of the carcass along the conveyor rail 1010 and perpendicular to the second imaging plane. In embodiments, the second imaging plane also includes a plurality of individual linear multi-focal point X-ray sources 1045b arranged around the inspection region 1050. In some embodiments, the second imaging plane 1043 includes, for example, five linear multi-focal point X-ray sources 1045b that are separated from one another and positioned around or along the perimeter of the inspection region 1050. In some embodiments, the five linear multi-focal point X-ray sources 1045b (in the second imaging plane 1043) are disposed or positioned to fill the gaps separating the five linear multi-focal point X-ray sources 1045a (in the first imaging plane 1042).

[0247] The inspection region or aperture 1050 is positioned around another array of X-ray detectors 1055b in a second imaging plane 1043 such that the X-ray detectors 1055b are located between the linear multi-focal point X-ray sources 1045b and the housing 1015. The array of detectors 1055b is also offset from the plane of the X-ray sources 1045b by a few millimeters such that X-rays from the multi-focal point X-ray source on one side of the inspection aperture 1050 can pass over an adjacent X-ray detector and interact with an X-ray detector on the other side of the inspection region 1050, thereby forming a transmission image through the inspected carcass.

[0248] A third cross-sectional view 1040c illustrates a composite representation of the first and second imaging planes 1042, 1043 as the carcass moves through the system 1000. The view 1040c illustrates the complete trajectory of the multi-focal point X-ray source points around the inspection region 1050 required to form a high-quality 3D tomographic image of the carcass. A small region 1060 of missing data is observable near the hook that transported the carcass. Accordingly, the image reconstruction algorithm of the system 1000 is configured to minimize the impact of the missing data in the final image.

[0249] During a scan operation, each X-ray source point within a single multi-focus X-ray source (1045a, 1045b) is turned on in sequence and projection data is collected through the carcass for that one source point. When the exposure is complete, a different X-ray source point within a different multi-focus X-ray source in the system 1000 is turned on, for example, to create the next X-ray projection. The scanning process continues until all X-ray sources have been fired in a sequence configured to optimize the X-ray image quality of the reconstruction.

[0250] In some embodiments, the inspection region 1050 has a cross-sectional shape that is a composite of a first rectangular shape mounted with a second triangular shape. In some embodiments, the first rectangular cross-sectional shape has exemplary dimensions defined by a width that is less than 20% of the height, preferably less than 40% of the height. In some embodiments, the first rectangular cross-sectional shape has exemplary dimensions (region) of 1500 mm (width) x 3900 mm (height). In some embodiments, the region of the second triangular shape is substantially smaller or negligible compared to the region of the first rectangular shape. Thus, for practical purposes, the exemplary dimensions (region) of 1500 mm (width) x 3900 mm (height) for the first rectangular cross-sectional shape represent the composite region - i.e., the inspection region 1050. It will be appreciated that such dimensions (region) of 1500 mm (width) x 3900 mm (height) for the inspection region or aperture 1050 are suitable for scanning beef carcasses in some embodiments.

[0251] FIG. 10BA fourth cross-sectional view 1040d of the 3D stationary gantry X-ray CT imaging system 1000 according to some embodiments of the present specification is shown. The fourth cross-sectional view 1040d is along a direction perpendicular to the movement of the carcass 1070 along the conveyor rail 1010 and parallel to the first and second imaging planes 1042, 1043. In embodiments, the first and second imaging planes 1042, 1043 are separated by a distance “d” to simplify service access. In some embodiments, the distance “d” ranges from 100 mm to 2000 mm. In an embodiment, the distance “d” ranges from 500 mm to 1000 mm. In embodiments, carcass motion that can occur between the first and second imaging planes 1042, 1043 other than in a simple linear direction (e.g., in a swaying and / or rotational motion from front to back and / or left to right) can be measured using standard 3D optical point cloud or radar imaging methods known to those of ordinary skill in the art. This 3D data can be converted to the actual carcass displacement at each point in the field and used to drive the tomographic image reconstruction back-projection process. This can be accomplished by methods known to those of skill in the art, for example, recalculating the direction of each X-ray projection from source to detector through the virtual carcass in computer memory as the image reconstruction process occurs.

[0252] According to aspects of the present specification, by deploying a specific imaging geometry, the size of the examination region can be configured for a specific carcass-based application, including a) selecting the number and location of multi-focal spot X-ray sources (e.g., sources 1045a, 1045b) to use, and b) configuring the array of X-ray detectors (e.g., detectors 1055a, 1055b) to fit the X-ray source locations. The specific imaging system geometry is passed to the X-ray 3D image reconstruction algorithm, where a re-computation of the weighting function is performed to ensure accurate image reconstruction. FIG. 9A 、 FIG. 9B 、 FIG. 10A and FIG. 10B Embodiments of the present specification represent the type of imaging system that can be deployed in a slaughterhouse that processes beef. Relatively smaller examination regions or bores are typically required in slaughterhouses that process pigs, goats, and lambs. It will be appreciated that under an examination region or bore of about 1 m in diameter, it is typically more cost effective to use a single plane scanning system such as a rotating gantry computed tomography system or a stationary gantry imaging system with a rectangular or circular tube configuration.

[0253] For example, FIG. 11First, second, and third cross-sectional views 1140a, 1140b, 1140c of a 3D stationary gantry X-ray CT imaging system 1100 configured for dual plane scanning of a carcass according to some embodiments of the present specification are shown. The first cross-sectional view 1140a is along a direction parallel to the motion of the carcass along the conveyor rail 1110 and perpendicular to the first imaging plane. In embodiments, the first imaging plane includes a plurality of individual linear multi-focal spot X-ray sources 1145a arranged around an examination region, area, or aperture 1150. In some embodiments, the first imaging plane includes, for example, three linear multi-focal spot X-ray sources 1145a separated from one another and positioned around or along a perimeter of the examination region 1150.

[0254] According to one aspect of the present specification, the examination region or aperture 1150 has a polygonal geometry or shape to approximate a circle or circular cross-section. The polygonal shape or geometry is suitable for scanning carcasses of lambs, pigs, and goats. In some embodiments, the examination region or aperture 1150 has a maximum width of 1500 mm and a maximum height of 2000 mm. In some embodiments, the maximum width of the examination region or aperture 1150 is less than 10% of the maximum height, preferably less than 20% of the maximum height.

[0255] In some embodiments, the examination region or aperture 1150 is bounded by a food-safe environmental enclosure or housing 1115. The examination region or aperture 1150 is positioned around an array of X-ray detectors 1155a in the first imaging plane such that the X-ray detectors 1155a are located between the linear multi-focal spot X-ray sources 1145a and the housing 1115. The array of detectors 1155a is offset by a few millimeters from the plane of the X-ray sources 1145a such that X-rays from a multi-focal spot X-ray source on one side of the examination aperture 1150 can pass over an adjacent X-ray detector and interact with an X-ray detector on the other side of the examination region 1150, thereby forming a transmission image through the examined carcass.

[0256] The second cross-sectional view 1140b is along a direction parallel to the motion of the carcass along the conveyor rail 1110 and perpendicular to the second imaging plane. In embodiments, the second imaging plane also includes a plurality of individual linear multi-focal spot X-ray sources 1145b arranged around the examination region 1150. In some embodiments, the second imaging plane includes, for example, three linear multi-focal spot X-ray sources 1145b separated from one another and positioned along a perimeter of the examination region 1150. In some embodiments, the three linear multi-focal spot X-ray sources 1145b (in the second imaging plane) are disposed or positioned to fill the gap separating the three linear multi-focal spot X-ray sources 1145a (in the first imaging plane).

[0257] An examination region or aperture 1150 is positioned around an array of another X-ray detector 1155b in a second imaging plane, such that the X-ray detector 1155b is located between the linear multi-focus X-ray source 1145b and the housing 1115. The array of detectors 1155b is also offset from the plane of the X-ray source 1145b by a few millimeters, such that X-rays from the multi-focus X-ray source on one side of the examination aperture 1150 can pass over an adjacent X-ray detector and interact with an X-ray detector on the other side of the examination region 1150, thereby forming a transmission image through the examined animal.

[0258] A third cross-sectional view 1140c illustrates a composite representation of the first and second imaging planes as the animal moves through the system 1100. The view 1140c illustrates the complete trajectory of the multi-focus X-ray source points around the examination region 1150 required to form a high-quality 3D tomographic image of the animal. A small region 1160 of missing data is observable near the hook that transports the animal. Accordingly, the image reconstruction algorithm of the system 1100 is configured to minimize the effect of the missing data in the final image.

[0259] As another example, FIG. 12 A cross-sectional view 1240 of a 3D stationary gantry X-ray CT imaging system 1200 configured for single-plane scanning of an animal according to some embodiments of the present specification is shown. The system 1200 includes a plurality of individual linear multi-focus X-ray sources 1245 arranged around the perimeter of an examination region, area or aperture 1250. The examination region, area or aperture 1250 is surrounded by an array of X-ray detectors 1255. The plurality of individual linear multi-focus X-ray sources 1245 and the array of X-ray detectors 1255 are enclosed in a housing 1215.

[0260] The figure also shows a plurality of first structures 1270 for dissipating heat from the plurality of X-ray sources 1245, and at least one second structure 1275 for dissipating heat from and providing a voltage supply to the plurality of X-ray sources 1245. In embodiments, the first structures 1270 are designed to maximize mechanical integrity and thermal conductivity. The at least one second structure 1275 includes thermally conductive elements to dissipate heat from the anode region, and also includes a metal rod passing through its center to provide the voltage.

[0261] According to an aspect of the present specification, the examination region, area or aperture 1250 has a substantially non-circular geometry or shape, such as a rectangular or square shape. The rectangular or square shape or geometry is suitable for scanning entire poultry and beef, lamb, pig and goat animal parts during deboning processes. In some embodiments, the examination region or aperture 1250 has dimensions of 600 mm (width) x 450 mm (height).

[0262] FIG. 13Bottom, top, longitudinal side, and end views 1305a, 1305b, 1305c, 1305d of a linear multi-focus X-ray source 1345 for a 3D stationary gantry X-ray CT imaging system according to embodiments of the present specification are shown. Referring now to views 1305a, 1305b, 1305c, 1305d concurrently, source 1345 includes a plurality of electron guns, cathodes, or source / emission points 1310 and an anode 1315 housed in a vacuum tube or enclosure 1320. In some embodiments, source 1345 includes 100 X-ray emission points 1310 at 10 mm intervals on a 1000 mm long active anode 1315.

[0263] In some embodiments, first, second, and third supports 1322a, 1322b, 1322c are deployed to support anode 1315 along the longitudinal axis. First and second supports 1322a, 1322b are deployed at the two ends, while third support 1322c is deployed at the center of anode 1315. In some embodiments, first and second supports 1322a, 1322b also serve as coolant feedthrough units, while third support 1322c enables high voltage feedthrough. In some embodiments, anode 1315 supports operating tube voltages in the range of 100 kV to 300 kV. In some embodiments, each electron gun, cathode, or source / emission point 1310 emits tube currents in the range of 1 mA to 500 mA, depending on the subject thickness and the inspection area, aperture, or size - the larger the inspection aperture and the thicker the subject, the higher the tube current required.

[0264] In some embodiments, each electron gun 1310 is configured to illuminate an area or lesion on anode 1315 ranging between 0.5 mm to 3.0 mm in diameter. The specific size of the lesion is chosen to maximize image quality and minimize heating of anode 1315 during X-ray exposure. The higher the product of tube current and tube voltage, the larger the lesion is typically designed to be.

[0265] According to aspects of the present specification, the 3D stationary gantry x-ray CT imaging system 100 includes multiple design features and manufacturing methods of the CT tube that have improved performance and stability from a physical and mechanical perspective, in addition to reducing production costs. In an embodiment, the housing is made of stainless steel and is formed using hydroforming rather than metal stamping. In an embodiment, the hydroforming manufacturing method uses high pressure fluid to press material into a mold to form the desired shape. In contrast, metal stamping uses a custom male mold and a custom female mold to press material into the desired shape. Using hydroforming has many advantages over metal stamping, including the advantages presented below. For example, there is less material waste during the forming process (approximately 0-10%); stamping is typically 20% or more waste. Additionally, hydroforming has lower upfront costs because it only requires one custom mold, as opposed to two molds in stamping. This also helps to shorten the production lead time and reduce the cost of part batch production. Furthermore, hydroforming provides the ability to form more complex shapes and features that would typically not be possible in stamping. Even with the added complexity, hydroformed parts are typically manufactured at a faster rate, which also helps to shorten the manufacturing time when the system is produced. Still further, hydroforming provides better surface finish because the water shapes the material rather than another metal part surface, which translates to better stability in the CT tube when high voltages are applied. Generally, hydroformed parts have greater strength properties than stamped parts. The uniform distribution of compression forces from the liquid during the forming process generally results in a more rigid part. Stamping has a greater likelihood of causing the sheet material to thin to an undesirable thickness with weaker strength properties in certain areas. Because the tube is placed under atmospheric vacuum pressure while trying to maintain a relatively compact shape and low weight, hydroforming contributes positively to the overall uniformity of the design. Additionally, hydroformed parts generally experience less material springback in the hydroforming, resulting in more accurate and consistent geometry, which is critical for the CT tube housing. It should be noted that the hydroforming process cannot produce a uniform material thickness or flatness as desired with machining.

[0266] According to some embodiments, the CT multi-energy detector module is composed of a printed circuit board (PCB), electrical components soldered to the PCB to form a printed circuit board assembly (PCBA), and a detector crystal (CdTe or CdZnTe) assembled onto the PCBA. The final processing steps to complete the detector module assembly include the attachment of high voltage flex circuit (HV flex) and the addition of a protective coating. FIG. 62A fully assembled sensor module 6200 is shown. Image 6204 shows the side of the board with the detector crystals at the top under the high voltage flexible board. Image 6206 shows the side of the board with 2 ASICs at the top. The PCB of the sensor board is preferably characterized by at least the following requirements: it should have a high density of traces and interconnections to carry the electronic signals from the detector crystals to the ASICs, and the area around the detector crystals and each pad in the detector crystal should be flat.

[0267] FIG. 14 is a block diagram illustration of a plurality of exemplary information, outputs or results derived from processing of carcass scan image data generated using a dual-plane 3D stationary gantry X-ray CT imaging system in accordance with some embodiments of the present specification. In embodiments, a controller in data communication with the 3D stationary gantry X-ray CT imaging system implements a plurality of instructions or programming code to receive 3D scan image data, process or analyze the scan image data and generate various outputs or results such as effective Z, density information, 3D structure of the animal, computed lean yield, intramuscular fat analysis, intermuscular fat volume, ratio of intramuscular fat (marbling) to tissue, absolute and relative size of individual organs, muscle volume, number of ribs, and presence or absence of cysts, tumors, pleurisy, and foreign objects.

[0268] In accordance with aspects of the present specification, 3D scan image data of a carcass provides effective Z (atomic number) and density information (block 1405) leading to insights related to the 3D structure of the carcass (block 1406) including bone, fat, and tissue structure, and thus can be used to drive a system for automatically cutting the carcass based on the structure of the carcass (block 1407). For example, it is well known that lamb carcasses typically have 8 ribs, but sometimes a lamb can have only 7 or even 9 ribs. Continuing this example, to plan the best yield for a slaughterhouse, it is necessary to determine how many lamb chops are needed instead of a rack of lamb that typically contains 7 ribs. Thus, a carcass can yield 1 rack, 1 rack and 1 chop, or 1 rack and 2 chops. Ideally, the decision of whether the carcass should be processed into individual chops or into a rack plus chops is made before starting the day’s production. Thus, in some embodiments, using 3D imaging can drive the best production plan (block 1408) and establish the correct cutting sequence for one or more automated cutting devices. As a non-limiting example, FIG. 20 A 3D scan image 2002 of a beef carcass 2004 is shown. The scan image 2002 provides insights related to the 3D structure of the carcass 2004 including bone, fat, and tissue structure. The insights related to the 3D structure enable volumetric bone localization for automatic alignment, positioning, and subsequent cutting of the carcass 2004.

[0269] In some embodiments, the 3D scan image data of the carcass can also be used to determine the eating quality in 3D as a whole within the carcass (block 1410). The densities of fat and muscle are known to be dissimilar. Thus, they appear at different gray levels in the reconstructed X-ray images. For example, indicators of the eating quality of beef are determined by a) the ratio of intramuscular fat to tissue (marbling) and b) the amount of intermuscular fat. At each point in each muscle, the eating quality analysis determines, at each point in each muscle, a first amount or portion of each muscle that is destined for the highest value output, a second amount or portion that is destined for standard output, and a third amount or portion that is destined for low value output. This analysis drives the overall valuation of the carcass (block 1412) and ensures that the farmer can be fairly rewarded for producing high quality animals, rather than just the weight of the carcass or the lean yield (the percentage of meat, fat, and bone in the carcass).

[0270] FIG. 21 A histogram analysis 2102 of a first scan image 2104a of a first beef sample and a second scan image 2104b of a second beef sample is shown in accordance with some embodiments of the present specification. The image histogram analysis 2102 shows first, second, and third gray level regions 2106 (fat), 2108 (lean), and 2110 (bone) corresponding to fat, lean, and bone as described in the first beef sample 2104a, and fourth, fifth, and sixth gray level regions 2112 (fat), 2114 (lean), and 2116 (bone) corresponding to fat, lean, and bone as described in the second beef sample 2104b. In embodiments, bins 22-38 correspond to fat; bins 39-54 correspond to lean; and bins 60-74 correspond to bone. These identified gray level regions are used to segment the first scan image 2104a and the second scan image 2104b to determine parameters such as volumetric content (e.g., muscle volume) and ratios (e.g., lean yield, and the ratio of intramuscular fat to tissue).

[0271] In some embodiments, further analysis of the 3D image data provides information about the health of the carcass / animal (block 1415), such as the presence of foreign objects such as syringe needles and barbed wire inclusions, and the presence of cysts and tumors, oversized organs, pleurisy, and other common ailments. Overall, this information also drives the valuation of the carcass, as unhealthy carcasses will be diverted to the low value food chain, while improving overall quality control in food safety (block 1417).

[0272] FIG. 15is a workflow illustrating the use of multiple 3D X-ray computed tomography processes for improving slaughterhouse management and automation in accordance with some embodiments of the present specification. In embodiments, a controller in data communication with a 3D X-ray computed tomography machine implements multiple instructions or programming code to receive 3D scan image data, process or analyze the scan image data and generate various outputs or results.

[0273] At step 1502, the animal is processed to remove skin, innards, limbs, and trim waste. At step 1504, a full carcass scan or inspection is performed using a 3D X-ray computed tomography system (such as those described with reference to FIG. 9A 、 FIG. 9B 、 FIG. 10A 、 FIG. 10B 、 FIG. 11 and FIG. 12 The carcass scan data can be analyzed to determine measurements or information related to eating quality, to understand animal health, to determine carcass value, to provide input to production planning processes, and to enable optimal processing of the animal to meet customer requirements. In various embodiments, the value of the carcass is determined based on at least one of lean meat yield, ratio of intramuscular fat to tissue, amount of intermuscular fat, absolute and relative size of individual organs, muscle volume, number of ribs, and presence or absence of cysts, tumors, pleurisy, and foreign objects.

[0274] Accordingly, at step 1506, the non-food products of the carcass are sent to an alternative processing stream. At step 1508, the innards and other byproducts are scanned to provide further input to animal health measurements (e.g., to check individual organs for abnormalities and the presence or absence of cysts, tumors, pleurisy, and foreign objects). This can again affect carcass health, carcass valuation, and subsequent production process planning. Thereafter, at step 1510, the carcass is sent for storage in a refrigerated room that is maintained at a temperature of less than 15 degrees Celsius, and preferably at about 12 degrees Celsius. At step 1512, production requirements are planned based on the cold carcass inventory.

[0275] Now, at step 1514, once the carcass has been in the chill room for 24 to 36 hours, the carcass is fully scanned. At this point, the carcass will have stabilized into a rigid shape, and the re-imaging with the 3D X-ray computed tomography system ensures the most accurate scan image data of the areas of continuous meat indicative of the skeletal, fat, and tissue structure and thus the predetermined level of quality (e.g., determined by the ratio of intramuscular fat to tissue and the amount of intermuscular fat) is sent to the automated cutting system which is used to segment the initial carcass into smaller pieces for more efficient processing in the deboning room. At step 1516, the carcass is sent to the deboning room, and thereafter, at step 1518, the automated cutting system performs the primary carcass cut to segment the carcass into manageable sizes for final dissection.

[0276] Next, at step 1520, in some embodiments, the smaller carcass pieces (resulting from step 1518) are scanned using a 3D X-ray screening system (e.g., the screening system of FIG. 4 ) having a smaller examination area, aperture, tunnel, or region in order to generate scan image data and determine therefrom the accurate 3D structure of the smaller carcass pieces, and then the expensive meat pieces (e.g., beef strip-loin) are automatically deboned. Here, the precise registration between the 3D scan image data and the automated cutting system is critical to avoid wasting valuable product and bone fragmentation into the final product. Subsequently, at step 1522, the meat is trimmed from the bone as needed from each smaller carcass piece.

[0277] At step 1524, in some embodiments, the meat is scanned using a 3D X-ray screening system having a smaller examination area, aperture, tunnel, or region, and the scan image data is analyzed to determine measurements related to individual dissected meat pieces (e.g., T-bone or rib eye steaks) for key quality indicators such as eating quality, fat thickness, and presence of foreign matter including bone fragments. The amount of meat remaining on the bone after deboning is also determined. If excess meat remains, the bone can be sent back for further processing to extract the remaining meat into the food chain. Subsequently, at step 1526, a quality control function is performed to ensure the final product meets customer requirements, and then at step 1528, the individual meat products are packaged.

[0278] Next, at step 1530, a quality control scan is performed on the individual meat pieces after packaging. The purpose of this inspection is to look for foreign matter as well as measurements such as fat thickness around a piece of steak, for example, to ensure customer requirements are met. In some embodiments, this step is done with a 3D X-ray CT system (e.g., the FIG. 12 ), a two-dimensional X-ray system, or a camera system. At step 1532, the packaged meat products are now boxed for each customer.

[0279] Now, at step 1534, the whole box of packaged meat is scanned by a 3D X-ray computed tomography system with a smaller inspection area, aperture, tunnel, or region for facilitating a final quality control function. During the final quality control function, at step 1536, the packaged list to be provided to the customer is compared to the actual contents of the box using an automated analysis method, such as a deep learning method, for example, to verify that the correct number of each type of product in the box has the expected eating quality, shape, and size specifications, where the eating quality is determined based on at least one of the ratio of intramuscular fat to tissue and the amount of intermuscular fat. Finally, at step 1538, the boxed product is dispatched to the customer.

[0280] In an embodiment, steps 1504, 1508, 1514, 1520, 1524, 1530, and 1534 highlight the process by which 3D X-ray carcass inspection adds value to improve overall abattoir production operations.

[0281] FIG. 16A is a workflow diagram illustrating an exemplary network layout of a semi-automated meat production process according to an embodiment of the present specification. In an embodiment, the meat production process workflow 1600 includes a 3D X-ray tomography 1602; a 2D X-ray tomography 1604; a hyperspectral and fluorescence scanner 1606; a handheld device 1608; a database 1610; an inspection workstation 1612; a quality control system and device 1614; an automation system 1616; a meat grading algorithm 1618; a carcass valuation algorithm 1620; a production planning algorithm 1622; an animal health algorithm 1624; and a product quality inspection and verification algorithm 1626, where all of the element blocks are coupled to a common communication / data network 1628. In an embodiment, the meat production process further includes 3D and 2D X-ray scanners and other sensing elements, such as RFID and / or barcode readers and / or cameras 1630.

[0282] In an embodiment, the common communication / data network 1628 enables real-time storage and retrieval of data from the database 1610, thereby providing a fast search facility to store and retrieve data.

[0283] The common communication / data network 1628 also facilitates real-time transmission of image data from the sensing elements, such as but not limited to the 3D X-ray tomography 1602, the 2D X-ray tomography 1604, the hyperspectral and fluorescence scanner 1606, and the handheld device 1608, to an algorithm processing unit, which can analyze the data from the sensing elements to generate information required for optimal operation of the meat production process.

[0284] The common communication / data network 1628 also enables real-time transfer of data from the sensing elements to automated cutting systems used in the meat production process as well as human operators to guide the cutting of carcasses and / or primal meat into retail meat cuts on an individual carcass basis. The common communication / data network 1628 also enables data from sensing elements employed in the meat processing plant to be analyzed by automated quality control processes 1626 and human quality control personnel to ensure accurate processing and food safety standards. In an embodiment, the common communication / data network 1628 provides a means for real-time display of production metrics and other data, such as financial reports, that support the meat production plant management to provide the highest possible productivity from the plant.

[0285] Reference FIG. 16A In an embodiment, the present specification provides an inspection workstation 1612 that operates as a plant management dashboard providing real-time updates to the status of all products within the meat processing plant for the operators of the plant. In an embodiment, the status information includes real-time location of a carcass, primal meat, retail meat cut, trim product, or packaged product identified by way of a unique ID. In an example scenario, if a sensing element of one or more of the 3D X-ray tomography 1602, 2D X-ray tomography 1604, hyperspectral and fluorescence scanners 1606, handheld devices 1608 detects fecal contamination of a particular primal meat by way of its unique ID, the inspection workstation / dashboard 1612 immediately displays the location of the remaining carcass, as well as any other primal meat, retail meat cut, trim, or packaged product originating from the same carcass for the operator to view.

[0286] The hyperspectral camera generates hyperspectral data that includes a plurality of different wavelengths detected at each pixel location. Thus, instead of a given pixel having a single color value assigned to it, the hyperspectral data includes a plurality of detected wavelengths at each pixel location. The plurality of wavelengths includes one or more wavelengths or any increment or sub-range of values therein ranging from 100 nm to 15,000 nm. Thus, the resulting image includes more than one wavelength from the continuum of the spectrum detected at each pixel.

[0287] In an embodiment, the status information displayed by the inspection workstation / dashboard 1612 includes at least one of the following: real-time notification of any package mislabeling or incorrect shipable case contents; real-time notification and location of any animal health defects identified by any sensing element or human operator within the plant; real-time production data, including output over adjustable time scales (e.g., current shift, day, week, month, or year); real-time plan variance; real-time notification of production backlogs or product non-conformance areas requiring managerial action; real-time financial data of retail product value based on objective measurements from suitable sensors within the plant; and other relevant data such as, but not limited to, employee utilization, employee efficiency, and work accuracy.

[0288] In an embodiment, the present specification provides a method of identifying the location of all workers working in a meat processing plant in real-time by providing each member of the workforce with a Wi-Fi, GPS, or other suitable location sensor. Reference is made to FIG. 16A In an embodiment, camera systems are installed in the premises of the meat processing plant for providing real-time data to analytical algorithms such as meat grading algorithms 1618, carcass valuation algorithms 1620, production planning algorithms 1622, animal health algorithms 1624, and product quality inspection and verification algorithms 1626. In an embodiment, the real-time data is used for: automatic time and motion studies to determine where plant efficiency can be realized by more efficient use of personnel and facilities; automatic technical analysis to determine the distinguishing characteristics of high performance operators, which can then be used to train lower performance / less efficient operators; automatic review of safety work practices of all workers using knives to determine best practices, thereby improving safety throughout the plant; and quality assurance.

[0289] In an embodiment, the present specification provides an augmented reality based method for achieving optimal cutting of carcasses, primal cuts, and retail cuts of meat in a meat processing plant. FIG. 16Bis a block diagram illustrating an augmented reality based system for cutting meat in a meat processing plant. In an embodiment, the system 1650 located in a meat processing plant includes a meat cutting station 1652 for cutting carcasses into primal and retail meat pieces. The meat cutting station 1652 is coupled to a controller workstation 1654. In an embodiment, the meat cutting station 1652 includes a light / laser projector 1656, one or more haptic feedback devices 1658, and one or more active viewers 1660 for electronically guiding an operator to cut a carcass in a desired manner. Each of the light / laser projector 1656, one or more haptic feedback devices 1658, and one or more viewers 1660 are electronically coupled to the controller workstation 1654, which in an embodiment is a computing device that controls the operation of the devices. In an embodiment, the light / laser projector 1656 is used to project an image of the desired primal and retail meat pieces onto the meat cutting station 1652 to guide the operator of the meat processing plant to cut the carcass as shown in the projected image. In another embodiment, the one or more haptic feedback devices 1658 include cutting tools such as but not limited to a blade haptic device that stops vibrating when the blade is in the desired position relative to the carcass to enable the operator to produce the desired primal and retail meat pieces from the carcass. In an embodiment, the one or more viewers 1660 include wearable active glasses to project or otherwise display how and where to cut or trim individual carcasses, primal, or retail meat pieces to provide the best results. It will be apparent to those skilled in the art that other feedback mechanisms such as but not limited to audible tones, indicators, or video monitors can also be used alone or in combination with other augmented reality devices to enable the operator to deliver the best cuts.

[0290] FIG. 16Cis a flowchart showing steps of an augmented reality based method for cutting meat in a meat processing plant. At step 1670, specifications of desired shapes, weights, and sizes of primal and retail cuts of meat required for a given carcass are received by a computer coupled to a meat cutting station in a meat processing plant. At step 1672, the computer generates one or more images showing how the carcass needs to be cut based on the received specifications. In an embodiment, the images include locations and angles of the cuts of meat required with respect to different parts of the carcass. At step 1674, the generated images are transmitted to a light / laser projector coupled to the meat cutting station. At step 1676, the light / laser projector projects the images on the meat cutting station to guide an operator of the meat processing plant to cut the carcass as shown in the projected images. At step 1678, the computer determines whether the meat cutting station is coupled to one or more active viewers, which in an embodiment include wearable active glasses to project or otherwise display how and where to cut or trim individual carcass, primal, or retail cuts of meat to provide optimal results. At step 1680, if the meat cutting station is coupled to one or more active viewers, the generated images are transmitted to the viewers to guide the operator of the meat processing plant to cut the carcass using the viewers as shown in the projected images. At step 1682, based on the received specifications, the computer transmits signals to a haptic feedback device coupled to the meat cutting station for guiding the device to cut the carcass in the required manner. In an embodiment, the haptic feedback device includes a cutting tool, such as but not limited to a blade haptic device that stops vibrating when the blade is in the required position with respect to the carcass to enable the operator to produce the required primal and retail cuts of meat from the carcass.

[0291] Reference is made to FIG. 16A In various embodiments, data produced by all sensor systems installed in the meat processing plant, such as but not limited to 3D X-ray tomography 1602, 2D X-ray tomography 1604, hyperspectral and fluorescence scanners 1606, and handheld devices 1608, are analyzed by automated algorithms, such as meat grading algorithm 1618, carcass valuation algorithm 1620, production planning algorithm 1622, animal health algorithm 1624, product quality inspection and verification algorithm 1626, to produce information for driving the meat production process.

[0292] In an embodiment, the automated real-time carcass valuation algorithm 1620 identifies the carcass / item derived from the carcass that is contaminated (e.g., by using the hyperspectral and fluorescence scanners 1606). The carcass valuation algorithm 1620 also identifies products (primal and cuts of meat) derived from the same carcass as the contaminated item and marks all of these products for decontamination or further analysis according to the type of contamination.

[0293] In an embodiment, the automated real-time carcass valuation algorithm 1620 also identifies health defects in the carcass, internal organs, and primal cuts. For example, pleuropneumonia; metal contamination from sources such as but not limited to fence wires or syringe needles; tumors or cysts can be identified in the carcass. In addition, tumors, cysts, enlarged organs, and worms can be identified in the internal organs using, for example, hyperspectral and fluorescence scanners 1606 and 3D X-ray tomography scanners 1602, 2D X-ray tomography scanners 1604. In addition, worm nodules, tumors, and cysts can be identified in the primal cuts; and discoloration, worms, tumors, and cysts can be identified in retail cuts processed in the meat processing plant using, for example, 3D X-ray computed tomography images sensors. In another embodiment, the automated real-time carcass valuation algorithm 1620 also identifies the 3D spatial location of the skeletal structure, muscle, intermuscular fat, or health defects within the carcass and primal cuts in order to drive automated cutting equipment and guide operators, for example, by using 3D X-ray computed tomography image sensors.

[0294] In an embodiment, the automated real-time product quality inspection and verification algorithm 1626 identifies the spatial distribution of meat quality within the carcass, primal cuts, or retail cuts or packaged products according to appropriate grading standards (e.g., Australian MSA standards or USDA meat quality standards) using image data obtained from the sensing equipment used in the meat processing plant, such as but not limited to 3D X-ray tomography scanners 1602, 2D X-ray tomography scanners 1604, hyperspectral and fluorescence scanners 1606, and handheld devices 1608.

[0295] In addition, in an embodiment, the automated real-time carcass production planning algorithm 1622 performs carcass valuation, including determining the optimal way to cut the carcass to maximize product revenue given the current customer product delivery requirements. In an embodiment, the production planning algorithm 1622 operates by combining objective measurement data from sensor systems, such as 3D X-ray tomography scanners 1602, 2D X-ray tomography scanners 1604, hyperspectral and fluorescence scanners 1606, and handheld devices 1608 installed in the meat processing plant, including spatially localized information about meat grading, muscle volume, animal health, number of ribs in the carcass, and animal health data obtained through the meat grading algorithm 1618, carcass valuation algorithm 1620, and animal health algorithm 1624.

[0296] In an embodiment, the real-time meat grading algorithm 1618 determines the composition of the trim box to determine the exact ratio of fat to lean. In an embodiment, the meat grading algorithm 1618 uses data from the sensing elements used in the meat production plant, such as 3D X-ray tomography scanners, to generate a percentage score of fat and lean and an indicator of the size distribution of lean and fat items within the trim box.

[0297] In an embodiment, the real-time product quality inspection and verification algorithm 1626 determines whether the labels of the packaged retail meat pieces are completed according to predetermined rules. In an embodiment, data from sensory elements (e.g., 3D X-ray tomography systems) are combined with hyperspectral imaging employed in the meat production plant to determine the weight, meat grade, meat color, fat content, fat thickness, and meat piece type of the products produced by the plant.

[0298] In an embodiment, the real-time product quality inspection and verification algorithm 1626 also determines whether the contents of a carton containing multiple packaged retail meat pieces meet predetermined customer requirements. In an embodiment, the product quality inspection and verification algorithm 1626 uses data from sensory elements (e.g., 3D X-ray tomography systems 1602 employed in the meat production plant) to determine the meat piece type, meat grading score, weight, and fat thickness of each retail meat piece within the carton, and then compares these parameters to product requirements provided by the customer obtained from the production database 1610. In an embodiment, the real-time product quality inspection and verification algorithm 1626 also performs automatic tracking of products throughout the plant by using sensing technologies such as, but not limited to, RFID, barcodes, video tracking, and methods based on time, speed, and distance.

[0299] In an embodiment, the real-time data analysis algorithms provided by the present specification also perform time and motion analysis of individual operators and groups of operators based on camera and position sensor measurements throughout the meat processing plant. It will be apparent to those skilled in the art that other automated analysis algorithms can also be employed in the meat processing plant. Examples of some such real-time automation algorithms include algorithms for monitoring temperature distribution, humidity changes, throughput, and other associated production metrics such as contact labor hours per animal, and the examples of real-time analysis algorithms provided herein are for representative purposes only and should not be considered limiting the scope of the present specification.

[0300] Reference FIG. 16AThe present specification provides for the use of image sensors such as a 3D X-ray tomography scanner 1602 with dual or multi-energy sensors placed in a robotically controlled rotating gantry or fixed gantry imaging geometry. In embodiments, the 3D X-ray tomography scanner 1602 can be used with or without motion correction methods depending on the application in which the scanning technology is deployed. In various embodiments, the 3D X-ray tomography scanner 1602 can be used in various applications in a meat production plant. In one embodiment, the scanner is used to perform a whole carcass scan while the carcass is still warm after entering the slaughter line in a meat processing plant, where the scanner acquires data including carcass volume, spatially localized meat quality grading, bone structure and initial cut line analysis, overall animal health defects including metal object inclusions, tumors and cysts, and where the data is used to obtain accurate carcass valuation and retail revenue estimates. In one embodiment, the 3D X-ray tomography scanner 1602 is used to perform a whole carcass scan after the carcass has hardened after cooling for a day or two, where the scanner acquires data to map the 3D carcass structure to sub-millimeter precision to determine the final cut lines for automatic or manual processing into primal cuts. In one embodiment, the scanner is used to perform an internal organ screening to determine the presence of cysts, tumors, metal and other foreign objects, and subsequent analysis of abnormal organ volume and density.

[0301] In another embodiment, the 3D X-ray tomography scanner 1602 is used to perform a primal cut scan to determine the sub-millimeter 3D location of carcass features immediately prior to automatic or manual cutting equipment in a deboning room. During such a scan, in some embodiments, the primal cut is fixed to a rigid support structure that can be used to transfer the primal cut from the imaging system (scanner 1602) to the automatic robotic cutting equipment in a known reference frame in the meat processing plant. In one embodiment, the 3D X-ray tomography scanner 1602 is used to perform a retail cut scan to determine the cut type, meat grade, weight, fat thickness and cut orientation of a retail cut within a package. The acquired scan data can then be used to cross correlate with a label applied to the package using optical character recognition techniques acquired from a video camera image or a bar code reader. In another embodiment, the scanned data can be used to automatically generate an accurate label that can then be applied directly to the retail cut package. In one embodiment, the 3D X-ray tomography scanner 1602 is used to scan a packaged carton to verify that the contents of a carton containing multiple retail cuts accurately reflect the label affixed to the outside of the carton. In embodiments, each retail cut within the carton is analyzed from the acquired 3D X-ray images to determine the cut type, meat grade, weight, fat thickness and 3D location of each retail cut within the carton.

[0302] Reference FIG. 16AThe present specification provides for the use of image sensors such as 2D X-ray tomography scanners 1604 with dual-energy or multi-energy sensors placed in a robotically controlled rotating gantry or fixed gantry imaging geometry. In embodiments, the 2D X-ray tomography scanners 1604 can be used with or without motion correction methods depending on the application in which the scanning technology is deployed. In various embodiments, the 2D X-ray tomography scanners 1604 can be used in various applications in a meat production plant. In an embodiment, the scanners are used to perform meat grading scans, typically for beef grading; where the X-ray scan data is analyzed to check meat quality, rib eye muscle area, intermuscular fat thickness, intramuscular fat content, fat marbling, and surface fat thickness. In an embodiment, the 2D X-ray tomography scanners 1604 are used to perform analysis of cross sections through a trim carton to determine the average fat-to-lean ratio of retail meat cuts and the average trim component size (fat and lean) within the cut.

[0303] FIG. 63 A line graph showing various stages of performing a scan and subsequent image analysis to generate data indicative of the quality of the meat ("meat grade") is shown in accordance with some embodiments of the present specification. In embodiments, the scan is performed using a tomography scanner 6300. At stage 6302, a carcass is loaded onto the scanner 6300 and the scan is initiated. In some embodiments, the process of stage 6302 takes about 3 seconds. At stage 6304, the scan is in progress. In some embodiments, at stage 6304, about 4 seconds have elapsed until about the midpoint of the scan. At stage 6306, the scan is complete. In some embodiments, at stage 6306, about 5 seconds have elapsed. At stage 6308, the carcass is unloaded from the scanner 6300. In some embodiments, at stage 6308, about 8 seconds have elapsed. At stage 6310, the scanner returns to the initial loading position (of stage 6302). In some embodiments, at stage 6310, about 9 seconds have elapsed. Thus, in some embodiments, all stages of a meat grading scan take about 9 seconds to complete. In some embodiments, the tomography scanner 6300 has a throughput of 400 faces / hour and produces reconstructed pixels of 0.5 mm x 0.5 mm.

[0304] In some embodiments, the system of the present specification applies a plurality of programming instructions to evaluate the X-ray images and generate data representative of the quality of the meat, wherein the quality is quantified by a first range of values; generate data indicative of intramuscular fat (IMF) deposition and / or content, which data is associated with a second range of values; and generate data indicative of the degree of marbling of the meat, which data is associated with a third range of values. In some embodiments, one or more camera systems are installed for cut location and meat color imaging.

[0305] The present specification also provides for the use of image sensors such as 2D projection X-ray imaging in single view or dual view configurations with dual energy or multi-energy X-ray sensors. In various embodiments, 2D projection X-ray imaging can be used in various applications in a meat processing plant. In one embodiment, the imaging is used to analyze internal organs after they are removed from the carcass into a tray, where each carcass produces one tray of green organs (e.g., stomach, small intestine, and large intestine) and one tray of red organs (e.g., heart, lungs, liver, kidneys). In embodiments, the X-ray system is used to find foreign objects such as metal objects and beetle nodules and health defects such as tumors, cysts, and swollen organs. In one embodiment, the imaging is also used to analyze cartons containing trimmings to determine the fraction of lean tissue to fat tissue averaged over the entire carton.

[0306] In some embodiments, the present specification provides for the use of image sensors (e.g., 2D projection X-ray imaging in single view and / or dual view configurations) with dual energy or multi-energy X-ray (MEXA) sensors to detect the presence of beetle nodules in scanned meat. It should be appreciated that the programming instructions are configured to process the X-ray image data to identify at least one of shape, attenuation values, clustering, density, or other values indicative of one or more beetle nodules. In some embodiments, the X-ray system uses a conveyor positioned on an upwardly sloped surface with a first end of the conveyor at a lower elevation position than a second, opposite end of the conveyor or a conveyor positioned on a downwardly sloped surface with a first end of the conveyor at a higher elevation position than a second, opposite end of the conveyor to minimize radiation dose and overall system size. In some embodiments, the X-ray system provides an inkjet, laser beam, LED strip, or augmented reality headset to indicate the presence of beetle nodules.

[0307] Reference is made to FIG. 16AThe present specification provides for the use of hyperspectral and fluorescence scanners 1606 operating on mid-infrared wavelengths in the range of 5,000 nm to 2,000 nm; short wave infrared wavelength range of 2,000 nm to 900 nm; near infrared wavelengths in the range of 900 nm to 800 nm; visible light wavelength range of 800 pm to 400 nm; and ultraviolet wavelengths in the range of 400 nm to 100 nm. Contrast between tissues varies as a function of wavelength, as a function of healthy versus diseased tissue, and as a function of contaminated versus clean tissue. In embodiments, ultraviolet, broadband visible, and infrared light can be used to illuminate the inspected offal / other meat products for reflectance image formation and analysis. In embodiments, the hyperspectral and fluorescence scanners 1606 can be used for contaminated carcass scale analysis (e.g., for detecting fecal matter after removal of the hide). In embodiments, the hyperspectral and fluorescence scanners 1606 can be used for analysis of the downer (offal) after removal from the carcass into a tray, which typically includes one tray of green offal (e.g., stomach, intestines, and large intestine) and one tray of red offal (e.g., heart, lungs, liver, kidneys). In embodiments, the hyperspectral and fluorescence scanners 1606 are used to determine a range of health defects in the meat products, such as, but not limited to, tumors, cysts, inflammation, rashes, and infections.

[0308] Reference is made to FIG. 16A The present specification provides for the use of a camera system 1630 operating in the visible wavelength range of 800 nm to 400 nm and the short wave infrared wavelength region range of 2,000 nm to 900 nm. In embodiments, the camera system is used for: object tracking along the conveyor belt and between the overhead conveyor system, automatic cutting system, and horizontal conveyor system; positioning data for positioning carcasses and primal meat within other inspection systems such as X-ray scanners and providing data for motion correction algorithms that can be needed; human factor and time motion study analysis to provide optimal production efficiency and optimal quality operator cutting procedures; and thermal imaging to analyze knife cutting methods for producing retail meat cuts from primal meat.

[0309] Reference is made to FIG. 16A The present specification provides for the use of a handheld device 1608 including an RFID and / or barcode reader and / or camera 1630. In various embodiments, the RFID and / or barcode reader and / or camera 1630, which can be handheld or fixed, is used for: tracking carcasses, offal, primal meat, retail meat cuts, trim containers, packaged products, and cartons or product within a production facility; and performing quick lookups of data associated with the barcode and / or RFID tag readings regarding carcasses, offal, primal meat, retail meat cuts, trim containers, packaged products, or product cartons.

[0310] In various embodiments, various different types of sensors and applications can be used in the slaughterhouse of a meat processing plant, such as, but not limited to, fixed equipment of 3D camera systems; radar ranging systems for determining carcass volume, meat grading, and meat color; handheld systems for measuring temperature, pH, color, contamination, and other parameters. Such and other sensors can be integrated within the overall framework disclosed in the present specification, without departing from the scope of the present specification, to further improve the efficiency and profitability of a meat processing plant.

[0311] In an embodiment of the present specification, each carcass processed in a meat processing plant, each primal cut from the carcass, and each subsequent retail meat piece from each of the primal cuts is provided with a unique identifier (ID) to ensure traceability of all products. For example, if a carcass enters the slaughterhouse chill room of a meat processing plant with an ID of "63" and six primal cuts are subsequently cut from the carcass, the primal cuts can be provided with IDs such as "63:1" through "63:6". If primal cut "63:1" is then processed into 26 retail meat pieces, the meat pieces can be provided with IDs such as "63:1:1" through "63:1:26". If primal cut 63:2 is processed into 15 retail meat pieces, the meat pieces can be provided with IDs such as "63:2:1" through "63:2:15". The IDs of the retail meat pieces from the remaining primal cuts of carcass ID "63" can be similarly provided. It will be apparent to those skilled in the art that multiple carcass, primal cut, and retail meat piece labeling schemes are possible and can be used in the present specification, and the example given above is merely one of these labeling schemes. In various embodiments, the IDs generated for carcasses, primal cuts, and retail meat pieces are also associated with the date and time stamp of when a primal cut or retail meat piece was cut from a carcass or separated from its primal cut.

[0312] In another embodiment, the present specification provides a method for tracking the location and time or arrival of each carcass, primal cut, and retail meat piece through a meat processing plant.

[0313] FIG. 17is a flowchart illustrating steps of assigning a carcass ID to track the location and time or arrival of each carcass through a meat processing plant according to embodiments of the present specification. At step 1702, each slaughter hook (in the meat processing plant) on which a carcass is hung on a moving rail is associated with an RFID tag and / or barcode. At step 1704, each animal is fitted with an RFID ear tag or other ID providing element upon arrival at the slaughterhouse (of the meat processing plant). At step 1706, each animal is slaughtered and the corresponding carcass is hung on a slaughter hook. At step 1708, the animal specific ID is directly associated with the slaughter hook RFID tag and / or barcode to generate a carcass ID. This ensures traceability of the animal to the corresponding carcass. In embodiments, the carcass ID is a combination of the slaughter hook ID and the animal ID to facilitate tracing of the carcass original ID to the farm where the animal was produced. In embodiments, a date and time stamp is included as a component of the carcass ID to facilitate human sorting of slaughterhouse data.

[0314] In an embodiment, the present specification employs camera technology to track primal meat cut from a carcass and transferred to a conveyor or secondary hanging rail. FIG. 18is a flowchart showing steps for assigning a carcass ID for tracking location and time when raw meat or retail meat pieces are obtained from a carcass by a meat processing plant according to embodiments of the present specification. At step 1802, each cut scene in the meat processing plant is viewed by one or more cameras and video data from the one or more cameras is processed in real time to determine when a new raw meat or retail meat piece is first separated from its originating carcass or raw meat. At step 1804, a new raw meat or retail meat piece ID is generated upon detection of the separation. At step 1806, after the separation, the one or more cameras continue to track the raw meat or carcass until it is placed on an adjacent conveyor or hook to be transported to the next processing step. At step 1808, it is determined whether the raw meat is secured to a hook. At step 1810, if the raw meat is secured to a hook, the raw meat ID is associated with the hook RFID and / or barcode. At step 1812, it is determined whether a remaining portion of the raw meat is removed from the hook. At step 1814, if a remaining portion of the raw meat is removed from the hook, the raw meat ID is transferred to a subsequent conveyor or waste chute of the meat processing plant. At step 1816, it is determined whether the raw meat or retail meat piece is placed on a conveyor. At step 1818, if the raw meat or retail meat piece is placed on a conveyor, the raw meat or retail meat piece ID is associated with the adjacent RFID tag and / or barcode embedded in the conveyor. In an embodiment, the conveyor ID is placed on the conveyor at 100mm-200mm intervals so that the location of each raw meat or retail meat piece on the conveyor is easily identifiable. At step 1820, it is determined whether the raw meat or retail meat piece is transferred from one conveyor to another. At step 1822, if the raw meat or retail meat piece is transferred from one conveyor to another, the conveyors are designed to automatically transfer the raw meat or retail meat piece ID directly from one conveyor to the next, by using camera tracking of the product in transition between conveyors to ensure that the product ID is accurately transferred from one conveyor to another. In an embodiment, if more than one retail meat piece is placed side-by-side on a conveyor such that multiple meat pieces are associated with the same conveyor barcode or RFID tag, camera tracking is used to determine the lateral position of each meat piece on the conveyor at the point the meat pieces are loaded onto or removed from the conveyor.

[0315] In an embodiment, for the point at which an operator lifts or otherwise removes a raw meat or product from a track or conveyor to enter a subsequent processing step (e.g. trimming fat from a raw meat or packaging a product), one or more cameras are used to monitor the location of the product and any portions that can be cut from it to maintain product location and ID assurance. FIG. 19is a flowchart showing the steps of assigning a carcass ID to track the location of a carcass / raw meat / retail meat piece removed by a human operator through a meat processing plant according to embodiments of the present specification. At step 1902, the trimmings from a raw or retail meat piece are placed in a trim bin containing trimmings from multiple raw and / or retail meat piece items. At step 1904, the unique ID of each product placed in the trim bin is recorded against the unique RFID and / or barcode ID of the bin. At step 1906, the trimmings from multiple smaller bins are aggregated into a single larger bin. At step 1908, the RFID and / or barcode of the larger bin is linked to the RFID and / or barcode of the multiple smaller bins emptied into it. At step 1910, the bin RFID and / or barcode data of the larger bin is also associated with the product data from each of the multiple smaller bins in order to maintain tracking of the items from each initial carcass. At step 1912, after the bin is emptied, any product associations are deleted from the database records associated with the bin so that when the bin is refilled it can be associated with corresponding new product IDs.

[0316] In embodiments, at the point where an automated processing device such as a rotary blade, band saw, draw equipment or water jet cutter removes or modifies a carcass, raw meat or retail meat piece, photographic data before and after is recorded and associated with the initial and final product for quality assurance purposes. In the case where the automated processing device moves the carcass, raw meat or retail meat piece from one location to another, the carcass, raw meat or retail meat piece ID is automatically transferred from the initial location to the final hook or conveyor location.

[0317] In embodiments, at every point in the meat processing plant, a sensor scans a carcass, raw meat, retail meat piece or packaged product, the carcass, raw meat, retail meat piece or packaged product ID is directly associated with the data produced by the sensor to allow instant recall of data from that sensor through the data network (e.g. 1628 in FIG. 16A for retrospective analysis and real time analysis through computer algorithms providing added value to the entire production process.

[0318] Multi-sensory imaging system / platform

[0319] In some embodiments, the present specification describes a multi-sensor imaging system / platform designed to use 2D (two-dimensional) projection X-ray imaging in single view or dual view configurations, where dual-energy or multi-energy X-ray (MEXA) sensors are combined with hyperspectral imaging for internal inspection and classification of organs. In some embodiments, the X-ray system uses a conveyor positioned on an upwardly sloped surface, where the first end of the conveyor is at a lower elevation than the second, opposite end of the conveyor, or a conveyor positioned on a downwardly sloped surface, where the first end of the conveyor is at a higher elevation than the second, opposite end of the conveyor, to minimize radiation dose and overall system size.

[0320] In some embodiments, the multi-sensor imaging system / platform combines multi-energy X-ray attenuation (MEXA) with visible light and short-wave infrared (SWIR) hyperspectral camera data and applies multiple programmed codes, instructions, or algorithms to automatically detect and sort defective bovine and ovine organs in a slaughterhouse. Hyperspectral data provides detailed information about the surface, while X-rays penetrate the tissue to provide information about the interior of the organ. In some embodiments, the multi-sensor imaging system / platform provides inkjet, laser beam, LED strip, or augmented reality headset to indicate the presence of health issues while scanning the meat.

[0321] The present specification describes a multi-sensor platform and associated multiple programmed codes, instructions, or algorithms to process X-ray scan data and hyperspectral imaging data for detecting defects in animal tissue, particularly bovine and ovine organs. It will be appreciated that, to collect data, normal and abnormal (where abnormal is diseased or sick) organs are obtained from a slaughterhouse, scanned by the multi-sensor system, and histopathologically examined by expert veterinarians. The collected data is then used to develop various algorithms for automatically detecting abnormal organs using various machine learning and deep learning algorithms (supervised and unsupervised). Using hyperspectral imaging data, defects in bovine and ovine organs are automatically identified with up to at least 92% accuracy. In embodiments, the multiple programmed codes, instructions, or algorithms can be used to automatically “flag” organs that are defective after classification, or generate images (which can be but are not limited to RGB, X-ray, and / or hyperspectral) with color or otherwise distinguished areas where abnormalities are detected, which can assist an inspector for further inspection. The multiple programmed codes, instructions, or algorithms are configured to generate at least one graphical user interface (GUI) to display the images and apply color or other demarcations (e.g., dot painting) to areas of the images to indicate that these areas contain one or more abnormalities.

[0322] Additionally, the plurality of programmed code, instructions, or algorithms analyze the target X-ray scan data to determine if the organ of interest (i.e., abnormal thickness or discoloration upon palpation) is too dense compared to healthy organs within a library of X-ray images (stored in a database). In some embodiments, each X-ray image in the library has associated thickness and density data. The plurality of programmed code, instructions, or algorithms are configured to process the associated thickness and density data to determine an appropriate threshold for accepting or rejecting the target X-ray image as containing healthy or unhealthy meat / organ, respectively. In some embodiments, the plurality of programmed code, instructions, or algorithms are further configured to identify a disease in the target X-ray scan data based on a library or database of X-ray images containing markers with information from several pathologies.

[0323] The multi-perception platform of the present specification provides improved visceral throughput, including optimization of rollers and protective lead shielding, and automation of image analysis to observe and inspect scanned organs in real-time and determine if a second scan is needed, and allows for longer projection times in order to scan more than one organ consecutively. The inter-roller spacing, intensity, and size can be optimized based on the weight and / or distribution of the scanned viscera. Thick viscera reduces the detected scan signal (due to more attenuation), and therefore reduces the statistical accuracy (i.e., more noise). In such cases, the multi-perception platform is configured to scan at a slower speed (longer time) in order to improve image / detection accuracy and precision.

[0324] FIG. 22 Perspective view 2202 and block diagram 2203 of a multi-perception imaging system / platform 2200 are shown, in accordance with some embodiments of the present specification. In some embodiments, system 2200 includes the following features described below.

[0325] In some embodiments, system 2220 includes a conveyor belt 2204 that translates at a speed in the range of 0.1 m / s to 1.0 m / s, and preferably at a speed of about 0.2 m / s.

[0326] In some embodiments, system 2200 includes an inspection tunnel 2206 having a length in the range of 1100 mm to 5000 mm, a width in the range of 500 mm to 1000 mm, a height in the range of 300 mm to 1000 mm, and preferably dimensions of 1360 mm long x 630 mm wide x 400 mm high.

[0327] In some embodiments, system 2200 includes a dual view x-ray scanning system 2210 including a first and second x-ray source each at 160 keV (with the first source in an overhead configuration and the second source in a lateral configuration). In some embodiments, system 2210 has 10 to 42 DABs, and preferably, 6 to 22 DABs for the overhead view and 4 to 20 DABs for the lateral view. In one embodiment, system 2210 has 20 data acquisition boards (DABs), specifically, 11 for the overhead view and 9 for the lateral view (112 pixels per board).

[0328] In some embodiments, system 2210 includes a high spatial resolution multi-energy photon counting x-ray sensor array, such as a cadmium telluride detector (CdTe: 0.8 mm x 1.2 mm x 2 mm).

[0329] In various embodiments, the x-ray imaging acquisition rate of system 2210 is in the range of 150 Hz to 500 Hz in 3 to 20 energy bands in the 20-160 keV range. In some embodiments, the x-ray imaging acquisition rate is 300 Hz in six energy bands in the 20-160 keV range.

[0330] In various embodiments, system 2210 includes a hyperspectral imaging system 2215 including a camera sensor. In some embodiments, the camera sensor includes a visible / IR (infrared) sensor operating in the wavelength range of 450 nm - 900 nm. In various embodiments, the camera sensor is configured for imaging in 200 to 1200 wavelength bands. In one embodiment, the camera sensor is configured for imaging in 300 wavelength bands. In some embodiments, the camera sensor includes a SWIR (short wave infrared) sensor operating in the wavelength range of 900 nm - 1700 nm. In various embodiments, the camera sensor is configured for imaging in 400 to 700 wavelength bands. In one embodiment, the camera sensor is configured for imaging in 512 wavelength bands. In some embodiments, the hyperspectral imaging acquisition rate ranges from 30 Hz - 150 Hz depending on the image resolution / size and scaled to the x-ray image capture.

[0331] The X-ray and hyperspectral imaging systems 2210, 2215 are in data communication with a computing device having a memory, an associated database system, and a controller / processor that implements a plurality of instructions, programming code, or algorithms (e.g., Ubuntu (Linux) Cube computer program) that are configured to control exposure times, image sizes, and acquisition rates, as well as perform various analyses of the X-ray images and / or hyperspectral images in order to identify abnormalities, diseases, and types of meat / organs / offal, classify healthy and unhealthy meat, and implement related functions and features as described in this specification.

[0332] The multi-sensing imaging system 2200 is configured to allow samples to be loaded from a first end, pass through the scanner, and emerge from a second end. In addition, the system 2200 features real-time energy and intensity calibration of the multi-energy X-ray sensor array, integration of two hyperspectral cameras on two cameras near full GigE bandwidth, synchronized storage to disk and calling of MEXA, DICOM (Digital Imaging and Communications in Medicine) format visible and SWIR camera data, and related TDR (Threat Detection Report), and a unified graphical user interface (GUI) for simultaneous detailed viewing of all image types.

[0333] In some embodiments, the system 2200 includes a number of features as described below. In some embodiments, general features include: a) the sensing system is designed to operate in a sanitary abattoir environment; b) the sensing system is washdown resistant; c) the sensing system is designed to meet food safety standards; d) the sensing system is designed to meet ARPANSA (Australian Radiation Protection and Nuclear Safety Agency) radiation safety requirements; e) the sensing system has a tunnel size of 630 mm (W) x 430 mm (H); and / or f) the sensing system has a conveyor speed of 200 mm / s.

[0334] In some embodiments, the system 2200 has the following imaging properties or specifications: a) the system is configured to implement dual view X-ray imaging. One view passes up through the center of the inspection region. The other view points horizontally through the inspection region; b) the X-ray imaging views use a 120 to 160 keV X-ray beam quality with 0.2 to 1.25 mA beam current (e.g., for low dose, low radiation exposure settings, such as for light curtain or no curtain mask, use 120 keV, 0.2 mA); c) the X-ray imaging views use a multi-energy X-ray (MEXA) sensor with 0.8 mm pitch sensor elements, where each sensor element counts photons to a resolution of up to 106 X-rays / mm2 2d) the visible wavelength hyperspectral imaging sensor operates in the range of 400 nm to 900 nm with a spectral resolution of at least 20 nm across the spectral region, with a pixel size no more than 2.0 mm across the width of the conveyor; e) the short wave infrared (SWIR) hyperspectral imaging sensor operates in the range of 900 nm to 1800 nm with a spectral resolution of at least 20 nm across the spectral range, with a pixel size no more than 2.0 mm across the width of the conveyor; f) in various embodiments, the X-ray, visible light, and SWIR camera systems are synchronized with an X-ray base frequency ranging from 150 Hz to 500 Hz. And in particular, the X-ray, visible light, and SWIR camera systems are synchronized with an X-ray base frequency ranging from 300 Hz; g) the X-ray scan data from the X-ray scanning system 2210 and the hyperspectral imaging data from the hyperspectral imaging system 2215 are transmitted to a computing device for subsequent real-time visualization (via one or more images displayed in one or more graphical user interfaces) and analysis using a plurality of programmed codes, instructions, or algorithms.

[0335] In some embodiments, the system 2200 has the following software features or configurations, which are implemented by a plurality of programmed codes, instructions, or algorithms. In embodiments, the computing device associated with the multi-sensor imaging system 2200 generates at least one graphical user interface (GUI) that provides a pass / fail risk indication for all offal items to the system operator. In embodiments, the at least one graphical user interface includes a scrolling image to display the offal currently in the X-ray tunnel along with superimposed inspection results from the automated health screening algorithm. Where available, the offal data is associated with the animal ID using RFID, QR code, barcode, or other similar ID technology by linking the scanner and the central abattoir database. The computing device associated with the system 2200 provides image review tools for retrospective analysis of offal samples, including X-ray manipulation and hyperspectral data manipulation tools. The software complies with relevant cybersecurity standards, such as ISO 27001.

[0336] In some embodiments, the system 2200 has the following algorithmic features or configurations. The system 2200 is configured to apply a plurality of programmed codes or instructions to combine the X-ray and hyperspectral image data to identify each type of offal as it passes through the scanning system. The target performance is at least 90% correct classification. The system 2200 is configured to apply a plurality of programmed codes or instructions to provide a risk assessment for each offal item as it passes through the scanning system, where the risk assessment indicates a probability of each offal item being healthy or unhealthy. In some embodiments, the data indicative of the risk assessment is associated with a first range of values for healthy offal and a second range of values for unhealthy offal. The system 2200 is configured to apply a plurality of programmed codes or instructions to combine the image-derived information with other abattoir-provided information, such as animal type, age, sex, and breeding data, when available, in order to maximize risk prediction accuracy. The system 2200 is configured to apply a plurality of programmed codes or instructions to generate a total risk score as a total of all underlying algorithm risk score results. The total risk score is used to generate a pass / fail result, which is further applied to apply a color to the X-ray and / or hyperspectral images of the particular offal item associated with the result for display in at least one graphical user interface.

[0337] In some embodiments, the system 2200 is configured to implement the following integrations. The system 2200 is configured to interface with an abattoir database system to call information about a particular carcass and to store pass / fail information for each offal item of each carcass. The system 2200 is configured to mechanically integrate with an abattoir conveyance system to pass offal items through the X-ray scanning and hyperspectral imaging system 2210, 2215 in a controlled manner. The system 2200 is configured to interface with a subsequent robotic system for automated offal selection and rejection.

[0338] Sample illumination

[0339] Referring back FIG. 22 To obtain good quality image data from the hyperspectral imaging system 2215, it is preferable to provide sufficient illumination at all wavelengths in the imaging region (which is otherwise dark). Various imaging light sources were evaluated, including white LEDs (light emitting diodes), halogen lamps, and QIR (quartz infrared) heat lamps, which are commonly used for food counters to keep food warm. To evaluate the emitted wavelengths from each light source, a white reflective card was placed in the X-ray scanning tunnel 2206 to reflect light from the illumination source back to the line-scan geometry hyperspectral imaging camera. A first plurality of curves 2302 and a second plurality of curves 2304 indicative of the results of these measurements are shown in FIG. 23 for the visible wavelength camera, and in FIG. 24 for the SWIR camera. FIG. 23A FIG. 23B

[0340] ​​In the SWIR region, the QIR source is nearly an order of magnitude brighter than the halogen lamp, and there is no illumination from the LED in this region. Thus, the QIR light source is employed for broadband illumination tasks. Given that the QIR light source is very efficient at producing heat, the X-ray scanner control system is modified so that the QIR light source is configured to turn on only when the X-ray beam is on. This limits the heating in the scan tunnel 2206 to only those seconds when the scan is actually being performed.

[0341] Synchronous capture

[0342] A series of tests are then performed using the X-ray absorption and optical reflection barcode patterns to verify that the correct X-ray data is associated with the correct visible light data, and that these are all associated with the correct SWIR data. FIG. 24 An example barcode is shown that is scanned simultaneously using the X-ray, visible light, and SWIR sensors. Alignment between the various sensors can be observed from the image alignment of the X-ray data 2402, visible light data 2404, and SWIR data 2406.

[0343] X-ray scanning and calibration

[0344] FIG. 25A and FIG. 25B MEXA X-ray image data 2502, 2504 are shown for a center numerically on-axis view FIG. 25A ) and for a side-on view FIG. 25B ) respectively. The left side of both figures is a standard X-ray test piece 2500 used in the aviation industry for image quality performance verification. The right side of both figures is an X-ray image 2502, 2504 of beef offal that was packaged and purchased from a local wholesale butcher shop. Calibration systems and methods can be employed to eliminate the vertical streaking artifacts and provide the necessary qualitative data to calculate the effective atomic number integrated along each line from the source to the detector.

[0345] Preliminary offal characterization

[0346] According to some embodiments, lamb offal (heart, liver, and lung) is acquired from a local butcher shop and X-ray image data is acquired as shown in FIG. 26 The sample 2600 is first imaged while still fresh to produce a first image 2602, and then imaged again after a few days of decay to produce a second image 2604.

[0347] In images 2602, 2604, the heart is identifiable in the X-ray data as distinct from the liver and lungs, and thus, an automated algorithm is configured to identify the heart in images 2602, 2604. In embodiments, a combination of vertical and horizontal view X-ray data is analyzed by a plurality of programmed codes or instructions to distinguish the lungs and liver and determine the thickness of the tissue to calculate the density and effective atomic number at each location in the image with reasonable accuracy. Over time, the loss of image contrast is due in part to blood leaching out of the organs.

[0348] Optimizing capture performance

[0349] As part of the imaging optimization, the optimal operating conditions for the multi-sensor system 2200 are determined as follows (in various embodiments as shown in Table 1).

[0350] Table 1

[0351]

[0352] With these optimized settings, in FIG. 27 synchronous first image data 2702, second image data 2704, and third image data 2706 for the MEXA, visible, and SWIR sensors, respectively, are shown. The MEXA data 2702 covers the entire field of view. The visible hyperspectral data 2704 covers 75% of the field of view, while the SWIR hyperspectral data 2706 covers 50% of the field of view. The limited fields of view for the two hyperspectral cameras are due to bandwidth constraints in the Gigabit Ethernet camera interface to the host computing device. In embodiments, the spectral resolution can be reduced to lower the camera output data rate and allow full field of view imaging.

[0353] Hyperspectral imaging performance

[0354] Image alignment

[0355] To verify the synchronization of data acquisition between the two hyperspectral cameras, a simulated pattern was developed for playback from each camera. In this case, one camera outputs magenta data and the other camera outputs green data. When perfectly aligned, the results will total white and / or will otherwise result in magenta or green leading or trailing pixels.

[0356] The results of this pattern 2800 for a hyperspectral imaging system with poor synchronization are shown in FIG. 28 Here, several green and magenta regions can be seen. In a perfectly synchronized system, all of the test bars appear white.

[0357] Image review

[0358] Hyperspectral image data from a lamb liver is shown in FIG. 29Data from the visible hyperspectral imaging camera 2902 is rendered as colored with a pseudo-color scale (from blue to red). Data from the SWIR hyperspectral camera 2904 is also shown. There is strong contrast between the fatty regions in the center of the image and the lung tissue regions at the periphery of the image.

[0359] Beef organ scans

[0360] All organs were collected from a cooperating slaughterhouse or butcher shop (approximately 30 km from the scanning and pathology lab), frozen on ice (2°C) for transport from the collection site to the lab, and scanned with the multi-sensory platform 2200 within 1.86 ± 0.16 days post-slaughter. Subsequently, all organs were examined by an experienced veterinary pathologist (gross and histologically) during a post-mortem examination to confirm abnormalities within 0.81 ± 0.12 days post-scanning. A total of 126 organs were collected as follows:

[0361] • 95 livers, 43 healthy and 52 not healthy

[0362] • 17 kidneys, 13 healthy, 4 not healthy at the slaughterhouse, but considered healthy at the post-mortem examination

[0363] • 14 lungs, 12 healthy and 2 not healthy, but only 1 was confirmed not healthy during the post-mortem examination

[0364] Livers were the most commonly affected organ and therefore provided the strongest dataset for algorithm development.

[0365] A total of 52 beef livers that were considered unfit or rejected for human consumption by the meat inspector were collected from a cooperating slaughterhouse. The organs were processed and stored at 2°C until scanning.

[0366] A total of 43 beef livers that were considered fit for human consumption were collected from a commercial sales point and stored at 2°C until scanning.

[0367] The organs were scanned using the multi-sensory platform 2200 (including multi-energy X-ray attenuation at six energy levels, as well as visible and shortwave infrared hyperspectral imaging), followed by gross examination, and subsequently histological examination by a veterinary pathologist to confirm abnormalities.

[0368] Organ scans

[0369] To scan the organs individually, the livers were placed into the opened containment bag and then scanned using the multi-sensor scanning system 2200 of the present specification. Each liver sample was positioned with the diaphragmatic surface facing up, the caudate lobe on the left bottom side, and then scanned following the standard protocol. A total of six radiographs were produced simultaneously for each of the six X-ray energy bands. Processing of the samples followed the standard PC2 workflow. Normal RGB (red green blue) images were also obtained at the scanned locations, and regions of interest were recorded for later image labeling.

[0370] Representative image results and spectral signals are shown in FIG. 30A to 30D . FIG. 30A A first RGB image 3002 of the bovine liver is shown, which shows the macroscopic aspects of the organ before scanning, a second X-ray image 3004 of low energy and a third X-ray image 3006 of high energy irradiation of the bovine liver. A red circle 3008 encloses a liver fluke found at histopathological analysis. The veterinary inspection showed that the organ had multiple nodules and flukes.

[0371] FIG. 30B A fourth RGB image 3010 and a corresponding fifth SWIR hyperspectral image 3012 are shown. FIG. 30C A sixth SWIR hyperspectral image 3014 and a corresponding spectral signal 3016 are shown. Box annotations 3018 mark the diseased areas. FIG. 30D A seventh visible hyperspectral image 3020 and a corresponding spectral signal 3022 are shown. Box annotations 3024 mark the diseased areas.

[0372] Postmortem inspection

[0373] All organs were systematically inspected for gross lesions by a qualified veterinarian specialized in pathology. If lesions were present, the following data were recorded: location, distribution, demarcation, color, shape, appearance and consistency of the meat surface. The most likely cause of the lesion was also recorded. The organs were also inspected for color abnormalities and texture inconsistencies by palpation, in some cases sectioned and sampled for histopathology to confirm the identity of the lesion. During this process, photographs were taken of the abnormal findings. After each postmortem inspection was completed, the findings were recorded.

[0374] Image processing

[0375] Machines and deep learning models are generally sensitive to the distribution of the dataset, and data variations or redundancies can lead to a decrease in the performance of the deep learning model. To mitigate the negative effects of interference or redundant information from the visible HS (hyperspectral) images (e.g., image 3100), the following are performed as FIG. 31The pre-processing operations shown include manual removal of irrelevant regions and selection of regions of interest (ROI) at step 3102, band filtering at step 3104, and value normalization at step 3106.

[0376] In step 3102, first, the regions outside the tray area are excluded to avoid misleading the deep learning model’s attention to these irrelevant regions, resulting in false predictions and classifications. The ROI is manually selected because the data is complex and insufficient. In addition, the non-meat tissue component (conveyor belt) is much higher than the meat component (organs and iron plate) in the original image. Therefore, the ROI is manually selected to speed up the training (time constraint) and obtain good results on small data. In the subsequent commercial scenario, there is no need to develop a specific automated preprocessing program, but only to fix the location of the iron plate containing meat when scanning to complete the ROI segmentation.

[0377] Second, at step 3104, unique spectra with high signal-to-noise ratio (SNR) are selected. In the case of images, SNR is the ratio of the pixel mean to the image variance. Because information usually presents a specific pattern or figure, it has a small variance, i.e., a high signal-to-noise ratio. Since different image bands do not carry equally important information, and some bands have high noise and low SNR, this poses a major challenge to distinguishing the overall outline of the organs. These randomly distributed noise data can be very destructive to the performance of the model’s classification and prediction.

[0378] The last step 3106 is to normalize the image values to normalize the intensity of each image band within a fixed range and maintain the intensity ratio between channels. The image pixel values are normalized to [0, 1] L2 norm, where the most important pixel value is mapped to 1.

[0379] Subsequently, the entire dataset is divided into a training set and a test set in a ratio of 2:1 to develop and evaluate the prediction or detection model, respectively.

[0380] Statistical analysis and algorithm development

[0381] Deep learning network classification

[0382] Automated inspection using deep learning classification is a viable solution to the challenge of improving economic efficiency and reducing the risk of human infection. In addition, in general, deep learning shows better performance for screening tasks. To develop a deep learning-based classification, the analysis is based on the assumption that abnormal images are caused by abnormal elements (i.e., anomalies) that do not appear in normal images. Since the deep learning network is trained by using a loss function so as to produce the expected result, the loss function is determined according to the assumption that abnormal images of unhealthy meat contain tissues that are not found in normal images of healthy meat. As FIG. 32As shown, the deep learning network 3200 is mainly composed of a discriminator 3202 and a training strategy 3204. The discriminator 3202 is composed of convolutional layers and fully connected layers. The convolutional layers capture spatial information in the image, while the fully connected layers transform the captured information into a prediction result. The discriminator 3202 is trained using the training strategy 3204 to produce the expected result.

[0383] In some embodiments, the deep learning network 3200 includes down-sampling and up-sampling stages. The down-sampling stage compresses spatial information to obtain a larger field of perception, which gives a complete image of the defect in the image. However, as the information is compressed, the network gradually loses the corresponding location information. Although the up-sampling stage can recover the compressed spatial information to find the exact location of the defect, the convolution itself is very sensitive to deformation.

[0384] The discriminator 3202 is trained according to the assumption that only abnormal pixels exist in the abnormal image. Abnormal pixels are conceptually abnormal, i.e., they cause the liver in the image to be abnormal, and it is not necessary to define exactly which pixels are abnormal in advance (unsupervised). The network 3200 automatically defines and discovers abnormalities during the training process. The discriminator 3202 is needed to predict each pixel of a given image and display the result as a heat map. The higher the heat map value, the higher the probability that the feature in the image is abnormal (0 means normal, 1 means abnormal). Therefore, the network 3200 must predict each pixel of a given image and display the result as a heat map. The higher the heat map value, the higher the probability that the feature in the image is a defect (0 means normal, 1 means defect). Although the location of the defect in the image containing the defect can be unknown, the defect in the image should be the maximum value in the image.

[0385] During training, the discriminator 3202 outputs normal organ images. After preprocessing, the images already contain only the liver and the iron plate. Therefore, the heat map of the normal organ can be all zeros and the same color. In contrast, the output of the abnormal image is set to at least one pixel of 1, i.e., the abnormal image should have an abnormal feature. In addition, the training strategy 3204 also calculates the difference between the heat map of the abnormal image and other normal images to ensure that the feature found in the abnormal image is not found abnormal with respect to all normal images. In some embodiments, the training strategy 3204 is to compare all tissues of a meat type (e.g., beef liver) with all corresponding meat types (beef liver) in the normal data set. If the predicted tissue does not appear in the normal data, it is identified as a defect.

[0386] In some embodiments, the training strategy 3204 is configured such that it automatically adjusts the learning rate to adapt to the computed gradients by computing the first and second moment estimates of the gradients. In some embodiments, for a defect screening task, where images are provided as input to the network 3200 to determine whether they contain defects, accuracy, precision, sensitivity, and specificity are computed to evaluate performance.

[0387] K-means clustering for automatic anomaly localization

[0388] Reference is now made to FIG. 33A automatic anomaly detection from SWIR images 3300 based on a k-means clustering algorithm (unsupervised classification). As shown in FIG. 33A the computational and analytical workflow for anomaly detection consists of three main stages, including a pre-processing stage 3302 for band selection and normalization to select regions of interest (ROIs) and reduce noise and intensity bias; a stage 3304 of k-means clustering for normal and abnormal tissues in the liver; and a stage 3306 of analysis of the clustering results.

[0389] Stage 3302: Image pre-processing stage

[0390] Each SWIR image is composed of sub-images of different bands, which can provide much more information than the corresponding RGB image. FIG. 33B A plot 3300b indicating the sum of intensities from each SWIR band of a beef organ is shown in accordance with some embodiments of the present specification. As shown, sub-images with lower total intensities, such as band 20 3310 and band 500 3312, have very low signal-to-noise ratio (SNR), which results in poor quality of the sub-images, while the SNR of sub-images with higher total intensities is high. Therefore, to enhance the SNR of the entire SWIR image, sub-images with total intensities exceeding 50% of the maximum total intensity sub-image are selected, as shown in FIG. 33B .

[0391] The pixel values of the same pixel position in each sub-image will form a pixel vector. However, the range of each pixel vector varies significantly, which makes the distance between different pairs of pixel vectors incomparable. Therefore, band-wise normalization is performed by the following equation:

[0392]

[0393] The normalized pixel vectors 3320 have the same range and become comparable, as shown in FIG. 33C .

[0394] Using k-means algorithm with distance as an indicator of similarity performs poorly in case of high dimensional data, as distance between vectors tends to be closer as dimension increases (curse of dimensionality). To reduce dimensionality, principal component analysis (PCA) algorithm is applied. Component size of 6 is chosen to ensure that the amount of variance explained by the selected components exceeds 97%. In FIG. 33D Reduced SWIR sub-image 3330 is shown in

[0395] Stage 3304: K-means clustering for anomaly detection

[0396] K-means clustering algorithm is used to divide the pixel PCA vectors into K clusters, where each pixel vector belongs to the cluster with the closest average distance to the cluster centroid.

[0397] In finding the best clustering, it contains two steps:

[0398] • Step 1 assigns cluster label i at iteration t:

[0399]

[0400] where each pixel vector is assigned to label i, and is the centroid of the label. ||.|| is the distance between vectors.

[0401] • Step 2 updates the centroid:

[0402]

[0403] The algorithm will converge when the assignment does not change.

[0404] Since K value is a key hyperparameter for k-means algorithm, K value is automatically selected according to within-set sum of squared error (WSS) and used in the model with the K of the elbow on curve 3335, as FIG. 33E shown.

[0405] After the k-means model converges, similar pixel vectors will have the same label, and the image can be segmented based on these labels. However, since k-means is an unsupervised algorithm, it cannot identify whether each cluster is normal or not.

[0406] Stage 3306: Cluster analysis for anomaly localization

[0407] FIG. 33F Cluster merging to detect anomalies in believers from SWIR hyperspectral data is shown in accordance with some embodiments of the present specification. To identify anomalies, a feature analysis of different cluster regions has been further performed. Pixel vectors within each label are randomly sampled for spectral analysis. FIG. 33FThe original mask 3340, the cluster centroid similarity matrix 3342, and the merged mask 3344 are shown. The similarity of each pair of cluster centroids is used to merge similar clusters to reduce over-segmentation, as shown. FIG. 33F

[0408] Since it is challenging to precisely distinguish the boundary between healthy tissue and diseased areas, an erosion process is performed on each label to avoid including labels with lower confidence. Thus, the k-means clustering algorithm produces a flawed localization. It should be appreciated that images with manually sketched defect locations are not used in the training phase of the deep learning network 3200, and instead the defect locations are completely automatically sketched or produced by the network 3200.

[0409] The deep learning network 3200 and associated image analysis method provide at least the following advantages. First, it provides a higher level of automation, mainly in terms of data processing. The deep learning network 3200 does not require a fixed size of ROIs. In prior art systems, the selection of ROIs requires a manual part. The manual selection of ROIs has uncertainty, which can lead to errors in the final prediction results. In some embodiments, the deep learning network 3200 uses a U-net structure, which allows pixel-level prediction with any input size. Thus, the network 3200 and associated image analysis method can reduce the complexity in processing data, and thus have a higher degree of automation. Second, as another advantage, the training strategy 3204 can be configured to perform semi-supervised localization training of defects. Further, the training strategy 3204 can be configured to allow training of the localization task without masks.

[0410] FIG. 64 An intelligent meat production system 6400 according to some embodiments of the present specification is shown. The system 6400 includes a plurality of geographically distributed meat production sites or slaughterhouses 6405a, 6405b, 6405c through 6405n (collectively, the numbers 6405) having associated multi-sensor imaging systems 6410a, 6410b, 6410c through 6410n (collectively, the numbers 6410 and similar to the multi-sensor imaging system 2200). In some embodiments, the plurality of meat production sites or slaughterhouses 6405a, 6405b, 6405c through 6405n also have associated livestock farms or feeders 6435a, 6435b, 6435c through 6435n (collectively, the numbers 6435).

[0411] ​In some embodiments, each of the plurality of meat production sites or slaughterhouses 6405a, 6405b, 6405c through 6405n is in data communication with at least one server 6420 over a network 6430. The at least one server 6420 has an associated database system 6425. In some embodiments, a plurality of end user computing devices 6440 are also in data communication with the at least one server 6420 over the network 6430. In non-limiting scenarios, for example, some of the plurality of end user computing devices 6440 can be co-located with some of the plurality of meat production sites or slaughterhouses 6405a, 6405b, 6405c through 6405n and / or associated livestock farms or feeders 6435, while some of the plurality of end user computing devices 6440 can be geographically distributed away from the plurality of meat production sites or slaughterhouses 6405a, 6405b, 6405c through 6405n and associated livestock farms or feeders 6435.

[0412] In some embodiments, each multi-sensor imaging system 6410 includes a single- or dual-view configured 2D projection X-ray imaging system with dual- or multi-energy X-ray attenuation (MEXA) sensors in combination with a hyperspectral imaging system in data communication with a computing device 6415 and a producer’s database system 6416. Each producer’s database system 6416 stores a plurality of local or site-specific meat production and quality data associated with the associated meat production site or slaughterhouse 6405 and livestock farm or feeder 6435. The plurality of local meat production and quality data includes, for example, but is not limited to, animal ID (corresponding to, for example, an identification tag associated with the animal), animal type (fish, chicken, pig, cow, lamb, etc.), animal breed, X-ray scan data corresponding to each different age of the animal, X-ray scan data of the animal carcass and / or raw meat, hyperspectral image data of the animal meat and organs, geographic location (geographic location of the livestock farm and / or meat production site or slaughterhouse), climate, weather, season, type of feed, time of year of meat production, vaccination history, medications, disease history, animal age (at the time of receiving in the meat production site or slaughterhouse), a plurality of post-sale parameters including lean meat yield (i.e., percentage of meat, fat, and bone), ratio of intramuscular fat to tissue, amount of intermuscular fat, absolute and relative size of individual organs, muscle volume, number of ribs, and presence or absence of diseases such as cysts, tumors, pleurisy, and foreign bodies.

[0413] According to aspects of the present specification, the plurality of local or site-specific meat production and quality data from each producer’s database system 6416 (in each of the plurality of meat production sites or slaughterhouses 6405) is aggregated and stored in a database system 6425 (associated with at least one server 6420) in order to generate a plurality of global meat production and quality data. In some embodiments, the plurality of local or site-specific meat production and quality data from each producer’s database system 6416 is aggregated based on one or more parameters such as, but not limited to, animal type, geographical location, feed type, and / or climate conditions.

[0414] According to aspects of the present specification, the at least one server 6420 implements a plurality of instructions or programming codes representative of at least one machine learning model. In some embodiments, the at least one machine learning model implements modeling techniques such as, but not limited to, partial least squares discriminant analysis, random forest, and artificial neural networks. In some embodiments, the at least one machine learning model implements at least one deep learning or artificial neural network (ANN), for example, a convolutional neural network (CNN).

[0415] In some embodiments, the at least one machine learning model is configured to detect (and thus differentiate) unhealthy / diseased scan data from healthy scan data from the scan data and thus infer and output global best livestock and meat production practices, patterns, and insights to maximize a plurality of positive parameters and minimize a plurality of negative parameters based on the plurality of global meat production and quality data. The plurality of positive parameters correspond to, for example, reduced drug requirements, lower carbon footprint, variable cost efficiency, reputation protection, lower health risks to consumers, and improvements in a plurality of post-sale parameters including lean meat yield, ratio of intramuscular fat to tissue, amount of intermuscular fat, absolute and relative size of individual organs, muscle volume, number of ribs, and absence of diseases such as cysts, tumors, pleurisy, and foreign bodies. The plurality of negative parameters correspond to, for example, presence of abnormalities / diseases such as cysts, tumors, pleurisy, and foreign bodies.

[0416] In some embodiments, the at least one machine learning model is configured to analyze the scan data (for example, any one, all, or any combination of X-ray scan data corresponding to each different age of the animal, X-ray scan data and hyperspectral image data of the animal carcass and / or raw meat) and additional information (for example, animal type, geographical location (livestock farm and / or meat production site or slaughterhouse), climate, weather, season, feed type, time of year of meat production, vaccination history, drugs, disease history, animal age (when received at the meat production site or slaughterhouse), and a plurality of post-sale parameters) to identify and output global best livestock and meat production practices, patterns, and insights that maximize content of a plurality of positive parameters and minimize content of a plurality of negative parameters.

[0417] In some embodiments, the analysis of the scan data with additional information is performed based on local meat production and quality data corresponding to each of the plurality of meat production sites or slaughterhouses 6405 and associated livestock farms or feeders 6435 in order to identify globally best livestock farming and meat production practices, patterns and insights. The globally best livestock farming and meat production practices, patterns and insights identified are then communicated back to the plurality of meat production sites or slaughterhouses 6405 and associated livestock farms or feeders 6435.

[0418] In some embodiments, the input data is used to train at least one machine learning model. In some embodiments, the input data includes at least a portion of the high spectral image data of the health or unhealth / disease of the animal’s meat and organs and the meat production and quality data with associated additional information, such as animal ID, animal type (fish, chicken, pig, cow, lamb, etc.), animal breed, X-ray scan data corresponding to each different age of the animal, X-ray scan data of the animal carcass and / or raw meat, geographical location (livestock farm and / or meat production site or slaughterhouse), climate, weather, season, type of feed, time of year of meat production, vaccination history, medication, disease history, age of the animal (when received at the meat production site or slaughterhouse), a plurality of postmortem parameters including lean meat yield (i.e. percentage of meat, fat and bone), intramuscular fat to tissue ratio, amount of intermuscular fat, absolute and relative size of each organ, muscle volume, number of ribs, and presence or absence of diseases such as cysts, tumors, pleurisy and foreign bodies.

[0419] The high spectral image data and additional information from the meat production and quality data are associated with the animal ID in the database and thus retrievable as training input data. In some embodiments, the training input data represents a sample from the global meat production and quality data stored in the database system 6425. In some embodiments, the sample is representative of each geographical location of the meat production site or slaughterhouse 6405 and associated livestock farm or feeder 6435.

[0420] In some embodiments, the input data is processed in order to generate processed training input data for training the at least one machine learning model. In some embodiments, the processing is applied directly to the input data (including the health or disease scans) and the input data does not require manual annotation or manual designation of health or disease. Thus, the training is based on the assumption that the diseased, unhealth or abnormal images are caused by abnormal elements (i.e. anomalies) that do not appear in the healthy / normal images.

[0421] As known to those of ordinary skill in the art, a hyperspectral image corresponding to the hyperspectral image data includes a plurality of pixels, where each pixel includes a plurality of hyperspectral bands. In some embodiments, the plurality of hyperspectral bands is decomposed into a plurality of bins, where each of the plurality of bins is associated with characterization data. In some embodiments, the characterization data is indicative of hyperspectral reflectance intensity from the surface of the target organ / meat. In some embodiments, the characterization data is further processed by a mathematical function (e.g., differentiation) in order to achieve feature highlighting. Thus, the hyperspectral image is segmented into bins of non-interest and bins of interest (e.g., bins associated with lean meat (meat), fat, bone, healthy organs, and diseased organs). For example, N spectral bands of the hyperspectral image are grouped into a predetermined number of M bins. In some embodiments, the number of bins M is 6. It should be appreciated that in some embodiments, a principal component analysis algorithm sets M = 6 in order to achieve an explained variance ratio (EVR) higher than 97%. However, in alternative embodiments, M can take lower values, e.g., 1, at a lower EVR, and up to 300 for visible light, and up to 512 for SWIR hyperspectral scan data (these values are maximum spectral bins = pixels on the imaging sensor). Thus, the characterization data is associated with one of the M bins. Preferably, the samples are uniformly distributed in the M bins, but can have a population that will vary the orthogonal data set. As a result of the normalization, the samples are uniformly distributed in the M bins such that each of the M bins has the same intensity, and thus the same statistical precision - i.e., each of the M bins carries the same weight in the final result.

[0422] In embodiments, the bins have the following characteristics: a) the bins need to be sufficiently separated or distinguished within the grid space. In addition, the data is presented for a multi-dimensional structure. Thus, a principal component is selected for each of the 3 axes; and b) the bins should have no more than 5% overlap, even more preferably no overlap in volume.

[0423] Each of the M bins corresponds to a separate set of processed training input data. At least one machine learning model is trained using data from the full set of M bins in order to be able to a) detect unhealthy or diseased hyperspectral image data from healthy hyperspectral image data, and b) infer and output global best livestock and meat production practices, patterns, and insights based on a plurality of global meat production and quality data related to the unhealthy / diseased and healthy hyperspectral image data.

[0424] In embodiments, the selection of the hyperspectral bands is designed to improve the quality of the data, enhance the accuracy of the at least one machine learning model, suppress overfitting, and improve efficiency, and ultimately improve the performance of the at least one machine learning model. Peak signal-to-noise ratio (PSNR) is the ratio between the maximum possible power of a signal and the power of disruptive noise that affects its representation, and this is used for band selection for K-means and deep learning models.

[0425] For each sample in the dataset, the hyperspectral image stack with shape (width, height, bands) has intensities ranging from 0 to where n is the intensity resolution. In some embodiments, n = 64. The image in band i is denoted as and the intensity of pixel j is denoted as . The total intensity is calculated using the following equation:

[0426]

[0427] After the total intensity calculation for each band, the maximum total intensity and its corresponding band can be found. The image is used as the reference image to calculate the peak signal-to-noise ratio (PSNR) for band i using the following equation:

[0428]

[0429]

[0430] where is and is and the mean squared error between The PSNR value is used as a threshold for band selection. According to some embodiments, the threshold is chosen to be 20 dB (total intensity is approximately 50% of

[0431] According to certain aspects of the present specification, the at least one server 6420 also implements a plurality of instructions or programming code to generate at least one graphical user interface (GUI) for access by a plurality of end user computing devices 6440 over the network 6430. The at least one GUI is configured to take in user queries. The taken in user queries are provided as input to the at least one trained machine learning model, which processes the user queries and outputs responses for display to the user in the at least one GUI. In some embodiments, the responses are based on global best livestock and meat production practices, patterns, and insights inferred by the at least one machine learning model. In some embodiments, the end user is a livestock farmer or raiser. In some embodiments, the end user is a meat producer.

[0432] In various embodiments, the at least one GUI enables an end user to input query data corresponding to various scenarios into the at least one trained machine learning model to simulate and understand how the scenarios (related to the input query data) would affect a plurality of positive and negative parameters. For example, the end user’s input query data can correspond to the scenario of “what would happen if I invested $3 in a vaccine?” The output response from the at least one trained machine learning model can be “this results in a $20 improvement in health / meat quality outcomes.” Similarly, the end user’s input query data can correspond to what would happen if, for example, no vaccine was used, the type of feed changed, the temperature changed, it rained or any other climate, weather or seasonal changes.

[0433] It will be appreciated that the analysis of the at least one machine learning model, and thus the responses, are highly geography (and climate) specific. Accordingly, in some embodiments, the at least one GUI enables an end user to specify (e.g., via selection of a pre-populated drop down list) a geographic location as well as input query data. In some embodiments, the at least one GUI enables an end user to filter responses based on one or more geographic locations and / or climate, weather and seasonal characteristics.

[0434] Beef organ scan results

[0435] Pathological histological findings

[0436] Table 2 shows the findings recorded during the scanning of diseased livers (meat inspectors rejected for human consumption) and the gross descriptions recorded during the postmortem inspection. Out of the 52 livers rejected for human consumption, a total of 32 showed varying degrees of discoloration, 7 livers had abscesses, 5 had duct thickening, 4 had fibrosis, 2 had flukes and 2 had cysts. The results of the kidneys and lungs were not presented as they were all healthy.

[0437] Table 2 - Postmortem gross descriptions of diseased livers and areas of interest set prior to scanning based on macroscopic aspects of the organs.

[0438]

[0439]

[0440] RGB images and X-ray images at postmortem inspection (collected for each of the six X-ray intensities of the irradiation)

[0441] FIG. 34AAn RGB image 3402 and an X-ray image 3404 of a kidney without macroscopic lesions are shown in accordance with some embodiments of the present specification. It should be noted that the generated images can be corrected after collection (e.g., to remove horizontal streaks) and do not affect the analysis. The clarity of the kidney lobes is also a demonstration of the penetration of tissue using X-ray imaging. FIG. 34B RGB image 3406 and X-ray image 3408 of two kidneys without macroscopic lesions are shown. Fatty tissue from the kidney parenchyma can also be visible, however when comparing FIG. 34A and 34B the images, the detection of discoloration does not appear to be evident.

[0442] FIG. 34C RGB image 3410 and X-ray image 3412 and macroscopic findings of two lungs are shown. Due to discoloration, the cranial lobe is marked 3414 as a region of interest. In FIG. 34D images 3416, 3418 of postmortem examination of the lungs show the cranial lobe of the organ with lesions from a past bout of pneumonia. However, the X-ray image can not be as effective in detecting these health issues as the lungs tissue is full of air and the disease does not cause a change in tissue density. FIG. 34E RGB image 3420 and X-ray image 3422 and macroscopic findings of a lung in accordance with some embodiments of the present specification are shown. FIG. 34F No lesions due to postmortem examination of a beef lung 3424 are shown in accordance with some embodiments of the present specification.

[0443] FIG. 34G RGB image 3426 and X-ray image 3428 and macroscopic findings of a liver with multifocal discoloration in accordance with some embodiments of the present specification are shown. The X-ray highlights the lighter pattern of the liver parenchyma in sections 3430 of the organ compared to healthy liver tissue. The organ was determined to have multifocal discoloration after histopathological analysis. In addition to processing the X-ray images collected for each of the six X-ray intensities irradiated, in some embodiments, a second processing method is employed that uses subtraction between all different energies to determine whether these would enhance the features of interest. FIG. 34H First and second X-ray images 3432, 3434 of a liver obtained from absorption assay data in accordance with some embodiments of the present specification, where the data is obtained by subtracting the high energy from the low energy or using low energy absorption assay data simultaneously. However, this analysis does not appear to provide unique superior performance. During postmortem examination of the liver described above, a lesion extensive pale discoloration was described on the diaphragmatic surface of the organ. This does not appear to be reflected in the X-ray data (3432, 3434), possibly because the effect of the defect on the tissue density was not enough to be picked up by the scanner.

[0444] FIG. 35A An RGB image 3502 and X-ray images 3504 of another liver sample are shown in accordance with some embodiments of the present specification. The RGB image 3502 shows macroscopic aspects of the organ before scanning. The six X-ray images 3504 correspond to the six X-ray energies that were irradiated. A circular marker 3506 encloses a liver fluke found at histopathology analysis. The veterinary inspection shows that the organ has multiple nodules and flukes. The images of the postmortem inspection of the liver show multiple nodules on the diaphragmatic surface and multiple thickened bile ducts and discoloration areas of the lesions. A cross-section of a 10 mm nodule on the left lobe shows a liver fluke, which is also reflected in the X-ray images 3504. As part of the postmortem inspection of the liver, FIG. 35B A first image 3508 showing the organ before dissection, a second image 3510 showing the nodule, and a third image 3512 showing the fluke are shown.

[0445] FIG. 36A An RGB image 3602 and X-ray images 3604 of yet another liver sample are shown in accordance with some embodiments of the present specification. The RGB image 3602 shows macroscopic aspects of the organ before scanning. The six X-ray images 3604 correspond to the six X-ray energies that were irradiated. The postmortem inspection of the liver finds two cysts on the diaphragmatic surface of the organ, one on the left lobe and one at the edge of the right lobe. The contents are liquid and watery. FIG. 36B A first image 3606 of the first cyst and a second image 3608 of the second cyst are shown during the postmortem inspection and histopathology of the liver.

[0446] FIG. 37A An RGB image 3702 and X-ray images 3704 of yet another liver sample are shown in accordance with some embodiments of the present specification. The liver sample has a nodular lesion found on the left lobe, and the region of interest 3706 is marked accordingly. The X-ray scan does not seem to reflect the shape and location of the lesion, although the pattern of the left lobe of the liver looks very uniform and brighter than the right lobe. FIG. 37B An RGB image 3708 of the liver is shown, which shows the large nodule at postmortem inspection.

[0447] FIG. 38A An RGB image 3802 and X-ray images 3804 of yet another liver sample are shown in accordance with some embodiments of the present specification. As can be seen in the image 3802, the liver has a large abscess 3806 on the left lobe of the diaphragmatic surface. The X-ray images 3804 show a lighter gray pattern shadow compared to the right counterpart. However, the nodule does not seem to be clearly reflected in the X-ray images 3804. FIG. 38BAn RGB image 3808 of a liver at postmortem inspection is shown, which shows a first abscess on the left lobe margin and a second abscess to the right of the bile duct. When inspected and incised, each abscess measured approximately 45 mm, showing a cheese content.

[0448] FIG. 39A An RGB image 3902 and an X-ray image 3904 of yet another liver sample according to some embodiments of the present specification are shown. The liver sample showed discoloration, bile duct thickening, and flukes upon inspection. Regions of interest 3906, 3908 are marked with discoloration areas on the left lobe and caudate lobe. The X-ray image 3904 shows the brighter margin of the organ on the entire lower part extending between the two lobes. FIG. 39B An RGB image 3910 of a liver is shown, which shows discoloration, bile duct thickening, and flukes at postmortem inspection. Thus, postmortem inspection of the liver confirmed a 30 x 25 mm irregular discoloration area on the diaphragmatic surface of the left lobe. The visceral surface of the bile ducts looked thickened, and flukes were found in them when cut.

[0449] Deep learning network classification

[0450] For the abnormality classification task, where images are provided to determine whether they are abnormal or normal images, the system and method of the present specification can achieve accuracy and sensitivity measures of over 90% (Table 3), and can also show the location of abnormal pixels.

[0451] Table 3. Model metrics on the test (validation) set (1 / 3 of the total dataset) of diseased and healthy liver samples.

[0452]

[0453] In addition to the binary automatic classification of abnormal and normal, as FIG. 40A shown, the method of the present specification is also configured to generate heat maps 4000 for abnormality detection. Notably, the manually classified abnormal locations are not used at all in the training phase, but are completely automatically generated by the discriminator. Moreover, the deep learning network generates reliable results compared to manual segmentation. For example, the network can detect the location of tumors. In livers 23, 54, 61, and 9, the tumor locations are shown in high contrast in the heat maps and are annotated or marked as green boxes 4002, 4004, 4006, 4008 in the heat maps. However, the network also finds other abnormalities - annotated or marked as red boxes 4010, 4012, 4014, 4016 in the heat maps. These abnormalities do not appear in any normal image, but do appear in abnormal images. Thus, the discriminator of the present specification also identifies them as abnormal tissue.

[0454] K-means clustering for abnormality detection

[0455] First Case Study

[0456] As shown in images 4020, 4022 of liver 9, the liver has an abscess within the yellow bounding box 4024. FIG. 40B SWIR image 4026 and normalized image 4028 of liver 9 are also shown. The k-means clustering results are shown in image 4030, and the merged mask is shown in image 4032. FIG. 40B

[0457] As shown in images 4020, 4022 of liver 9, the liver has an abscess within the yellow bounding box 4024. FIG. 40C The pixel vectors within each cluster are sampled for further spectral feature comparison 4040, as shown in images 4040, 4042. FIG. 40C According to the k-means clustering algorithm, spectral features with the same cluster label are very similar, while they are different from each other if they do not have the same cluster label. The dimension is further reduced to 3 to visualize the pixel vectors in 3D space.

[0458] As can be seen, the abnormal tissues marked in the diamond 4045 are closer to each other. The light yellow cross cluster 4047 is a tray feature, and the features with pink square 4049 and orange cross 4051 belong to normal tissues. The features of the purple dots 4053 can also be abnormal in this image, as they are scattered far away from normal tissues. FIG. 40D

[0459] Second Case Study As shown in images 4120, 4122 of liver 8, the liver has multiple nodules and flukes within the yellow bounding box 4124.

[0460] SWIR image 4126 and normalized image 4128 of liver 8 are also shown. The k-means clustering results are shown in image 4130, and the merged mask is shown in image 4132. FIG. 41A FIG. 41A The pixel vectors within each cluster are sampled for further spectral feature comparison 4140, as shown in images 4140, 4142. According to the k-means clustering algorithm, spectral features with the same cluster label are very similar, while they are different from each other if they do not have the same cluster label. However, the features of multiple nodules are more difficult to segment when compared to the abscess features of liver 9 (

[0461] ), as the features are closer. The dimension is further reduced to 3 to visualize the pixel vectors in 3D space. FIG. 41B FIG. 41B As can be seen, the abnormal tissues marked in the diamond 4045 are closer to each other. The light yellow cross cluster 4047 is a tray feature, and the features with pink square 4049 and orange cross 4051 belong to normal tissues. The features of the purple dots 4053 can also be abnormal in this image, as they are scattered far away from normal tissues. FIG. 40A As can be seen, the abnormal tissues marked in the diamond 4045 are closer to each other. The light yellow cross cluster 4047 is a tray feature, and the features with pink square 4049 and orange cross 4051 belong to normal tissues. The features of the purple dots 4053 can also be abnormal in this image, as they are scattered far away from normal tissues.

[0462] FIG. 41C ​​​It can be seen that the abnormal tissue marked with purple diamonds 4145 are closer to each other. The light yellow cross cluster 4147 is a tray feature. The features with pink squares 4149, orange crosses 4151, and purple dots 4153 belong to normal tissue.

[0463] Sheep organ scans

[0464] Scanning of sheep organs and postmortem inspection were performed in a similar way as for beef organs. However, the developed image data pre-processing, feature extraction, and machine learning models were different from those described above for beef cattle. The main difference is that data extraction was performed manually in selected regions of interest and spatial pixel information was not analyzed.

[0465] Materials and methods for sheep organs

[0466] Sheep organs were inspected using the same X-ray procedure as for the beef trial. Different organs were sampled from cooperating slaughterhouses and sales points. In addition, the innards of one lamb (heart, lungs, and liver) were evaluated to distinguish between organ types within the same image. The color differences due to tissue density make this easy for the human eye, and therefore, FIG. 42A The labeled X-ray image 4204 in FIG. 21 shows the clear differences in color that can be confirmed by the RGB image 4202. The RGB image 4202 is a photo of the innards of a healthy lamb, and the X-ray image 4204 shows the labels or annotations, with the heart being green 4206, the lungs being yellow 4208, and the liver being red 4210.

[0467] Results for sheep organs

[0468] FIG. 42A A predictive algorithm for organ type differentiation using multi-energy X-ray attenuation can be developed. Note that in FIG. 42A and FIG. 42B the ROIs have been labeled, avoiding visible fat, distinct color changes, or transition areas between organs.

[0469] FIG. 42B An X-ray image 4204 is shown in accordance with some embodiments of the present specification, indicating the differences in multi-energy X-ray intensity between the three organ types of a lamb’s innards (heart, lungs, and liver), each labeled and with a corresponding intensity histogram 4212a, 4212b, 4212c of the yellow-labeled organ of interest 4208.

[0470] Image processing software was employed to determine the intensity of X-rays passing through each type of organ tissue in FIG. 42A , where a lower intensity means that the organ tissue is denser. FIG. 42BThe images show that the heart (int = 83) is the most dense, followed by the liver (int = 149), and then the lungs (int = 172). These intensities are... FIG. 42B The color changes are clearly shown (heart is green 4206, lungs are yellow 4208, liver is red 4210).

[0471] Sheep lungs are selected from cooperative slaughterhouses because of the presence of caseous lymphadenitis (CLA), commonly known as caseous glands, in the lymph nodes surrounding the lungs. FIG. 43A to 43D The image shows two sheep lungs suspected of having CLA within the lymph nodes of the lung tissue, which can be easily identified using X-ray attenuation. Meanwhile, FIG. 43E and 43F The labeled X-ray images and small areas of interest (ROIs) are shown to compare the intensity of healthy and diseased areas of the same lung.

[0472] Specifically, FIG. 43A The image 4302 shows an RGB image of a sheep lung according to some embodiments of this specification, which illustrates evidence of CLA and six X-ray images 4304 taken at different energy levels.

[0473] FIG. 43B The images show an RGB image 4302 with markings or annotations 4306 indicating CLA in the mediastinal lymph nodes via palpation, another RGB image 4308 after resection (right-hand cross-section), and an X-ray image 4310 with markings or annotations 4312. The darker areas marked in X-ray image 4310 correspond to CLA in the mediastinal lymph nodes of the lung tissue, and are marked in both images. The intermediate image 4308 shows a large caseous gland corresponding to the left lung, which is the right-hand side of the first image 4302 and the last image 4310. No image of the right lung (left-hand side of the image) was taken; the right lung is marked as a potential area for the presence of caseous glands, as shown in X-ray image 4310.

[0474] FIG. 43C An RGB image 4320 of another sheep lung, illustrating evidence of abscess, and six X-ray images 4322 taken at six different energy levels, are shown according to some embodiments of this specification. FIG. 43C Six X-ray images, 4322, show a large lesion in the right lung, in which the lung was deemed unfit for human consumption upon palpation. The darker color in the corresponding area of ​​the right lung (left-hand side) is... FIG. 43C As can be seen in, and in FIG. 43D It is marked as comment 4325. FIG. 43DA large abscess filled with pus (labeled as annotation 4327) is shown to be revealed as the cause of the rejection when the slice is taken. The lesion is clearly seen and the ROI 4329 is easily labeled in the X-ray image 4330 as there is a region of higher density (darker and lower intensity region, int = 27562) in the corresponding location on the image in FIG. 43E FIG. 43E Specifically, FIG. 43C and FIG. 43D First and second intensity histograms 4335 and 4337 of abscessed and healthy regions of sheep lungs from

[0475] Similarly, the intensity differences encountered when imaging sheep lungs with abscesses also occur in sheep lungs showing evidence of CLA. FIG. 43F Images 4340, 4342 in

[0476] FIG. 44 RGB image 4402 of a diseased sheep liver in

[0477] FIG. 45 RGB image 4502 and X-ray image 4504 of a healthy sheep liver according to some embodiments of the present specification are shown. Also shown is an intensity histogram 4508 within a yellow polygon 4506 of the X-ray image 4504. FIG. 44 and 45 Intensity histograms 4408, 4508 in

[0478] In FIG. 46A ​An RGB image 4602 of a sheep liver containing visible evidence of lesions is shown, whereby a liver lesion (1 cm diameter) is defined on one sheep liver, which was diagnosed to have hepatitis. This is visible in the X-ray image 4604 in FIG. 46A . Image 4603 shows the liver meat piece for the cross-section. In the same liver, the "normal" liver tissue is compared to the local liver lesion, which was found to have a slightly increased intensity (35583 vs. 39020). FIG. 46B An X-ray image 4604 with first and second markers 4606 and 4608 corresponding to normal tissue and local liver lesion in a sheep liver, and their respective intensity histograms are shown.

[0479] Lamb lung. A lamb lung was also scanned using the multisensory system of the present specification. An example of MEXA image data 4704 is presented in FIG. 47 . Image data 4704 provides contrast between airways and lung tissue and can be used to detect pneumonia.

[0480] Cottage cheese gland. Another important issue in slaughterhouses is the detection of cottage cheese glands. FIG. 48 An example photo 4802 (closed and open) of a cottage cheese gland in mutton is shown in

[0481] Hyperspectral data of sheep organs and detection algorithms

[0482] Visible and SWIR surface reflectance hyperspectral intensity spectra 4902 of 102 mixed sheep and beef organs (Table 4) are shown in FIG. 49 . Differences between these spectra make it possible to differentiate between organ types. Plots 5002, 5004 of automatic organ classification accuracy using visible, SWIR, and their combination using two classification models (partial least squares discriminant analysis and random forest) are shown in FIG. 50 .

[0483] Table 4. Distribution of sheep and beef organs collected by organ type.

[0484]

[0485] Disease status - sheep organs

[0486] First derivatives of absorbance of visible and SWIR hyperspectra 5102, 5104 of 89 healthy and diseased sheep organs (Table 5) are shown in FIG. 51 . Differences between these spectra make it possible to differentiate between organ health. Plots 5202, 5204 of automatic organ classification accuracy using visible, SWIR, and their combination using two classification models (partial least squares discriminant analysis and random forest) are shown in FIG. 52 .

[0487] Table 5. Distribution of sheep 89 organs by organ type and disease status.

[0488]

[0489] Grain vs. forage fed beef

[0490] Average visible and SWIR hyperspectral reflectance spectra 5302, 5304 of 108 (54 grain and 54 forage, frozen) steaks are shown in FIG. 53 Differences between these spectra allow for the differentiation of steak origin. Automatic steak classification tables using visible and SWIR HS with three classification models (partial least squares discriminant analysis, linear discriminant analysis, and random forest) are shown in Tables 6 and 7, respectively, where all three models achieved significant differentiation.

[0491] Table 6. Model metrics on a validation dataset (n = 28, 25% of total samples) for differentiating forage and grain fed beef using visible hyperspectral imaging. Models were developed on a training dataset (n = 80, 75% of samples) using partial least squares discriminant analysis (PLS-DA), linear discriminant analysis (LDA), and random forest (RF) methods.

[0492]

[0493] Table 7. Model metrics on a validation dataset (n = 28, 25% of total samples) for differentiating forage and grain fed beef using shortwave infrared (SWIR) hyperspectral imaging. Models were developed on a training dataset (n = 80, 75% of samples) using partial least squares discriminant analysis (PLS-DA), linear discriminant analysis (LDA), and random forest (RF) methods.

[0494]

[0495] Additional uses of multi-sensory (MEXA) imaging system

[0496] Scanning of beef primal cuts

[0497] Beef primal cuts (wholesale rib sections) were scanned as part of a larger trial involving the development of optimal carcass endpoints for feedlot cattle by breed. Cattle were slaughtered at 0, 50, 100, 150, and 200 days on commercial feedlots and selected wholesale rib sections were chosen for X-ray scanning to develop prediction algorithms for fat, muscle, and bone proportions. FIG. 54X-ray images of the same raw meat at three different energy levels (low, medium, and high) 5402, 5404, 5406 using two views (upper projector view 5410 and side projector view 5412) are shown, as well as a summed signal 5408 of all six energy levels. The X-ray data shown in the images is raw data that is not pre-processed or normalized.

[0498] Steak scans

[0499] Data collection runs were performed on steak samples. In Figure 55 the images, low, medium, high energy and summed X-ray images 5502, 5504, 5506, 5508 of the six meat portions are shown. In addition to these, there are three wavelength SWIR images 5502', 5504', 5506' and an assembled image of maximum intensity pixel values 5508' for qualitative visualization.

[0500] Lamb scans

[0501] In Figure 56 sample first and second X-ray image data 5602, 5604 of two lambs are provided. As shown, the inherent image quality appears reasonable for analysis of health issues such as pneumonia, brain inflammation, foreign object inclusions (e.g., syringe needle), and other gross anatomical issues. The X-ray data 5602, 5604 is able to detect differences between lambs that died from head trauma during birth, and lambs that did not consume milk between birth and death. These results show other potential applications for avoiding post-slaughter inspection, and potentially contribute to understanding causes of lamb mortality and reducing lamb mortality.

[0502] This specification relates to the use of X-ray technology in detecting pathology in foodborne issues. Out of 52 livers scanned that were rejected for human consumption, 32 were found to have varying degrees of discoloration (focal and multifocal, located in different lobes, extended and localized), 7 abscesses, 5 duct thickening, 4 fibrosis, 2 flukes, and 2 cysts. Thus, discoloration without changes in tissue density is not expected to be captured by X-ray absorption assays.

[0503] Lesions such as abscesses and flukes cause changes in liver tissue, including calcification and thickening processes, which alter the physiological radiodensity of the organ and can be detected by X-ray images. In the lung, a less dense tissue than the liver, abscesses and CLA lesions are more easily noticed. Visual and X-ray intensity comparisons show differences between the liver, kidney, and lung, in addition to their size and shape.

[0504] Soft tissue abscesses are loci or localized collections of pus caused by bacteria or other pathogens, which are surrounded by a peripheral rim or abscess membrane found within soft tissue anywhere in the body. Even though X-rays often have limited value for the assessment of soft tissue abscesses, they can show soft tissue gas or foreign bodies, thus increasing suspicion of an infectious disease process or revealing other causes of underlying soft tissue swelling.

[0505] Liver fluke disease or liver distome disease is a foodborne liver fluke zoonosis caused by liver fluke and distome fluke. Liver fluke is a flat, leaf-shaped hermaphroditic parasite. Radiologic findings can often demonstrate characteristic changes, thus aiding in the diagnosis of liver fluke disease. Early parenchymal stages of the disease can show subcapsular hypoattenuating areas in the liver.

[0506] While X-ray technology does not seem to identify the shape of the lesion, the images depict various degrees of modification of the liver pattern depending on the lesion found. For each liver scanned, the areas of interest are marked based on the macroscopic aspect of the organ, then confirmed during the postmortem inspection. With the marked lesions, the six radiographs show a lighter gray shadow when compared to healthy tissue. Liver lesions caused by observed pathologies (i.e., duct thickening, calcification, etc.) cannot be accurately identified by radiographs, and this technique can only show an unusual gray shadow by marked areas.

[0507] This specification demonstrates the use of multi-energy X-ray technology to differentiate organs in a simulated slaughterhouse environment and presents X-ray images compared to whole and anatomical photography images of the same organs, with markings and annotations to train a neural network to differentiate lesions on X-ray images compared to corresponding areas of interest on healthy organs.

[0508] Livers from cattle are significantly larger than sheep, and some of them present obvious lesions indicative of disease processes. The images in Trial 2 are a mix of species (mainly sheep) and organ types, where lamb offal X-rays show density differences of different types of organs, while cattle livers show significantly denser than sheep livers, and and cattle (Wagyu) livers show denser than non-and cattle livers. Thus, the size of the tissue or organ being scanned can affect the ability of the X-ray sensor to detect differences in abnormal tissue. In embodiments, the areas and distances between X-ray images can be adjusted for different animal species.

[0509] When lesions are visible to the naked eye or felt as abnormalities in the tissue through palpation, X-ray images can identify their shape and further information without the need for sectioning prior to sectioning. However, when organs are simply discolored or have some subconscious evidence of disease processes due to an overly thick surface, such as liver flukes deep in a large beef liver or cystic fibrosis, X-ray images cannot be flagged and thus the intensity histogram from a given ROI will be to determine whether the organ can be fit for human consumption in the best method possible.

[0510] Hyperspectral data

[0511] Hyperspectral (HS) imaging in the form of two sensors within the multi-sensory platform 2200 is a non-contact technique that includes the visible spectrum (400-900 nm) and the shortwave infrared spectrum (900-1700 nm). The HS images generated from the frame-by-frame slicing of the hypercube within the region of interest (ROI) are surface-based and can detect differences in spectral signatures within different products (e.g., organs, meat, and agricultural products).

[0512] These spectral signatures are extracted from a given ROI across each sample and can be compared and contrasted with one another using machine learning modeling techniques such as partial least squares discriminant analysis, random forests, and artificial neural networks. As a non-contact, non-destructive classification tool, HS can be used to classify organs by organ type and determine whether a particular type of organ is diseased. The various algorithms of the present specification can be integrated with the multi-sensory platform.

[0513] According to aspects of the present specification, the multi-sensory imaging system / platform 2200 can be used in organ handling scenarios under commercial conditions, such as slaughterhouses or processing plants. One non-limiting example is a situation where organs are mixed and need to be identified by both species and type. The results of the spectral of each organ and the classification algorithm that distinguishes the organs by species and type are described below. Figure 57The image shows the visible (VIS) reflectance spectra 5700a, 5700b and short-wave infrared (SWIR) reflectance spectra 5700c, 5700d for each of the four organs (liver, heart, lung, and kidney) and for each species (beef 5700b, 5700d and sheep 5700a, 5700c), where the heart and lung show higher intensities compared to the liver and kidney in both the VIS and SWIR regions. For the VIS spectra 5700a and 5700b, the intensity difference occurs between 500 and 850 nm. The heart and lung have similar spectral characteristics throughout the VIS spectrum, except between 500 and 600 nm, where the heart has greater intensity. Similarly, between 500 and 600 nm, the kidney shows slightly greater intensity than the liver. A stronger distinction occurs in the SWIR regions of 5700c and 5700d, particularly between 1050 and 1350 nm, with the lung exhibiting the highest intensity, followed by the heart, then the liver, and finally the kidney with the lowest intensity. Species-specific differences are negligible for both spectra.

[0514] In some embodiments, cross-species datasets are aggregated to develop organ and species discrimination algorithms, as if scanning a mixture of organs from two species via a platform. These algorithms are developed and validated using 5-fold cross-validation. When using two substances, three different spectral regions (VIS, SWIR, and a combination of VIS and SWIR – COMB) each showed optimal predictive performance using different discriminant analysis models. Figure 58A As shown, model indices 5800a, which indicate predictions from VIS spectral data, are the most accurate for linear discriminant analysis (LDA) modeling. In contrast, model indices 5800b, which indicate predictions from SWIR data, are modeled using random forest (RF). Figure 58B The most accurate and indicative prediction of the COMB model index 5800c is modeled using partial least squares discriminant analysis (PLS-DA). Figure 58C ) is the most accurate. However, the three classification methods evaluated using VIS or SWIR produced similar accuracies, but COMB was significantly more accurate using PLS-DA compared to RF and LDA. COMB's PLS-DA model also had the highest sensitivity, specificity, under-curve region, and Kappa consistency coefficient (). Figure 58C Compared to using each sensor individually, using COMB yielded the highest overall accuracy (71–92%) and KPa (63–89%), regardless of the discriminant analysis method used. Therefore, these results demonstrate that HS imaging is accurate for distinguishing species and organs in industrial applications when these are combined and passed through the platform.

[0515] Figure 59Spectra after different smoothing processes are shown: a) raw VIS 5900a; b) centered moving average VIS 5900b; c) Savitzky-Golay filtered VIS 5900c; d) raw SWIR 5900d; e) centered moving average SWIR 5900e; f) Savitzky-Golay filtered SWIR 5900f. The Savitzky-Golay filter did not smooth the spectra to the same extent as the centered moving average. In some embodiments, the most accurate model was absorbance with first derivative. However, in most cases, the difference in prediction accuracy using different data pre-processing methods was not large, and thus, data pre-processing can not be critical in all cases.

[0516] In some embodiments, the algorithm can scan the entire organ and then search for abnormal areas, which can be aided by X-ray spectroscopy. Similarly, identifying different components of an organ sampled from a slaughterhouse, such as lymph nodes, fat, and bile ducts, can help identify the organ, as well as detect defects, diseases, or abnormalities. However, this would require a larger ROI, for example, marking the entire organ from the HS image.

[0517] Hyperspectral sensors can only measure electromagnetic radiation from the surface of the product, and thus cannot measure properties inside the organ to detect defects or abnormalities below the surface. However, the multi-perception platform of the present specification is used to collect data from multi-energy X-ray sensors that can penetrate tissue further. In some embodiments, the platform 2200 contains six X-ray sensors that penetrate and identify abnormalities that the HS sensors cannot identify at different depths.

[0518] As Figure 60 shown, based on visual inspection, several multi-energy X-ray attenuation (MEXA) images of a sheep lung generated by the multi-perception platform 2200 show the shape and features of the organ, and in some cases, allow for marking of defects such as caseous lymphadenitis (CLA). In Figure 60 the MEXA image 6002a shows a CLA lesion that was felt by an inspector at the slaughterhouse during palpation, and then confirmed by a veterinary pathologist after examination / incision, as shown in image 6002b. The X-ray image 6002c of the sheep lung shows a caseous lymphadenitis (CLA) lesion.

[0519] On the other hand, Figure 61It is shown that in other organs (e.g. kidney with pyelonephritis), the defects are not obvious to the naked eye, although they can become apparent after image analysis or if the scan shows the anatomical organ of the lesion. Image 6100d is a bisected kidney showing a focal lesion suggestive of pyelonephritis. However, the MEXA image 6100e of the defective kidney appears lighter in color than the image 6100b of a healthy kidney that is passed for human consumption. Image 6100a is a passed sheep kidney suitable for human consumption, while image 6100c is a sheep kidney rejected at the slaughterhouse due to defects. This indicates that diseases that can cause changes in tissue density can be reflected in X-ray absorption, allowing the creation of thresholds for heart, liver, lung, kidney and different defects that can replace palpation in slaughterhouses. The system 2200 can also be used to detect foreign bodies, such as needles from vaccinations or other metallic bodies. The multi-sensory imaging system 2200 is able to detect organs with defects in animal tissue (particularly beef offal) with an average accuracy greater than 70%, preferably 90%.

[0520] The presently disclosed embodiments can be used to assist inspectors in slaughterhouses and infer the presence of potential lesions and their location within the organ. Furthermore, if the image library is extended to have more organs, the X-ray can determine whether the organ of interest (i.e. abnormal thickness or discoloration when palpated) is too dense compared to healthy organs within the image library, with appropriate thresholds confirmed for acceptance or rejection. Moreover, once there is a large amount of labeled data with information from several lesions, disease processes can also be identified through X-ray imaging.

[0521] Hyperspectral imaging technology is non-invasive and non-contact and can enable automatic classification and disease detection of livestock organs in slaughterhouses, allowing animal health reports to be provided to producers.

[0522] The above examples are merely illustrative of numerous applications of the system of the present specification. Although only a few embodiments of the present application are described herein, it should be understood that the present application can be put into practice in many other specific forms and implementations without departing from the spirit or essential characteristics thereof. Therefore, the present examples and embodiments are to be considered as illustrative and not restrictive, and the application can be modified in the scope of the appended claims.

Claims

1. An imaging system configured to evaluate meat, comprising: an X-ray scanning system configured to generate X-ray scanning data of meat; a hyperspectral imaging system configured to generate hyperspectral imaging data; a computing device in data communication with the X-ray scanning system and the hyperspectral imaging system, wherein the computing device comprises a processor and a memory storing a plurality of programming instructions that, when executed by the processor, configure the processor to: acquire the X-ray scanning data and the hyperspectral imaging data; automatically determine a quality of the meat by analyzing the acquired X-ray scanning data in combination with the hyperspectral imaging data; classify the meat into one of an acceptable quality category and an unacceptable quality category based on the determined quality; and generate data indicative of the quality of the meat.

2. The system of claim 1, wherein the X-ray scanning system comprises a two-dimensional projection X-ray imaging system in combination with a multi-energy X-ray (MEXA) sensor, the two-dimensional projection X-ray imaging system having at least one of a single view configuration or a dual view configuration.

3. The system of claim 2, wherein the X-ray scanning system comprises a tilted conveyor such that an inlet end of the conveyor is at a lower elevation than an outlet end of the conveyor.

4. The system of claim 2, wherein the X-ray scanning system uses a downwardly tilted conveyor such that an inlet end of the conveyor is at a higher elevation than an outlet end of the conveyor.

5. The system of claim 1, wherein the hyperspectral scanning data comprises data in a visible light wavelength range and a short wave infrared wavelength range.

6. The system of claim 1, wherein the meat comprises offal and organs.

7. The system of claim 1, further comprising at least one of an inkjet, a laser beam, an LED strip, or an augmented reality headset adapted to produce a visual indication of the quality associated with the meat.

8. The system of claim 1, wherein the processor is further configured to: generate at least one graphical user interface to display at least one image corresponding to the X-ray scanning data, and determine the quality based on data indicative of a thickness and / or density of the meat.

9. The system of claim 1, further comprising a conveyor that translates the meat through the system at a speed in a range of 0.1 m / s to 1.0 m / s.

10. The system of claim 1, wherein the multi-sensor imaging system has an inspection tunnel having a length ranging from 1100 mm to 5000 mm, a width ranging from 500 mm to 1000 mm, and a height ranging from 300 mm to 1000 mm. ​ 11. The system of claim 1, wherein the X-ray scanning system comprises a first X-ray source of 120 to 160 keV with a beam current of 0.2 to 1.25 mA and a second X-ray source of 120 to 160 keV with a beam current of 0.2 to 1.25 mA, wherein the first X-ray source is configured in an overhead shooter configuration and the second X-ray source is configured in a side shooter configuration.

12. The system of claim 11, wherein the X-ray scanning system comprises a multi-energy photon counting X-ray sensor array.

13. The system of claim 11, wherein the X-ray scanning system comprises 6 to 22 data acquisition panels corresponding to the first X-ray source and 4 to 20 data acquisition panels corresponding to the second X-ray source.

14. The system of claim 1, wherein the X-ray scanning system is configured to acquire data in a plurality of energy bands, wherein the plurality of energy bands ranges from 3 to 20, and wherein each of the energy bands is in a range of 20-160 keV.

15. The system of claim 1, wherein the hyperspectral imaging system comprises a first camera sensor configured for visible light imaging in a 200 to 1200 wavelength band and a second camera sensor configured for short wave infrared imaging in a 400 to 700 wavelength band.

16. The system of claim 15, wherein the first camera sensor is configured to operate in a range of 400 nm to 900 nm and has a spectral resolution of at least 20 nm across a width of the conveyor, wherein a pixel size is no more than 2.0 mm.

17. The system of claim 16, wherein the second camera sensor is configured to operate in a range of 900 nm to 1800 nm and has a spectral resolution of at least 20 nm across a width of the conveyor, wherein a pixel size is no more than 2.0 mm.

18. The system of claim 1, wherein the hyperspectral imaging system is configured to have an acquisition rate of 30 to 150 Hz.

19. The system of claim 1, wherein the X-ray scanning system and the hyperspectral imaging system are synchronized with an X-ray base frequency ranging from 150 to 500 Hz.

20. The system of claim 1, wherein the processor is further configured to determine a type of meat based on the acquired X-ray scanning data and hyperspectral imaging data.

21. The system of claim 1, wherein the processor is further configured to: generate at least one graphical user interface to display at least one image corresponding to the hyperspectral imaging data; identify a region in the at least one image that is indicative of an anomaly; and apply an annotation to the identified region, wherein the annotation is at least one of a shape or a color.

22. The system of claim 21, wherein the processor is configured to implement at least one machine learning model, and wherein the machine learning model is configured to analyze the hyperspectral imaging data in order to determine the quality of the meat and the region that is indicative of an anomaly.

23. The system of claim 22, wherein the machine learning model is adapted to be trained using K-means clustering in order to identify areas indicative of abnormalities.

24. The system of claim 1, wherein the data indicative of the quality of the meat includes at least one of lean meat yield, intramuscular fat to tissue ratio, amount of intermuscular fat, absolute size of individual organs, relative size of individual organs, muscle volume, number of ribs, presence or absence of disease, presence or absence of cysts, presence or absence of tumors, presence or absence of pleurisy, or presence or absence of foreign objects.

25. A system for generating data indicative of animal feeding practices and meat production practices, comprising: a plurality of geographically distributed meat production sites having associated multi-sensor imaging systems, wherein each of the multi-sensor imaging systems includes an X-ray scanning system and a hyperspectral imaging system; at least one server in data communication with a database and each of the multi-sensor imaging systems, wherein the at least one server contains a processor and a memory storing a plurality of programming instructions that, when executed by the processor, configure the processor to: implement at least one machine learning model; provide a plurality of data accessed from the database as input to the at least one machine learning model, wherein the at least one machine learning model is configured to analyze the plurality of data to generate the data, wherein the data is intended to maximize a plurality of positive parameters related to animal feeding and meat production and minimize a plurality of negative parameters related to animal feeding and meat production; and enable a plurality of geographically distributed computing devices to access the generated data.

26. The system of claim 25, wherein the plurality of data corresponds to a collection of a plurality of animal and meat related data from each of the plurality of geographically distributed meat production sites, and wherein the plurality of animal and meat related data includes at least one of: animal ID, animal type, animal breed, X-ray scan data corresponding to each different age of the animal, X-ray scan data of the animal carcass and / or raw meat, hyperspectral image data of the animal meat and organs, geographic location of the feedlot and / or meat production site, climate, weather, season, type of feed, time of year of meat production, vaccination history, medications, disease history, age of the animal when received at the meat production site, lean meat yield, intramuscular fat to tissue ratio, amount of intermuscular fat, absolute size of individual organs, relative size of individual organs, muscle volume, number of ribs, presence or absence of disease, presence or absence of cysts, presence or absence of tumors, presence or absence of pleurisy, or presence or absence of foreign objects.

27. The system of claim 26, wherein the plurality of positive parameters comprises at least one of: reduced drug requirements, lower carbon footprint, variable cost efficiency, reputation protection, lower health risks to consumers, or improvements in lean meat production, ratio of intramuscular fat to tissue, amount of intermuscular fat, absolute size of individual organs, relative size of individual organs, muscle volume, number of ribs, absence of disease, absence of cysts, absence of tumors, absence of pleurisy, or absence of foreign objects.

28. The system of claim 26, wherein the plurality of negative parameters comprises at least one of an increase in presence of disease, presence of cysts, presence of tumors, presence of pleurisy, or presence of foreign objects.

Images