Antimicrobial Susceptibility Testing Using Machine Learning

Abstract

The optimized test method is used to determine the minimum inhibitory concentration (MIC) of a specific antimicrobial agent for a sample. This may include iterative imaging of wells inoculated with the sample and containing various concentrations of the antimicrobial agent. The images are then processed to determine various feature values ​​that are provided as input to a machine learning model.

Description

Background Art

[0001] The various types of tests related to patient diagnosis and treatment can be performed by analyzing the patient's microorganism or "microbe". Microorganisms are microscopic living organisms, such as bacteria, fungi or viruses, which can be unicellular or multicellular. The biological sample containing the patient's microorganism can be taken from the patient's infection, body fluid or abscess, and can be placed in a test plate or array, combined with a variety of reagents, incubated and analyzed to help the patient's treatment. Automatic biochemical analyzers have been developed to meet the needs of health care institutions and other institutions, so that patient samples are analyzed, and the accuracy and reliability of the assay results are improved compared with the analysis using manual operations, and the effectiveness of multiple antimicrobial agents is determined. Antimicrobial agents are reagents that kill microorganisms or inhibit their growth, such as antibiotics for bacteria and antifungal agents for fungi. However, with the ever-changing bacterial genus and newly discovered antimicrobial agents, the demand for biochemical tests has increased in both complexity and volume.

[0002] An important class of automatic microbiological analyzers functions as a diagnostic tool for determining the characteristics of both infectious microorganisms and antimicrobial agents that effectively control the growth of the microorganism. Automatic microbiological analyzers function as a diagnostic tool for determining the characteristics of both infectious microorganisms and antimicrobial agents that effectively control the growth of the microorganism. When performing diagnostic tests, the identification and in vitro antimicrobial susceptibility patterns of microorganisms separated from biological samples are determined. Conventional forms of such analyzers can place small samples to be tested into multiple small sample test wells in a plate or array containing different enzyme substrates or antimicrobial agents diluted in series. Microbial identification (ID) tests and antimicrobial susceptibility tests (AST) for determining the minimum inhibitory concentration (MIC) of antimicrobial agents effective for the microorganism can utilize color changes, fluorescence changes, turbidity (turbidity) in the sample test wells established in the array, or other information derived from the test. Both AST and ID measurements and subsequent analysis can be performed by a computer-controlled microbiological analyzer to provide advantages of reproducibility, reduced processing time, avoidance of transcription errors, and standardization of all tests run in a laboratory.

[0003] In an ID test for microorganisms, a standardized dilution of a patient's microbial sample, called an inoculum, is first prepared to provide a bacterial or cell suspension with a predetermined known concentration. The inoculum is placed in a plurality of test wells, which may contain or be subsequently provided with a predetermined test medium. Depending on the type of microorganism present, the medium will promote changes in color, turbidity, fluorescence, or other characteristics after incubation. These changes are used to identify the microorganism in the ID test.

[0004] In the AST test, a plurality of test wells are filled with many different antimicrobials, such as antibiotics, that inoculum and concentration are progressively increased. Different antimicrobials can be diluted to include clinically interested concentrations in growth medium or liquid culture medium. After incubation, when growth is not inhibited by the antimicrobial in those test wells, turbidity will improve or remain unchanged in the test wells. The MIC of every kind of antimicrobial is measured by the lack of growth relative to every kind of antimicrobial concentration. Therefore, the minimum antimicrobial concentration that demonstrates growth lacks is the MIC. BRIEF DESCRIPTION OF THE DRAWINGS

[0005] While the specification concludes with claims which particularly point out and distinctly claim the invention, it is believed that the invention will be better understood from the following description of certain embodiments taken in conjunction with the accompanying drawings, wherein like reference numerals represent like elements, and wherein:

[0006] FIG. 1A depicts a portion of a schematic diagram of an exemplary biological testing system;

[0007] Figure 1B Depicted Figure 1A Another portion of a schematic diagram of a biological test system;

[0008] Figure 2 Depicted Figure 1B A perspective view of an exemplary incubator system and an exemplary optical system of a biological testing system;

[0009] Figure 3 Depicted Figure 2 Perspective view of the optical system;

[0010] Figure 4 Depicted Figure 2 Another perspective view of the optical system of FIG. 1 , showing the XY stage of the optical system;

[0011] Figure 5 Depicted Figure 2 A schematic diagram of a portion of an optical system;

[0012] Figure 6 depicts a schematic diagram of an exemplary computer system;

[0013] Figure 7 A graph depicting the theoretical growth curve of a microorganism;

[0014] Figure 8 Describes the use Figure 1A and 1B Schematic diagram of an exemplary growth test well of a biological test system;

[0015] Fig. 9 Describes the use Figure 1A and 1B An exemplary image analysis cycle of a biological testing system;

[0016] Fig.10 Depicted by Figure 2 An exemplary raw image captured by the optical system;

[0017] Fig.11 Describes from Fig.10 An exemplary enhanced image obtained from an original image of

[0018] Fig.12 Describes from Fig.10 An exemplary gradient image derived from an original image of

[0019] Fig.13 depicts an exemplary image enhancement flow chart;

[0020] Fig.14 depicts an exemplary dynamic image enhancement flow chart;

[0021] Fig.15 Describes from Fig.10 An exemplary segmented image obtained from the original image;

[0022] Fig.16 depicts an exemplary image segmentation flow chart;

[0023] Fig.17 An exemplary microorganism having elongated cells is depicted;

[0024] Fig.18 An exemplary microorganism having non-elongated units is depicted;

[0025] Fig.19 Depicting an exemplary optimized antimicrobial susceptibility testing method of the present invention;

[0026] Fig. 20 depicts an exemplary decision tree that can be used to determine the MIC;

[0027] Fig.21 depicts an exemplary decision tree that can be used to determine the MIC;

[0028] Fig. 22 depicts an exemplary decision tree that can be used to determine the MIC;

[0029] Fig.23A depicts a portion of an exemplary decision tree that can be used to determine the MIC;

[0030] Fig. 23B depicts a portion of an exemplary decision tree that can be used to determine the MIC;

[0031] Fig.24 depicts an exemplary structure of a neural network that can provide MIC predictions; and

[0032] Fig.25 An exemplary structure of a neural network that can make growth / inhibition predictions is depicted.

[0033] The accompanying drawings are not intended to be limiting in any way, and it is contemplated that various embodiments of the invention may be carried out in a variety of other ways, including ways not necessarily shown in the drawings. The accompanying drawings, which are incorporated in and form a part of the specification, illustrate several aspects of the invention and, together with the description, serve to explain the principles of the invention; however, it should be understood that the invention is not limited to the precise arrangements shown. DETAILED DESCRIPTION

[0034] The following description of certain embodiments of the present invention should not be used to limit the scope of the present invention. Other examples, features, aspects, embodiments and advantages of the present invention will become apparent to those skilled in the art from the following description, which is by way of example, one of the best ways to implement the present invention. As will be appreciated, the present invention can have other different and obvious aspects, all of which do not depart from the present invention. Therefore, the drawings and description should be regarded as illustrative and not restrictive in nature.

[0035] It should be understood that any one or more of the teachings, expressions, forms, examples, etc. described herein may be combined with any one or more of the other teachings, expressions, forms, examples, etc. described herein. Therefore, the teachings, expressions, forms, examples, etc. described below should not be viewed in isolation from each other. In view of the teachings herein, a variety of suitable ways in which the teachings herein can be combined will be apparent to one of ordinary skill in the art. Such modifications and variations are intended to be included within the scope of the claims.

[0036] I. Biological Testing System Hardware

[0037] Figure 1A and 1B Schematic illustrations of various hardware components that may be used in the biological testing system 1 are depicted. The biological testing system 1 facilitates an optimized antimicrobial susceptibility testing (AST) method 101 ( Fig.19 ). The biological testing system 1 broadly includes a consumable preparation system 3, an inoculation system 5, an incubator system 7 and an optical system 9. Once sufficient materials are loaded by the user, the various systems within the biological testing system 1 coordinate with each other and work automatically.

[0038] To operate the biological testing system 1, the user first obtains an appropriate microbial sample. Figure 1A As shown, the microbial sample can be obtained from an agar plate 11, or in some cases, from a blood sample. Next, the user prepares an inoculum suspension by transferring the microorganisms into a tube containing a suitable liquid culture medium or broth. Figure 1A One such tube is shown as inoculum 13 in the figure. In some forms of the biological test system 1, the liquid culture medium or broth can be an approximately 0.5 mM phosphate buffer solution containing a small amount of sodium chloride and potassium chloride to help maintain the viability of the microorganisms introduced into the solution without adversely interfering with the MIC determination or other related tests. The phosphate buffer solution can be used in both the ID test and the AST test to minimize the need for different inocula in the two systems and take advantage of the efficiency of having a single broth. Each inoculum 13 is placed in an inoculation rack 15, and the entire inoculation rack 15 is placed in the inoculation system 5. Once in the inoculation system 5, if necessary, the inoculum in each inoculum 13 is adjusted to a 0.5 McFarland standard turbidity value to produce an inoculum 17. In some forms of the biological test system 1, a 1 microliter plastic loop or swab can be provided to the user to easily pick colonies from the agar plate and minimize the amount of adjustment required to bring the inoculum to the desired turbidity value. Once adjusted to the desired turbidity value, the inoculum is completed. The final inoculum is hereinafter referred to as inoculum 17. Figure 1B As shown. Inoculum 17 can be further diluted to a 1:250 dilution and converted into inoculum 18. As will be discussed in more detail below, the inoculum contained in each inoculum 17 is applied to an identification (ID) array holder 21, while the inoculum contained in each inoculum 18 is applied to an AST array holder 23. Both the ID array holder 21 and the AST array holder 23 are assembled by the consumable preparation system 3 and provided to the inoculation system 5 for use with the inoculum 17 and the inoculum 18.

[0039] The consumables preparation system 3 is loaded with a magazine of test arrays 19, which may contain a variety of antimicrobial agents or other reagents required for the biological test system 1 arranged in a series of test holes 20. For example, the test array 19 may contain an antimicrobial dilution array or an identification array. The consumables preparation system 3 may also be loaded with a body phase diluent (not shown) and / or a variety of other elements for preparing and completing the ID array holder 21 and the AST array holder 23 and the inoculation therein. Mainly, the consumables preparation system 3 operates as needed to retrieve the test array 19 and combine each retrieved test array 19 into the ID array holder 21 or the AST array holder 23. The test array 19 may be selected and assembled by a robot fixture (not shown) or other mechanical features determined by a prescribed test. For example, a doctor may use an antibiotic amoxicillin to indicate a biological test. Thus, the test array 19 related to the amoxicillin test is retrieved and assembled into the appropriate ID array holder 21 and the AST array holder 23. All or some portions of the test array 19 may be formed from a styrene material to help reduce fluorescence crosstalk, fallout, and / or bubbles when digitally inspecting each test well 20 .

[0040] Once the inoculum 17, inoculum 18, ID array holder 21 and AST array holder 23 are assembled, the inoculation system 5 distributes the substantially undiluted inoculum from the inoculum 17 into the test wells 20 of the ID array holder 21, and distributes the diluted inoculum from the inoculum 18 into the test wells 20 of the AST array holder 23. The time between the application of the inoculum 17 to the ID array holder 21 or the application of the inoculum 18 to the AST array holder 23 and the start of the logarithmic growth of the microorganisms disposed therein is called the "lag time". The lag time can be reduced by using an enhanced broth (e.g., a broth containing yeast extract, vitamins and / or minerals). The lag time can also be reduced by increasing the inoculum. In some forms of the biological testing system 1, the amount of the inoculum can be doubled to reduce the lag time by approximately 30 minutes without affecting the accuracy of the MIC determination. The distribution can be completed by an elevator assembly 26 having an XY robot or an XYZ robot (not shown), which has a clamp (not shown) and a pipette (not shown), as well as various circuits, channels and pipes as necessary. The task of the XYZ robot is to retrieve the inoculum from the inoculation rack 15 and distribute the inoculum to the test holes 20 of the ID array holder 21 and the AST array holder 23. Once the ID array holder 21 and the AST array holder 23 are fully loaded with inoculum, each ID array holder 21 and the AST array holder 23 are moved into the incubator system 7 by the elevator assembly 26.

[0041] like Figure 2As shown, the incubator system 7 includes slots 27 for holding a large number of ID array holders 21 and AST array holders 23. Each array holder is placed in a corresponding slot 27 by an XYZ robot 29 using a clamp 31. The XYZ robot 29 operates to move in any part of the XYZ plane and position the clamp 31 near the desired ID array holder 21 or AST array holder 23. When in the incubator system 7, each array holder is incubated under specific desired environmental conditions. For example, the incubator system 7 can be set to incubate the array holders at 35 degrees Celsius. At certain time intervals during incubation, the XYZ robot 29 retrieves a specific ID array holder 21 or AST array holder 23 and moves the selected array holder to the optical system 9.

[0042] like Figures 2 to 4 As shown, the optical system 9 includes features configured to observe, monitor, view and / or capture images of each test well 20 of the ID array holder 21 or the AST array holder 23. Specifically, each ID array holder 21 is monitored by an ID fluorometer 33, and each AST array holder 23 is monitored by an AST camera 35. To accomplish the monitoring, the XYZ robot 29 retrieves a specific array holder with a gripper 31 and places the selected array holder on an XY stage 37. The XY stage 37 moves in the XY plane to position the array holder under the associated monitoring element, i.e., the ID array holder 21 is disposed under the ID fluorometer 33, and the AST array holder 23 is disposed under the AST camera 35, for monitoring and observation in the optical system 9. The XY stage 37 includes finely adjusted motor controls to allow each test well 20 of the associated array holder to be accurately positioned within the observation frame of the ID fluorometer 33 or the AST camera 35.

[0043] Figure 5An exemplary architecture of the AST optical portion 39 of the optical system 9 is shown. The AST optical portion 39 includes an illumination source 41, an objective lens 43, a tube lens 45, and a folding mirror 47. The illumination source 41 may include a condenser LED system for providing monochromatic illumination of each test hole 20 of the AST array holder 23. The objective lens 43 may include a Nikon 20×0.45NA ELWD (Extra Long Working Distance) objective lens, an Olympus 10X objective lens, or any other suitable type of lens. The objective lens 43 may include a 20× objective lens, wherein each pixel covers approximately 0.33 microns. The 20× objective lens provides the ability to detect a reasonable number of cells (approximately 100 to 200 cells) at the beginning of cell growth and the ability to detect cell morphology. The objective lens 43 may include a 10× objective lens and / or a 5MP camera for a larger dynamic range and / or larger samples for each test hole 20, while maintaining sufficient resolution to count the microorganisms therein. In some exemplary embodiments of the optical system 9, the objective lens 43 only acquires one photo or image of each test hole 20 passed each time. The objective lens 43 may be slightly offset from the bottom of the test hole 20 to eliminate background noise from the bottom of the test hole. In some forms of the optical system 9, the objective lens 43 is configured to focus about 5 to 10 microns away from the bottom of the test hole 20. In some forms of the optical system 9, the objective lens 43 is configured to focus 8 microns away from the bottom. The objective lens 43 may also include a Z platform 44 for allowing the objective lens 43 to move in the Z axis relative to the XY platform 37. Therefore, between the XY platform 37 that moves the AST array holder 23 in the XY platform and the Z platform 44 that moves the objective lens 43 in the Z axis, each test hole 20 of the array holder 23 can be moved in any three-dimensional space to accurately align the test hole 20 with the frame of the AST camera 35. The tube lens 45 may be embodied in an achromatic tube lens. The AST camera 35 may include a Sony IMX253 camera, various types of 5 megapixel cameras provided by other manufacturers (e.g., Canon, Thorlabs, or Sentech), or any other suitable type of camera. In some versions of the optical system 9, the XY stage 37 and the Z stage 44 are replaced by an XYZ stage to provide all three axes of three-dimensional movement of the test hole 20.

[0044] Reference now Figure 6 , various components of the biological test system 1 can include one or more computing devices or systems, such as the exemplary computer system 49. For example, any of the consumable preparation system 3, the inoculation system 5, the incubator system 7, and / or the optical system 9 can include one or more computing systems, such as the exemplary computer system 49. Alternatively, each of these subsystems of the biological test system 1 can function via commands from an overall computing system, such as the exemplary computer system 49.

[0045] The computer system 49 may include a processor 51, a memory 53, a mass storage memory device 55, an input / output (I / O) interface 57, and a human-machine interface (HMI) 59. The computer system 49 may also be operably coupled to one or more external resources 61 via a network 63 or the I / O interface 57. The external resources may include, but are not limited to, servers, databases, mass storage devices, peripheral devices, cloud-based network services, or any other suitable computer resources that the computer system 49 may use.

[0046] The processor 51 may include one or more devices selected from the following: a microprocessor, a microcontroller, a digital signal processor, a microcomputer, a central processing unit, a field programmable gate array, a programmable logic device, a state machine, a logic circuit, an analog circuit, a digital circuit, or any other device that manipulates signals (analog or digital) based on operating instructions stored in the memory 53. The memory 53 may include a single memory device or multiple memory devices, which include but are not limited to read-only memory (ROM), random access memory (RAM), volatile memory, non-volatile memory, static random access memory (SRAM), dynamic random access memory (DRAM), flash memory, cache memory, or any other device capable of storing information. The mass storage memory device 55 may include a data storage device, such as a hard disk drive, an optical drive, a tape drive, a non-volatile solid-state device, or any other device capable of storing information.

[0047] The processor 51 may operate under the control of an operating system 65 residing in the memory 53. The operating system 65 may manage computer resources so that computer program code embodied as one or more computer software applications (e.g., application 67 residing in the memory 53) may have instructions executed by the processor 51. In an alternative embodiment, the processor 51 may execute the application 67 directly, in which case the operating system 65 may be omitted. One or more data structures 69 may also reside in the memory 53 and may be used by the processor 51, the operating system 65, or the application 67 to store or manipulate data.

[0048] The I / O interface 57 may provide a machine interface that operably couples the processor 51 to other devices and systems, such as a network 63 or external resources 61. Thus, the application 67 may work in conjunction with the network 63 or external resources 61 through communications via the I / O interface 57 to provide various features, functions, applications, processes, or modules that comprise embodiments of the present invention. The application 67 may also have program code that is executed by one or more external resources 61, or otherwise rely on functions or signals provided by other systems or network components external to the computer system 49. Indeed, given the nearly infinite hardware and software configurations that are possible, one of ordinary skill in the art will appreciate that different forms of the present invention may include applications that are external to the computer system 49, distributed among multiple computers or other external resources 61, or provided by computing resources (hardware and software) provided as a service (e.g., cloud computing services) over the network 63.

[0049] The HMI 59 may be operably coupled to the processor 51 of the computer system 49 in a known manner to allow a user to directly interact with the computer system 49. The HMI 59 may include a video or alphanumeric display, a touch screen, a speaker, and any other suitable audio and / or visual indicator capable of providing data to the user. The HMI 59 may also include input devices and controls, such as an alphanumeric keyboard, a pointing device, a keyboard, buttons, a control knob, a microphone, etc., capable of accepting commands or input from a user and sending the incoming input to the processor 51.

[0050] The database 71 may reside on the mass storage memory device 55 and may be used to collect and organize data used by the various systems and modules described herein. The database 71 may contain data and supporting data structures for storing and organizing data. In particular, the database 71 may be arranged with any database organization or structure, including but not limited to a relational database, a hierarchical database, a network database, or a combination thereof. A database management system in the form of a computer software application executed as an instruction to the processor 51 may be used to access information or data stored in records of the database 71 in response to a query, wherein the query may be dynamically determined and executed by the operating system 65, other application programs 67, or one or more modules.

[0051] II. Optimized AST System and Method

[0052] In some forms, the system 1 described above can be used to facilitate an optimized AST method 101 (e.g. Fig.19 Some or all of the features provided in ).

[0053] In some conventional methods, the MIC is determined by manual visual inspection of the test wells after a waiting period to allow for microbial growth. Figure 7 As shown, the growth of the microbial sample in the test wells is not observed with the naked eye until 4 to 10 hours after the inoculum is placed in the test wells. Traditional methods of determining MIC are limited by what a person can visually perceive relative to the growth of the microorganisms in the test wells. In addition, the presence of dilutions of antimicrobial agents at concentrations below the MIC concentration may slow the growth rate and take even longer to detect. Figure 7 As shown in , the human eye cannot visually observe the first six to seven doublings of a microbial sample. However, the information provided by these initial doublings is generally indicative of the MIC. Figures 7 to 22 As shown, the optimized AST method 101 utilizes digital microscopy to monitor the area and count of microorganisms in one test well after another starting from the inoculation point and thus allows for more rapid and accurate detection of MIC.

[0054] Figure 8 An illustration of how data that can be used in an optimized AST method 101 can be captured from a test well is provided. The test well 20 is embodied by a "384" type test well with a clear viewing bottom. The volume of the inoculum in the test well 20 can be set to 20 microliters to reduce the amount of material (e.g., bulk diluent) required for the AST method 101 and / or to minimize light artifacts and provide sampling of a consistent number of microorganisms from a 1:250 dilution of a 0.5 McFarland inoculum. Reducing the volume of the inoculum in the test well 20 generally increases light artifacts. In some cases, capturing a single vertical plane 103 with a 20× objective, thereby setting the AST camera 35 to 0.33 microns / pixel may be sufficient to sample the inoculum and ensure that each individual microorganism is identifiable. However, these parameters are configurable and may be changed as desired by the user or the underlying needs of the system. Capturing three focus positions approximately 5 microns apart within a single vertical plane 103 may also provide a sufficient number of microorganisms in the sample to be counted for use in the optimized AST method 101. The three focus positions are Figure 8 105, 107, and 109. In an exemplary form of the optimized AST method 101, each focus position is approximately 700 x 700 microns within a single vertical plane 103, rather than capturing three positions in three different vertical planes. The AST camera 35, optical system 9, and computer 49 are configured to capture images of each focus position over a continuous period of time, manipulate each of these images, and thereafter use data derived from these manipulated images to perform MIC determinations.

[0055] The image analysis loop 102 generally operates in Fig. 9, which includes an image capture step 111, an image enhancement step 113, an image segmentation step 115, and an object counting step 117. The optimized AST method 101 may include repeatedly performing an image analysis loop 102 until the MIC is determined. The image analysis loop 102 utilizes one or more instances of the computer 49 and its various elements to perform the image capture step 111, the image enhancement step 113, the image segmentation step 115, and the data extraction step 117, as well as any sub-steps provided therein.

[0056] The image analysis cycle 102 begins with an image capture step 111 and the capture of a raw image 119 of the inoculum via the AST camera 35 of the optical system 9 ( Fig.10 ) and stores the raw image 119 in the memory 53. Although the raw image 119 is depicted as a single image, the raw image 119 may also be a composite of several individual images acquired in the vertical plane 103, for example, a composite of the position 105, the position 107, and the position 109. The raw image 119 may also be a composite of several images acquired in different planes within the inoculum sample, or may be a single image. Fig.10 As shown in , the raw image 119 may contain defects such as uneven illumination. The uneven illumination may be the result of a meniscus generated by the inoculum in combination with the walls of the associated test well 20, which may affect the path of the illumination source 41. Uneven background intensity may also be caused by environmental issues (e.g., plastic deformation) or from a variety of other sources. After the raw image 119 is captured in the image capture step 111, the image capture step 111 moves to the image enhancement step 113.

[0057] like Fig. 9 and 11 As shown in FIG. 14 , the image enhancement step 113 processes the original image 119 to create an enhanced image 121 ( Fig.11 ), thereby enhancing the image 121 to be more suitable for the process of counting and identifying microorganisms such as bacteria. During the image enhancement step 113, one or more properties of the original image 119 are modified. These properties may include basic grayscale transformation, noise filtering, and median filtering. For example, to solve the problem of non-uniform illumination, a median filter may be applied to the original image 119 to achieve a gradient image 123 ( Fig.12 ). This can be accomplished by selecting a pixel radius sufficient to provide a resulting image containing only the gradient of the background illumination of the original image 119. The gradient image 123 is then subtracted from the original image 119 to correct for the uneven illumination and generate an enhanced image 121 that does not contain the uneven illumination.

[0058] The image enhancement step 113 may apply image enhancement statically, dynamically, or both. For example, the pixel radius of the median filter may be a statically set constant value, or may be dynamically and adaptively derived from the features of each original image 119 captured by the optical system 9. Fig.13 As shown, some forms of image enhancement step 113 may include step 125, whereby it is determined whether the working image should undergo enhancement. If step 125 determines that the working image should be enhanced, then step 125 proceeds to step 127. In step 127, it is determined whether static enhancement should be applied. If step 127 determines that static enhancement should be applied, then in the case of applying static enhancement, step 127 proceeds to step 129. If step 127 determines that static enhancement should not be applied, then step 127 proceeds to step 131. In step 127, static enhancement is applied to the working image, and step 127 returns to step 125. In step 131, dynamic enhancement is applied to the working image, and step 131 returns to step 125. If step 125 determines that the working image should not be further enhanced, then step 125 proceeds to the end.

[0059] Fig.14 An example of a method for dynamic image enhancement 133 is illustrated. The method for dynamic image enhancement 133 involves dynamically determining an appropriate pixel radius for median filter enhancement. The method for dynamic image enhancement 133 starts from step 135, whereby the size of the microorganism shown in the original image 119 is determined. The size of the microorganism in the original image 119 can change according to a variety of environments and parameters related to the inoculum and the entire optical system 9. Therefore, although the literal size of the specific microorganism being tested is usually constant in nature, the relative size of the microorganism shown in the original image 119 is dynamic and variable because, for example, the parameters of lens objectification are different. Once step 135 determines the size of the microorganism in the original image 119, step 135 proceeds to step 137. In step 137, the pixel radius is derived from the determined microorganism size. As a general example, if step 135 determines that the maximum length of any given microorganism in the original image 119 is five pixels, it can be determined that the pixel radius is greater than five, so that the filtered image only represents the gradient contained in the background. After step 135 derives the pixel radius based on the dynamically determined microorganism size, step 135 proceeds to step 139. In step 139, a gradient image 123 is generated based on processing the original image 119 with the derived pixel radius. After generating the gradient image 123, step 139 moves to step 141. In step 141, the gradient image 123 is subtracted from the original image 119 to generate the enhanced image 121. Thereafter, the method of dynamic image enhancement 133 proceeds to the end.

[0060] like Fig. 9 , 15 As shown in FIG. 16 , the image segmentation step 115 is used to segment the image into different regions containing pixels representing microorganisms as foreground or background. The image segmentation step 115 converts the enhanced image 121 into a segmented image 143, as shown in FIG. Fig.15 The image segmentation step 115 generates a binary image from the enhanced image 121, where each pixel is equal to a value of 0 or 1, where 0 refers to the background and 1 refers to a portion of a specific microorganism.

[0061] Image artifacts such as noise can be removed by applying a noise reduction filter before applying the segmentation algorithm. Segmentation can be obtained using a static threshold or an adaptive image thresholding method using, for example, an Otsu cluster-based thresholding algorithm. In this algorithm, grayscale samples are clustered into two parts as background and foreground (object), or alternatively modeled as a mixture of two Gaussians. The threshold of the specific image thresholding algorithm used can be determined dynamically, depending on the overall image provided to the image segmentation step 115 and the relative grayscale of the image. For example, the inoculation system 5 or another element of the system 1 can be configured to apply Negro black to each test hole 20 to enhance the image because Negro black does not attach to certain microorganisms (e.g., bacteria). Compared to the original image 119 without Negro black, this can change the relative grayscale in the original image 119 and require different thresholds for the segmentation algorithm. In some forms of the optimized AST method 101, the threshold can be dynamically determined by searching for edges in several regions of the image. These edges are transition points between the background and the microorganisms. Therefore, the threshold can then be calculated as the average gray value of the pixels on each side of the located edge.

[0062] like Fig.16 As shown, some forms of the image segmentation step 115 may begin at step 145. In step 145, a determination is made whether to apply a noise reduction filter to the enhanced image 121. If step 145 determines that a noise reduction filter should be applied, then step 145 proceeds to step 147, where the noise reduction filter is applied. Thereafter step 147 proceeds to step 149. If step 145 determines that a noise reduction filter should not be applied, then step 145 proceeds directly to step 149. In step 149, a decision is made as to whether a threshold value is to be determined dynamically. If step 149 determines that a threshold value should be determined dynamically, then step 149 proceeds to step 151, where the threshold value is determined. Thereafter step 149 proceeds to step 153. If step 149 determines that a threshold value is not to be determined dynamically, then a static predetermined threshold value is used, and step 149 proceeds directly to step 153. In step 153, the enhanced image 121 is segmented using the selected threshold value, and step 153 and the image segmentation step 115 thereafter proceed to an end.

[0063] Once the image segmentation step 115 generates the segmented image 143, the image segmentation step 115 proceeds to the data extraction step 117. In the data extraction step 117, the background and foreground pixels are considered to yield information about the number of microorganisms in the sample (count-C), the area occupied by the microorganisms in the sample (area-A), and the ratio of the area occupied by the microorganisms to the number of microorganisms (A / C). In some forms of the data extraction step 117, the actual microorganism count is compared to the average microorganism count to determine whether an error occurred during the image capture process. The comparison may include a standard deviation from the average microorganism count to summarize the microorganism comparison.

[0064] The data extraction step 117 can be configured to derive information about the number of microorganisms in the image. In some embodiments of the data extraction step 117, the number of foreground pixels in the segmented image 143 can be counted according to a predefined width and / or length to determine the number of microorganisms in the imaged portion of the inoculum. The counting algorithm can be divided into two separate algorithms, one for counting rod-shaped microorganisms and one for counting spherical microorganisms, because the outline of the underlying microorganism provides corresponding different foreground pixel shapes in the segmented image 143. For example, the counting algorithm can be configured to consider a square of 2×2 pixels of microorganisms for the purpose of counting spherical microorganisms, or a rectangle of 1×4 pixels of microorganisms for the purpose of counting rod-shaped microorganisms can be considered. In addition, the counting algorithm can be configured to process two algorithms to capture different three-dimensional orientations of rod-shaped microorganisms. For example, if the slender rod is positioned toward the end of the AST camera 35, it will have a very different outline when viewed in two dimensions by the AST camera 35. Therefore, two counting algorithms can be used during the counting phase of the image analysis cycle 102. Alternatively, the counting algorithm can be configured to consider any foreground pixel surrounded by background pixels and count it as a microorganism.

[0065] Data extraction step 117 can be configured to derive information about the total area occupied by all microorganisms in the image. In some forms of data extraction step 117, the total number of foreground pixels in segmented image 143 can be counted and compared to the total number of background pixels in segmented image 143. Data extraction step 117 can present the area count information in any format, including, for example, a percentage of 30%, or as a literal number of pixels, such as "138 foreground pixels out of a total of 450 pixels" or "138 foreground pixels and 312 background pixels."

[0066] The data extraction step 117 can be configured to derive information regarding the ratio of the total microorganism count to the total area occupied by all microorganisms in the image. This information can be used to determine whether the microorganisms experience elongation over time. Elongation is a precursor to death, indicating that the concentration of the antimicrobial dilution has a negative impact on the microorganism. More specifically, elongation may occur when microorganisms (e.g., bacteria) encounter an effective amount of an antibiotic drug.

[0067] For example, Fig.17 and 18 A microorganism represented by the Escherichia coli (E. coli) bacterium is depicted. Fig.17 Depicts E. coli after exposure to dilutions of the antibiotic ampicillin. Fig.18 E. coli without any ampicillin exposure is depicted. Fig.17 The example shows that ampicillin exposure results in abnormal growth in the form of elongations, which is a precursor to bacterial death, and that the concentration of ampicillin is sufficient to neutralize the bacteria. Fig.18 Normal growth is illustrated. It follows that the ratio of the total area occupied by all bacteria to the bacterial count will increase over time to indicate that each bacterium is elongated. For example, if a particular image sample contains 100 bacteria, and each bacterium measures approximately 4 pixels each, then the area occupied by 100 bacteria within the image is 400 pixels. In this image sample, the ratio of bacterial area to bacterial count is 4.0. Over time, if each bacterium in the sample elongates to approximately 12 pixels each, then the area occupied by 100 bacteria in the latter image is 1200 pixels, and the ratio of bacterial area to bacterial count in the latter image is 12.0. This increase in ratio indicates that the concentration of ampicillin is effectively neutralizing the E. coli bacteria because the bacteria are elongated, which in turn indicates imminent death.

[0068] Once the data extraction step 117 obtains the desired information from the segmented image 143, the image analysis loop 102 terminates. The optimized AST method 101 iteratively performs the image analysis loop 102 at set time intervals to determine how the microorganisms in each test well 20 change and react to a specific antimicrobial dilution pairing. In addition, the optimized AST method 101 iteratively performs the image analysis loop 102 for each test well 20 associated with the microorganism to determine how the microorganism reacts with each concentration of the antimicrobial dilution. For example, assuming that the tested microorganism is an Escherichia coli bacterium, and three test wells 20 are tested, each of which has 20 microliters of solution. The first test well 20 may contain an antimicrobial dilution of 1 microgram / milliliter (μg / ml), the second test well 20 may contain an antimicrobial dilution of 2μg / ml, and the third test well 20 may contain an antimicrobial dilution of 4μg / ml. The optimized AST method 101 performs an image analysis cycle 102 on each of the three test wells at each set time interval to determine (a) how each antimicrobial dilution affects the microorganisms; (b) how each antimicrobial dilution performs relative to the other antimicrobial dilutions. If the data shows that the 1 μg / ml antimicrobial dilution is as effective as the 2 and 4 μg / ml antimicrobial dilutions in neutralizing the microorganisms, then the 1 μg / ml antimicrobial dilution is the MIC.

[0069] An exemplary form of the optimized AST method 101 is Fig.19 154 and 155. In step 155, the system waits for the time period threshold value of setting, so that the microorganism in the selected test hole 20 has enough time to provide new information about the growth rate or the reaction to the specific antimicrobial dilution. The optimized AST method 101 can be preferably configured to utilize a 30-minute time period as the threshold value, but different threshold values ​​can be used in different embodiments and / or in the case of different microorganisms. For example, some bacteria or yeast or other microorganisms can react very quickly and provide relevant information to carry out MIC determination within an hour. In this case, the optimized AST method 101 can be configured to perform image analysis circulation 102 every five minutes to capture the data about the rapidly changing environment in the test hole 20. Other microorganisms can react relatively slowly to the antimicrobial dilution, so the time threshold value is that one hour may be more appropriate. Once step 155 waits for the specified time period threshold value, step 155 just moves to step 157.

[0070] In step 157, one iteration of the image analysis loop 102 is performed for the particular microorganism using the selected test well 20. As described above, the iteration of the image analysis loop 102 yields data regarding the growth rate of the microorganism within the selected test well 20. After performing the iteration of the image analysis loop 102, step 157 moves to step 159. In step 159, the data collected in step 157 is stored and / or updated in a memory, which may be in the form of a database, a flat file, or any other similar memory or storage device. In some embodiments of the optimized AST method 101, step 159 stores the data collected in step 157 in a database 71 ( Figure 6 ). Once step 159 stores / updates the collected data, step 159 moves to step 161.

[0071] In step 161, the optimized AST method 101 determines whether sufficient data has been collected to determine the MIC. For example, this can be done by determining whether the data collected for the test wells matches the data used to train the machine learning classifier for determining the MIC. If more data is needed to accurately determine the MIC, step 161 returns to step 155 and waits to execute another image analysis cycle 102 to collect more data at a future time interval. If step 161 determines that a sufficient amount of data has been collected, step 161 proceeds to step 163.

[0072] In step 163, the MIC is determined and reported (e.g., by HMI 59) in step 165. This determination is based on data collected during each iteration of the image analysis cycle 102 for all antimicrobial dilutions of the microbial sample and control wells in which the antimicrobial is not present. As described in the following section, this determination may involve applying the collected data to a machine learning classifier that has been previously trained to make growth predictions, and then determining the MIC based on which of the predicted test wells exhibit growth or are inhibited.

[0073] III. Machine Learning for Determining Minimum Inhibitory Concentration (MIC)

[0074] Fig. 20 Depicted is an exemplary decision tree classifier that can be used in some embodiments to determine MIC using the parameters listed in Table 1. Such a classifier can be created, for example, by collecting a training data set by periodically (e.g., every 30 minutes starting at 90 minutes after inoculation and ending at 4 hours after inoculation; every 30 minutes starting at 90 minutes after inoculation and ending at 6 hours after inoculation; etc.) capturing images of test well groups inoculated with a particular microorganism and different concentrations of antimicrobial agent and images of control wells inoculated with a microorganism but not containing any antimicrobial agent, and extracting data from those images, such as previously described in Fig. 9 The captured data, or a subset thereof, and the observation of which wells showed growth (G) or inhibition (I) after 16 hours can then be fed to various algorithms, such as QUEST (Quick, Unbiased and Efficient Statistical Tree) or CHAID (Chi-squared Automatic Interaction Detector) to create a decision tree that will make a prediction of whether a particular well exhibits growth (G) or inhibition (I). Figures 20 to 22 Some examples of predictions of whether coagulase negative staphylococci show growth or inhibition after 16 hours in the presence of different concentrations of daptomycin, erythromycin, and oxacillin are made after 4 hours based on the following decision trees created in this way using QUEST: area covered after 4 hours (referred to in this embodiment as c9), percent increase in area between 3.5 and 4 hours (calculated as area at 4 hours - area at 3.5 hours divided by area at 3.5 hours, and referred to in this embodiment as r9, or ratio 9-8), and growth in control wells at 4 hours (referred to in this embodiment as growth 9, or g9). The resulting decision tree is shown in FIG. Fig. 20 (for oxacillin), 21 (for daptomycin), and 22 (erythromycin).

[0075] In general, the methods described can be used to create decision trees that will make growth predictions for a variety of microorganisms in the presence of various antimicrobial agents at varying concentrations. Tables 1 and 2 list categories of microorganisms and antimicrobial agents that are considered suitable for use in decision tree classifiers of the type described above, as well as some examples of specific organisms and antimicrobial agents within those categories:

[0076] Table 1

[0077]

[0078]

[0079] Table 2

[0080]

[0081]

[0082]

[0083] It is contemplated that the above methods may also be used for microorganisms and antimicrobial agents not listed in Tables 1 and 2. For example, it is contemplated that a decision tree classifier such as that described above may be used for growth prediction of microorganisms in the Enterobacteriaceae, Pseudomonadaceae, Acinetobacteraceae, Micrococcaceae, and / or Streptococcusaceae families.

[0084] Variations on the above methods are possible. For example, in some cases, rather than viewing the classifications directly as predictions of growth or inhibition for the test wells whose data are provided to a decision tree, they may be further processed to create classifications. For example, in some embodiments, when data are provided to, for example, Fig.23A and 23B When using the decision tree shown in , the number of leaves connected to their parent node by a yes (i.e., inhibition) branch can be compared to the number of leaves connected to their parent node by a no (i.e., growth) branch, and in either case, the larger number can be considered a growth / inhibition prediction for that well. Regardless of how they are made, once predictions can be made for each test well, the concentration in the test well with the lowest antimicrobial concentration that still produces an inhibition (I) prediction can be considered the MIC.

[0085] Table 3: Exemplary factors for growing or inhibiting a decision tree

[0086]

[0087]

[0088] Other approaches are possible and will be immediately apparent to one of ordinary skill in the art. For example, in some embodiments, rather than training a single decision tree, a training technique (e.g., XG Boost) can be used to generate an ensemble of multiple trees whose outputs can be combined to provide a higher reliability prediction. There can be different numbers of trees in such an ensemble, from 2 trees to hundreds of trees. During testing, when XG Boost software was used to create an ensemble of 500 decision trees using the factors listed in Table 3 based on training data collected using the following: Acinetobacter baumannii (A. baumannii), Citrobacter freundii cplx (C. freundiicplx), Enterobacter aerogenes (E. aerogenes), Enterobacter cloacae (E. cloacae), Escherichia coli (E. coli), Klebsiella oxytoca (K. oxytoca), Klebsiella pneumoniae (K. pneumoniae), Pseudomonas aeruginosa (P. aeruginosa), Pseudomonas fluorescens (P. fluorescens), Serratia marcescens (S. marcescens), Salmonella spp. (Salmonella spp.) sp.), Proteus mirabilis, P. rettani; and the antibiotics cefepime (β-lactam), cefotaxime, ceftazidime, gentamicin, levofloxacin, meropenem, and tetracycline.

[0089] Table 4: Accuracy measures obtained during decision tree testing

[0090]

[0091]

[0092] (There was an indeterminate result where two consecutive wells when sorted by concentration were given an inhibition classification while the next well by concentration gave a growth classification - a situation which is preferably considered in production to indicate a technical error or borderline growth and should be determined by manual intervention after a full 16 to 18 hour growth period).

[0093] Unexpectedly, during this test, it is found that this ensemble is suitable for every kind of tested microorganism to carry out growth prediction, rather than every kind of microorganism or every kind of antimicrobial agent need independent ensemble or tree.As a result, in some embodiments, the biological test system can be configured with the ensemble of decision tree that will be used for any kind of Gram-negative bacteria regardless of the specific Gram-negative bacteria in the tested antimicrobial agent or sample how, rather than using different decision trees or ensembles for every kind of microorganism and antimicrobial agent.Certainly, independent decision tree or ensemble can be trained for the data of the specific combination of microorganism and antimicrobial agent, and it is also possible to provide MIC prediction for the specific combination of microorganism and antimicrobial agent.Therefore, the above discussion about application decision tree sorter and ensemble and the ensemble sorter that is trained to use the parameter specified in table 3 can be applied to various microorganisms and antimicrobial agent and should be understood as only illustrative, and should not be considered as restriction.

[0094] Other types of machine learning models can also be used to make MIC predictions. For example, in some embodiments, rather than using, for example, Figures 20 to 23B Instead of the decision tree shown in (or an ensemble of such trees), one can use a decision tree with, for example Fig.25The neural network of the structure shown in .When the neural network of this type is trained to use the following to carry out growth / inhibition prediction, it is found that the results shown in Table 5 are provided when 130 kinds of microorganisms / antimicrobial combinations are tested: the area covered by the microorganisms of the reading obtained from 0 to 4 hours every 30 minutes under 10X magnification using a digital microscope, the count of microorganisms (limited to one micron of objects for this training purpose), the ratio of area to count, and the area covered by microorganisms in the test wells and the ratio of the area covered by microorganisms in the growth (no antimicrobial) wells. The specific organisms and antimicrobial agents used in this test are Acinetobacter baumannii, Citrobacter freundii cplx, Enterobacter cloacae, Escherichia coli, Klebsiella pneumoniae, Klebsiella oxytoca, Pseudomonas aeruginosa, Pseudomonas fluorescens, Serratia marcescens, Salmonella, Proteus mirabilis, Pruvidonella rettgeri and cefepime, cefotaxime, ceftazidime, gentamicin, levofloxacin, meropenem, tetracycline. The input nodes, output nodes, hidden layer codes, and response mapping code sets of the neural network trained in this manner are shown in Table 6. Of the 130 microorganism / antimicrobial combinations tested, the neural network provided G or I predictions that resulted in interpretable MICs for 126 of the tested substances. Four test results were determined to be uninterpretable because they contained an I, I, G sequence (e.g., 0.03G, 0.06G, 0.12G, 0.25G, 0.5I, 1G, 2I, 4I, 8G, 16I) following a previous G, I, G pattern, which (and can be in production) is considered to indicate a technical error or weak signal requiring manual intervention after a full 16 to 18 hour growth period.

[0095] Table 5: Accuracy measures obtained during testing of neural network G / I predictions.

[0096]

[0097]

[0098] Table 6: Example neural network parameters and code

[0099]

[0100]

[0101]

[0102]

[0103]

[0104]

[0105]

[0106]

[0107]

[0108]

[0109]

[0110]

[0111]

[0112]

[0113] There may also be variations in the way the machine learning module is used. For example, in some cases, rather than training the model to simply make growth / inhibition predictions as described above, the model may be trained to make MIC predictions directly. Fig.24 An example of the structure of a neural network that can be used to make this type of prediction is provided.

[0114] In some embodiments, in order to use e.g. Fig.24 MIC prediction can be performed using a neural network as shown in , and test wells containing a certain concentration of antimicrobial agent can be inoculated with microorganisms. Figures 20 to 23B Data can be collected by taking measurements in the manner described in the context of decision trees and neural network classifiers of 2000 and 25. These measurements, along with the identity and concentration of the antimicrobial agent in the test well, can then be provided as input to the input nodes of the neural network. The values ​​provided at the output nodes of the network can then be considered as an indication of the likelihood that the concentration corresponding to the output node is the MIC of the microorganism added to the test well. For example, when using a network that has been trained to provide the likelihood that a first concentration is the MIC at a first output node, the likelihood that a second concentration is the MIC at a second output node, and the likelihood that a third concentration is the MIC at a third output node, the concentration corresponding to the output node with the highest value (which may be different from the concentration of the antimicrobial agent in the test well from which the data was collected) can be considered as the MIC.

[0115] In some embodiments, rather than using e.g. Fig.24 Rather than taking the values ​​of the output nodes of the network shown in as directly indicating the MIC, the values ​​from such output nodes can be used as input to another function that will provide a (potentially more reliable) MIC value. For example, in some embodiments, a separate set of measurements can be collected from each of a plurality of test wells having different concentrations of the same antimicrobial agent, and these sets of measurements can each be provided as input to, for example, Fig.24 Each of those neural networks will then provide a prediction of the most likely MIC, and those predictions can be provided as input to another function (e.g., a statistical function such as a mode or median determination function) which will provide a prediction of what will be considered the true MIC of the sample. To illustrate, consider Table 7 below, which shows how the predicted MIC is obtained from the output of multiple neural networks during testing.

[0116] Table 7: Example MIC derivation from multiple neural network outputs

[0117]

[0118]

[0119] During testing, it was unexpectedly discovered that using this approach allowed a single neural network with the input nodes, output nodes, hidden layer codes, and response mapping codes listed in Table 8 to create MIC predictions with substantial and absolute agreement listed in Table 9 for data covering species of Acinetobacter baumannii, Citrobacter freundii, Enterobacter aerogenes, Enterobacter cloacae, Escherichia coli, Klebsiella oxytoca, Klebsiella pneumoniae, Proteus mirabilis, Pseudomonas aeruginosa, Pseudomonas fluorescens, Proteus mirabilis, Pluvedenella rettgeri, Serratia marcescens, and Salmonella, as well as cefepime, cefotaxime, ceftazidime, gentamicin, levofloxacin, meropenem, and tetracycline.

[0120] Table 8: Example neural network parameters and code

[0121]

[0122]

[0123]

[0124]

[0125]

[0126]

[0127]

[0128]

[0129]

[0130]

[0131]

[0132]

[0133]

[0134]

[0135]

[0136]

[0137]

[0138]

[0139]

[0140]

[0141] Table 9: Accuracy measures obtained during testing of neural network MIC predictions.

[0142]

[0143]

[0144] As described above, neural networks can be used to simulate the growth pattern of bacterial isolates over time under various drug concentrations. As long as bacteria grow in holes without drugs, neural networks can be created to use the count and area of ​​the growth holes and make early predictions of MIC under various drug concentrations. Such a network can predict MIC growth within the first 6 hours (or within the first 3.5 to 4 hours of the network discussed in the case of Tables 6 and 9), which is highly consistent with the actual MIC readings at 16 to 24 hours. This can be applied to Gram-negative bacteria (including Enterobacteriaceae, Pseudomonas, Acinetobacter, Stenotrophomonas and Haemophilus) and Gram-positive bacteria (including Staphylococcus, Enterococcus and Streptococcus), as well as the following drug classes: penicillins, β-lactams, β-lactams with inhibitors, carbapenems, cephalosporins, monolactams, tetracyclines, aminoglycosides, macrolides, glycopeptides, lincosamides, oxazolidinones, quinolones, folate inhibitors and ansamycins.

[0145] After creating a model, such as a decision tree classifier, decision tree ensemble, neural network, or other type of model, it may be combined with multiple other models (e.g., where a particular model, such as Figures 20 to 22Decision trees for certain microorganism-antimicrobial combinations in some embodiments) are preferably deployed together with, for example, Figure 6 In some embodiments, the computer system shown in the embodiment of the present invention can be used to accelerate the MIC determination relative to the situation that is possible with traditional visual analysis. Then, when the sample is analyzed on a machine controlled by such a computer system or integrated therewith, the image and data corresponding to the image and data used to train the model can be captured and extracted using, for example, the technology described herein, and the data can be used to generate MIC predictions. In some variations, this can include identifying the microorganisms in the sample (e.g., using a classifier, such as described in Venkatesh Vijaykumar, Classifying Bacterial Species using Computer Vision and Machine Learning, International Journal of Computer Applications Vol.151, No.8, October, 2016 at 23-26, the disclosure of which is incorporated by reference herein) and selecting an appropriate decision tree or other type of model. In other variations, this type of model selection can be omitted, and replaced by a model that has been trained to predict a variety of microorganisms and / or antimicrobial agents that can be used.

[0146] IV. Exemplary Combinations

[0147] The following examples relate to a variety of non-exhaustive ways in which the teachings herein may be combined or applied. It should be understood that the following examples are not intended to limit the scope of any claims that may be proposed at any time in this application or in subsequent documents of this application. There is no disclaimer. The following examples are provided for illustrative purposes only. It is expected that the various teachings herein can be arranged and applied in many other ways. It is also expected that some changes may omit certain features mentioned in the following examples. Therefore, unless otherwise explicitly pointed out by the inventor or a successor with an interest in the inventor later, the aspects or features mentioned below should not be considered critical. If any claim proposed in this application or in a subsequent document related to this application includes additional features other than those mentioned below, it should not be presumed that these additional features are added for any reason related to patentability.

[0148] Example 1

[0149] The method comprises: (a) incubating a first plurality of test mixtures in a plurality of test wells 20 using an incubator system 7 of a biological test system 1, wherein: (i) each test mixture from the first plurality of test mixtures is inoculated with a first biological sample; (ii) each test mixture from the first plurality of test mixtures has a concentration of a first antimicrobial agent; (iii) in each test mixture from the first plurality of test mixtures, a concentration of a first antimicrobial solution in the test mixture is different from a concentration of the first antimicrobial solution in each other test mixture from the first plurality of test mixtures; (iv) each test mixture from the first plurality of test mixtures is inoculated with the same first biological sample; and (v) the first plurality of test mixtures contain a growth mixture having a zero concentration of the first antimicrobial agent; (b) in a plurality of imaging (c) for each test mixture from the first plurality of test mixtures having a first antimicrobial agent concentration greater than zero, obtaining a plurality of machine learning outputs by executing, by a processor 51 of a computer system 49, steps comprising: (i) determining a plurality of feature values ​​for the test mixture, wherein: (A) each feature value corresponds to a parameter from a plurality of parameters; and (B) determining the plurality of feature values ​​for the test mixture based on the images of the test mixture captured at the plurality of imaging times; and (ii) providing the plurality of feature values ​​determined for the test mixture to a machine learning model; and (d) the processor 51 of the computer system 49 generating a MIC prediction for the first biological sample based on the plurality of machine learning outputs.

[0150] Example 2

[0151] The method described in Example 1, wherein: (a) the multiple imaging times include an earliest imaging time that is separated from the start of incubation by a first duration; (b) each imaging time from the multiple imaging times except the earliest imaging time is separated from its previous imaging time by a second duration; and (c) the second duration is shorter than the first duration.

[0152] Example 3

[0153] The method of Example 2, wherein the first duration is 90 minutes and the second duration is 30 minutes.

[0154] Example 4

[0155] The method of any one of Embodiments 1 to 3, wherein the plurality of imaging times comprises a latest imaging time that is four hours from the start of incubation.

[0156] Example 5

[0157] The method of any of Examples 1 to 4, wherein, for each test mixture from the first plurality of test mixtures having a first antimicrobial agent concentration greater than zero: (a) the plurality of feature values ​​for the test mixture comprises a first set of feature values ​​and a second set of feature values; (b) the first set of feature values ​​is based on an image captured at a first time from the plurality of imaging times; (c) the second set of feature values ​​is based on an image captured at a second time from the plurality of imaging times; and (d) the plurality of parameters comprises a parameter group, wherein each parameter from the parameter group corresponds to one value from the first set of feature values ​​and one value from the second set of feature values.

[0158] Example 6

[0159] The method of Example 5, wherein, for each test mixture from the first plurality of test mixtures having a first antimicrobial agent concentration greater than zero, for each imaging time from the plurality of imaging times, the plurality of characteristic values ​​for the test mixture comprises: a rate of change value for each parameter from the parameter set.

[0160] Example 7

[0161] The method of any of Examples 5 to 6, wherein, for each test mixture from the plurality of test mixtures having a first antimicrobial agent concentration greater than zero: (a) the plurality of feature values ​​comprises a growing set of feature values; (b) for each imaging time from the plurality of imaging times, each parameter from the parameter set corresponds to a feature value from the growing set of feature values; and (c) the feature values ​​from the growing set of feature values ​​are based on images of the growth mixture captured at the plurality of imaging times.

[0162] Example 8

[0163] The method of any one of Embodiments 5-7, wherein the parameter set comprises microbial count.

[0164] Example 9

[0165] The method of any one of Examples 1 to 8, wherein: (a) the method further comprises: (i) incubating a second plurality of test mixtures, wherein: (A) each test mixture from the second plurality of test mixtures is inoculated with a second biological sample; (B) each test mixture from the second plurality of test mixtures has a concentration of a second antimicrobial agent; and (C) in each test mixture from the second plurality of test mixtures, the concentration of the second antimicrobial agent in the test mixture is different from the concentration of the second antimicrobial agent in each other test mixture from the second plurality of test mixtures; (ii) for each test mixture from the second plurality of test mixtures having a second antimicrobial agent concentration greater than zero, obtaining a second plurality of machine learning outputs by performing the steps comprising: providing a plurality of feature values ​​determined for the test mixture to a machine learning model; and (b) for each test mixture from the first plurality of test mixtures having a first antimicrobial agent concentration greater than zero and each test mixture from the second plurality of test mixtures having a second antimicrobial agent concentration greater than zero, the machine learning model provided with the plurality of feature values ​​determined for the test mixture is the same machine learning model.

[0166] Example 10

[0167] The method of Embodiment 9, wherein the first antimicrobial agent is different from the second antimicrobial agent.

[0168] Embodiment 11

[0169] The method of any of embodiments 9 to 10, wherein the first antimicrobial agent and the second antimicrobial agent are each classified into the group consisting of: (a) β-lactams; (b) aminoglycosides; (c) fluoroquinolones; (d) carbapenems; and (e) tetracyclines.

[0170] Example 12

[0171] The method of any one of embodiments 9 to 11, wherein the first antimicrobial agent and the second antimicrobial agent are each from the group consisting of: (a) cefepime; (b) cefotaxime; (c) ceftazidime; (d) gentamicin; (e) levofloxacin; (f) meropenem; and (g) tetracycline.

[0172] Embodiment 13

[0173] The method of any one of Examples 9 to 12, wherein: (a) the first biological sample comprises a first microorganism; (b) the second biological sample comprises a second microorganism; and (c) the first microorganism is different from the second microorganism.

[0174] Embodiment 14

[0175] The method of any one of Embodiments 9 to 13, wherein the first microorganism and the second microorganism are both Gram-negative.

[0176] Embodiment 15

[0177] The method of any one of embodiments 9 to 14, wherein the first microorganism and the second microorganism are both from the group consisting of: (a) Acinetobacter baumannii; (b) Citrobacter freundii; (c) Enterobacter aerogenes; (d) Enterobacter cloacae; (e) Escherichia coli; (f) Klebsiella oxytoca; (g) Klebsiella pneumoniae; (h) Proteus mirabilis; (i) Pseudomonas aeruginosa; (j) Pseudomonas fluorescens; (k) Providencia rettgeri; (1) Serratia marcescens; (m) Salmonella typhi; and (n) Salmonella paratyphi.

[0178] Example 16

[0179] The method of any of Examples 1 to 15, wherein: (a) for each test mixture from the first plurality of test mixtures having a first antimicrobial agent concentration greater than zero, obtaining the plurality of machine learning outputs comprises: obtaining an intermediate MIC prediction as a machine learning output for the test mixture after providing the plurality of feature values ​​determined for the test mixture to the machine learning model; and (b) generating a MIC prediction for the first biological sample comprises providing the plurality of machine learning outputs to a MIC creation function.

[0180] Embodiment 17

[0181] The method of Example 16, wherein the MIC prediction for the first biological sample is a median intermediate MIC prediction from the plurality of machine learning outputs.

[0182] Embodiment 18

[0183] The method of any of Examples 16-17, wherein, for at least one test mixture from the first plurality of test mixtures, the intermediate MIC prediction obtained as the machine learning output for the test mixture is a concentration of the first antimicrobial agent that is lower than the concentration of the first antimicrobial agent in the test mixture.

[0184] Embodiment 19

[0185] The method of any of Examples 16 to 18, wherein: (a) the machine learning model is a neural network having a plurality of output nodes; (b) each output node from the plurality of output nodes corresponds to a potential MIC; and (c) for each test mixture from the first plurality of test mixtures, the intermediate MIC prediction obtained as the machine learning output for the test mixture is the potential MIC corresponding to the output node having the highest value when the plurality of feature values ​​determined for the test mixture are provided to the neural network.

[0186] Embodiment 20

[0187] The method of any of Examples 1 to 19, wherein, for each test mixture from the first plurality of test mixtures having a first antimicrobial agent concentration greater than zero, providing an identification of the first antimicrobial agent to the machine learning model along with the plurality of feature values ​​determined for the test mixture.

[0188] Embodiment 21

[0189] The method of any of Examples 1 to 14, wherein: (a) for each test mixture from the first plurality of test mixtures having a first antimicrobial agent concentration greater than zero, obtaining the plurality of machine learning outputs comprises: obtaining a growth prediction as a machine learning output for the test mixture after providing the plurality of feature values ​​determined for the test mixture to the machine learning model; and (b) generating a MIC prediction for the first biological sample comprises: identifying the test mixture having the lowest concentration of the first antimicrobial agent for which an inhibition growth prediction is obtained as a machine learning output for the test mixture.

[0190] Embodiment 22

[0191] The method of Example 21, wherein: (a) the machine learning model is an ensemble comprising one or more decision trees, each decision tree having multiple leaf nodes, each leaf node connected to a parent node by a branch indicating growth or inhibition; and (b) for each test mixture from the first plurality of test mixtures having a first antimicrobial agent concentration greater than zero, obtaining a growth prediction for the test mixture comprises: when multiple feature values ​​of the test mixture are provided to the machine learning model, predicting growth or inhibition based on whether there are more leaf nodes connected to the parent node by branches indicating growth or whether there are more leaf nodes connected to the parent node by branches indicating inhibition.

[0192] Embodiment 23

[0193] The method of Example 21, wherein the machine learning model is a machine learning model trained to provide growth predictions at output nodes.

[0194] Embodiment 24

[0195] The method of any one of embodiments 9 to 14, wherein the first microorganism and the second microorganism are both from the group consisting of: (a) Acinetobacter baumannii; (b) Citrobacter freundii; (c) Enterobacter cloacae; (d) Escherichia coli; (e) Klebsiella oxytoca; (f) Pseudomonas aeruginosa; (g) Pseudomonas fluorescens; (h) Serratia marcescens; (i) Salmonella typhi; (j) Salmonella paratyphi; (k) Proteus mirabilis; and (1) Providencia rettgeri.

[0196] Embodiment 25

[0197] The method of any one of embodiments 1-21 or 23-24, wherein the parameter set comprises: (a) microbial area; and (b) a ratio of microbial area to microbial count.

[0198] Embodiment 26

[0199] The method of any one of embodiments 1 to 4 or 21, wherein: (a) the machine learning model is a decision tree classifier; (b) the method comprises: (i) determining the identification of a microorganism contained in the first biological sample; and (ii) selecting a decision tree classifier from a plurality of decision tree classifiers based on the identification of the microorganism.

[0200] Embodiment 27

[0201] The method described in Example 26, wherein the type of microorganism contained in the first biological sample is selected from the group of microorganism types consisting of: (a) Gram-negative fermenting bacteria; (b) Gram-negative non-fermenting bacteria; (c) Gram-positive Micrococcaceae; and (d) Gram-positive Streptococcusaceae.

[0202] Embodiment 28

[0203] The method of Example 27, wherein: (a) the type of microorganism contained in the first biological sample is a Gram-negative fermenting bacterium; and (b) the microorganism contained in the first biological sample is selected from the group consisting of: (i) Aeromonas hydrophila; (ii) Citrobacter freundii; (iii) Citrobacter krusei; (iv) Enterobacter aerogenes; (v) Enterobacter cloacae; (vi) Escherichia coli; (vii) Klebsiella oxytoca; (viii) Klebsiella pneumoniae; (ix) Morganella morganii; (x) Proteus mirabilis; (xi) Proteus pennisetum; (xii) Proteus vulgaris; (xiii) Pruvidenia stuartii; (xiv) Pruvidenia rettgeri; (xv) Salmonella paratyphi; (xvi) Salmonella typhi; (xvii) Serratia marcescens; (xviii) Shigella dysenteriae; (xix) Shigella flexneri; (xx) Shigella sonnei; (xxi) Vibrio cholerae; and (xxii) Yersinia enterocolitica.

[0204] Embodiment 29

[0205] The method described in Example 27, wherein: (a) the type of microorganism contained in the first biological sample is a Gram-negative non-fermenting bacterium; and (b) the microorganism contained in the first biological sample is selected from the group of microorganisms consisting of: (i) Acinetobacter baumannii; (ii) Acinetobacter hemolyticus; (iii) Acinetobacter loffei; (iv) Moraxella species; (v) Pseudomonas aeruginosa; (vi) Pseudomonas alcaligenes; (vii) Pseudomonas fluorescens; (viii) Pseudomonas putida; and (ix) Stenotrophomonas maltophilia.

[0206] Embodiment 30

[0207] The method described in Example 27, wherein: (a) the type of microorganism contained in the first biological sample is Gram-positive Micrococcaceae; and (b) the microorganism contained in the first biological sample is selected from the group of microorganisms consisting of: (i) Staphylococcus aureus; (ii) Staphylococcus capitis; (iii) Staphylococcus epidermidis; (iv) Staphylococcus haemolyticus; (v) Staphylococcus saprophyticus; (vi) Staphylococcus worderi; (vii) Staphylococcus simulans; and (viii) Staphylococcus lugdunensis.

[0208] Embodiment 31

[0209] The method described in Example 27, wherein: (a) the type of microorganism contained in the first biological sample is Gram-positive Streptococcus family; and (b) the microorganism contained in the first biological sample is selected from the group of microorganisms consisting of: (i) Enterococcus avium; (ii) Enterococcus faecalis; (iii) Enterococcus faecium; (iv) Enterococcus gallinarum; (v) Enterococcus raffinose; (vi) Streptococcus bovis; (vii) Streptococcus pyogenes; and (viii) Streptococcus pneumoniae.

[0210] Embodiment 32

[0211] The method described in Example 26, wherein the type of microorganism contained in the first biological sample is selected from the group of microorganism types consisting of: (a) Enterobacteriaceae; (b) Pseudomonadaceae; (c) Acinetobacteraceae; (d) Micrococcaceae; and (e) Streptococcus.

[0212] Embodiment 33

[0213] The method of Example 26, wherein the type of the first antimicrobial agent is selected from the group of antimicrobial agent types consisting of: (a) aminoglycosides; (b) ansamycins; (c) beta-lactams, penicillins; (d) beta-lactam plus inhibitor combination; (e) carbapenems; (f) cephems; (g) fluoroquinolones; (h) folate pathway inhibitors; (i) fosfomycin; (j) glycopeptides; (k) lincosamides; (l) lipopeptides; (m) macrolides; (n) Oxazolidinones; and (o) tetracyclines.

[0214] Embodiment 34

[0215] The method of Example 33, wherein: (a) the first antimicrobial agent is an aminoglycoside; and (b) the first antimicrobial agent is selected from the group of antimicrobials consisting of: (i) amikacin; (ii) gentamicin; and (iii) tobramycin.

[0216] Embodiment 35

[0217] The method of Example 33, wherein: (a) the first antimicrobial agent is a type of β-lactam, penicillin; and (b) the first antimicrobial agent is selected from the group of antimicrobials consisting of: (i) penicillin; and (ii) oxacillin.

[0218] Embodiment 36

[0219] The method of Example 33, wherein: (a) the type of the first antimicrobial agent is a β-lactam plus an inhibitor combination; and (b) the first antimicrobial agent is selected from the group of antimicrobials consisting of: (i) amoxicillin + clavulanic acid; (ii) ampicillin + sulbactam; (iii) ceftazidime + avibactam; (iv) cefuroxime + tazobactam; (v) piperacillin + tazobactam; and (vi) ticarcillin + clavulanic acid.

[0220] Embodiment 37

[0221] The method of embodiment 33, wherein: (a) the type of the first antimicrobial agent is a carbapenem; and (b) the first antimicrobial agent is selected from the group of antimicrobials consisting of: (i) doripenem; (ii) ertapenem; (iii) imipenem; and (iv) meropenem.

[0222] Embodiment 38

[0223] The method of Example 33, wherein: (a) the type of the first antimicrobial agent is a cephem; and (b) the first antimicrobial agent is selected from the group of antimicrobials consisting of: (i) cefazolin; (ii) cefepime; (iii) cefotaxime; (iv) cefoxitin; (v) ceftaroline; (vi) ceftazidime; (vii) ceftriaxone; (viii) ceftizoxime; and (ix) cefuroxime.

[0224] Embodiment 39

[0225] The method of Example 33, wherein: (a) the first antimicrobial agent is a fluoroquinolone; and (b) the first antimicrobial agent is selected from the group of antimicrobials consisting of: (i) ciprofloxacin; and (ii) levofloxacin.

[0226] Embodiment 40

[0227] The method described in Example 26, wherein: (a) the multiple parameters consist of: (i) microbial area; (ii) the difference in microbial area; and (iii) microbial growth; (b) the multiple characteristic values ​​include: (i) the microbial area at the last imaging time from the multiple imaging times; (ii) the difference in microbial area from the second to last imaging time to the last imaging time; and (iii) microbial growth at the last imaging time.

[0228] Embodiment 41

[0229] A non-transitory computer readable medium encoding instructions for performing the method of any one of embodiments 1 to 40.

[0230] Embodiment 42

[0231] A computer system 49 is configured with instructions for causing the biological testing system 1 to execute the method of any one of embodiments 1 to 40.

[0232] Embodiment 43

[0233] A biological test system (1) comprising a processor (51) configured with a set of operable computer instructions, which when executed cause the system (1) to perform a method comprising: (a) incubating a first plurality of test mixtures in a plurality of test wells (20) using an incubator system (7) of the biological test system (1), wherein: (i) each test mixture from the first plurality of test mixtures is inoculated with a first biological sample; (ii) each test mixture from the first plurality of test mixtures has a concentration of a first antimicrobial agent; (iii) in each test mixture from the first plurality of test mixtures, the concentration of the first antimicrobial agent solution in the test mixture is different from the concentration of the first antimicrobial agent solution in each other test mixture from the first plurality of test mixtures; (iv) each test mixture from the first plurality of test mixtures is inoculated with the same first biological sample; and (v) The first plurality of test mixtures comprises a growth mixture in which the first antimicrobial concentration is zero; (b) at a plurality of imaging times, for each test mixture from the first plurality of test mixtures, capturing an image of the test mixture using an AST camera (35), wherein each imaging time occurs after initiation of incubation; (c) for each test mixture from the first plurality of test mixtures in which the first antimicrobial concentration is greater than zero, obtaining a plurality of machine learning outputs by a processor (51) performing steps comprising: (i) determining a plurality of feature values ​​for the test mixture, wherein: (A) each feature value corresponds to a parameter from a plurality of parameters; and (B) determining the plurality of feature values ​​for the test mixture based on the images of the test mixture captured at the plurality of imaging times; and (ii) providing the plurality of feature values ​​determined for the test mixture to a machine learning model; and (d) the processor (51) generating a MIC prediction for the first biological sample based on the plurality of machine learning outputs.

[0234] Embodiment 44

[0235] The biological testing system (1) described in Example 43, wherein: (a) the multiple imaging times include an earliest imaging time that is separated from the start of incubation by a first duration; (b) each imaging time from the multiple imaging times except the earliest imaging time is separated from its previous imaging time by a second duration; and (c) the second duration is shorter than the first duration.

[0236] Embodiment 45

[0237] The biological testing system (1) of Example 44, wherein, for each test mixture from the first plurality of test mixtures having a first antimicrobial agent concentration greater than zero: (a) the plurality of feature values ​​for the test mixture comprises a first set of feature values ​​and a second set of feature values; (b) the first set of feature values ​​is based on an image captured at a first time from the plurality of imaging times; (c) the second set of feature values ​​is based on an image captured at a second time from the plurality of imaging times; and (d) the plurality of parameters comprises a parameter set, wherein each parameter from the parameter set corresponds to one value from the first set of feature values ​​and one value from the second set of feature values.

[0238] Embodiment 46

[0239] The biological testing system of embodiment 45, wherein, for each test mixture from the first plurality of test mixtures having a first antimicrobial agent concentration greater than zero, the plurality of characteristic values ​​for the test mixture, for each imaging time from the plurality of imaging times, comprises: a rate of change value for each parameter from the set of parameters.

[0240] Embodiment 47

[0241] The biological test system (1) of any of Examples 45 to 46, wherein, for each test mixture from the plurality of test mixtures having a first antimicrobial agent concentration greater than zero: (a) the plurality of feature values ​​comprises a growing set of feature values; (b) for each imaging time from the plurality of imaging times, each parameter from the parameter set corresponds to a feature value from the growing set of feature values; and (c) the feature values ​​from the growing set of feature values ​​are based on images of the growth mixture captured at the plurality of imaging times.

[0242] Embodiment 48

[0243] The biological test system (1) of any one of embodiments 43 to 47, wherein: (a) the method further comprises: (i) incubating a second plurality of test mixtures, wherein: (A) each test mixture from the second plurality of test mixtures is inoculated with a second biological sample; (B) each test mixture from the second plurality of test mixtures has a concentration of a second antimicrobial agent; and (C) in each test mixture from the second plurality of test mixtures, the concentration of the second antimicrobial agent in the test mixture is different from the concentration of the second antimicrobial agent in each other test mixture from the second plurality of test mixtures; (ii) for each test mixture from the second plurality of test mixtures having a second antimicrobial agent concentration greater than zero, obtaining a second plurality of machine learning outputs by performing the steps comprising: providing a plurality of feature values ​​determined for the test mixture to a machine learning model; and (b) for each test mixture from the first plurality of test mixtures having a first antimicrobial agent concentration greater than zero and each test mixture from the second plurality of test mixtures having a second antimicrobial agent concentration greater than zero, the machine learning model provided with the plurality of feature values ​​determined for the test mixture is the same machine learning model.

[0244] Embodiment 49

[0245] The biological test system (1) of embodiment 48, wherein the first antimicrobial agent is different from the second antimicrobial agent.

[0246] Embodiment 50

[0247] The biological testing system (1) of any one of embodiments 48 to 49, wherein: (a) the first biological sample comprises a first microorganism; (b) the second biological sample comprises a second microorganism; and (c) the first microorganism is different from the second microorganism.

[0248] Embodiment 51

[0249] The biological testing system (1) of any one of embodiments 43 to 50, wherein: (a) for each test mixture from the first plurality of test mixtures having a first antimicrobial agent concentration greater than zero, obtaining the plurality of machine learning outputs comprises: obtaining an intermediate MIC prediction as a machine learning output for the test mixture after providing the plurality of feature values ​​determined for the test mixture to the machine learning model; and (b) generating the MIC prediction for the first biological sample comprises providing the plurality of machine learning outputs to a MIC creation function.

[0250] Embodiment 52

[0251] The biological testing system (1) described in Example 51, wherein, for at least one test mixture from the first plurality of test mixtures, the intermediate MIC prediction obtained as the machine learning output for the test mixture is a concentration of the first antimicrobial agent that is lower than the concentration of the first antimicrobial agent in the test mixture.

[0252] Embodiment 53

[0253] The biological test system (1) of any one of Examples 51 to 52, wherein: (a) the machine learning model is a neural network having multiple output nodes; (b) each output node from the multiple output nodes corresponds to a potential MIC; and (c) for each test mixture from the first plurality of test mixtures, the intermediate MIC prediction obtained as the machine learning output for the test mixture is the potential MIC corresponding to the output node having the highest value when the multiple feature values ​​determined for the test mixture are provided to the neural network.

[0254] Embodiment 54

[0255] The biological testing system (1) of any one of embodiments 43 to 53, wherein, for each test mixture from the first plurality of test mixtures having a first antimicrobial agent concentration greater than zero, an identification of the first antimicrobial agent is provided to the machine learning model along with a plurality of feature values ​​determined for the test mixture.

[0256] Embodiment 55

[0257] The biological testing system (1) of any one of Examples 43 to 54, wherein: (a) for each test mixture from the first plurality of test mixtures having a first antimicrobial agent concentration greater than zero, obtaining the plurality of machine learning outputs comprises: obtaining a growth prediction as a machine learning output for the test mixture after providing the plurality of feature values ​​determined for the test mixture to the machine learning model; and (b) generating a MIC prediction for the first biological sample comprises: identifying the test mixture having the lowest concentration of the first antimicrobial agent for which an inhibition growth prediction is obtained as a machine learning output for the test mixture.

[0258] Embodiment 56

[0259] The biological testing system (1) described in Example 55, wherein: (a) the machine learning model is an ensemble comprising one or more decision trees, each decision tree having multiple leaf nodes, each leaf node connected to a parent node by a branch indicating growth or inhibition; and (b) for each test mixture from the first plurality of test mixtures having a first antimicrobial agent concentration greater than zero, obtaining a growth prediction for the test mixture includes: when multiple feature values ​​of the test mixture are provided to the machine learning model, predicting growth or inhibition based on whether there are more leaf nodes connected to the parent node by branches indicating growth or whether there are more leaf nodes connected to the parent node by branches indicating inhibition.

[0260] Embodiment 57

[0261] The biological testing system (1) described in Example 55, wherein the machine learning model is a machine learning model trained to provide growth predictions at output nodes.

[0262] Embodiment 58

[0263] The biological testing system (1) described in Example 43, wherein: (a) the machine learning model is a decision tree classifier; (b) the method includes: (i) determining the identification of the microorganism contained in the first biological sample; and (ii) selecting a decision tree classifier from a plurality of decision tree classifiers based on the identification of the microorganism.

[0264] Embodiment 59

[0265] The biological testing system (1) described in Example 58, wherein: (a) the multiple parameters are composed of the following: (i) microbial area; (ii) the difference in microbial area; and (iii) microbial growth; (b) the multiple characteristic values ​​include: (i) the microbial area at the last imaging time from the multiple imaging times; (ii) the difference in microbial area from the second to last imaging time to the last imaging time; and (iii) microbial growth at the last imaging time.

[0266] VOther

[0267] It should be understood that in the above embodiments and claims, the statement that something is "based on" other things should be understood to mean that the thing is at least partially determined by the thing indicated as being based on it. In order to indicate that something must be completely determined based on other things, it is described as "exclusively based on" anything that it must be completely determined by.

[0268] It should be understood that any embodiment described herein may include a variety of other features in addition to or in place of those described above. By way of example only, any embodiment described herein may also include one or more of the various features disclosed in any of the various references incorporated herein by reference.

[0269] It should be understood that any one or more of the teachings, expressions, embodiments, examples, etc. described herein may be combined with any one or more of the other teachings, expressions, embodiments, examples, etc. described herein. Therefore, the above teachings, expressions, embodiments, examples, etc. should not be viewed in isolation from each other. In view of the teachings herein, a variety of suitable ways in which the teachings herein may be combined will be apparent to those of ordinary skill in the art. Such modifications and variations are intended to be included within the scope of the claims.

[0270] It is understood that any patent, publication, or other public material allegedly incorporated herein by reference is incorporated herein in whole or in part only to the extent that the incorporated material does not conflict with existing definitions, statements, or other public materials set forth in the present disclosure. Therefore, and to the extent necessary, the disclosure explicitly set forth herein supersedes any conflicting material incorporated herein by reference. Any material or portion thereof allegedly incorporated herein but conflicting with existing definitions, statements, or other public materials set forth herein will be incorporated only to the extent that there is no conflict between the incorporated material and the existing public materials.

[0271] Various forms of the present invention have been shown and described, and further improvements of the methods and systems described herein may be achieved by appropriate modifications by those of ordinary skill in the art without departing from the scope of the present invention. Several such potential modifications have been mentioned, and other modifications will be apparent to those skilled in the art. For example, the embodiments, forms, geometric figures, materials, dimensions, ratios, steps, etc. discussed above are illustrative and not essential. Therefore, the scope of the present invention should be considered in light of the appended claims, and should be understood as not being limited to the details of the structures and operations shown and described in the specification and drawings.

Claims

1. A method comprising: (a) incubating a first plurality of test mixtures (19) in a plurality of test wells (20) using an incubator system (7) of a biological test system (1), wherein: (i) inoculating each test mixture from the first plurality of test mixtures (19) with a first biological sample; (ii) each test mixture from the first plurality of test mixtures (19) has a concentration of a first antimicrobial agent; (iii) in each test mixture from the first plurality of test mixtures (19), a concentration of a first antimicrobial agent in the test mixture is different from a concentration of the first antimicrobial agent in each other test mixture from the first plurality of test mixtures (19); (iv) inoculating each test mixture from the first plurality of test mixtures (19) with the same first biological sample; and (v) the first plurality of test mixtures (19) comprises a growth mixture having a first antimicrobial agent concentration of zero; (b) capturing, for each test mixture from said first plurality of test mixtures (19), an image (119) of the test mixture using an antimicrobial susceptibility testing (AST) camera (35) at a plurality of imaging times, wherein each of said imaging times occurs after initiation of incubation; (c) for each test mixture from the first plurality of test mixtures (19) having a first antimicrobial agent concentration greater than zero, obtaining a plurality of machine learning outputs by a processor (51) of a computer system (49) performing the steps comprising: (i) determining a plurality of characteristic values ​​of the test mixture, wherein: (A) each eigenvalue corresponds to a parameter from a plurality of parameters; and (B) determining a plurality of characteristic values ​​of the test mixture based on images (119) of the test mixture captured at the plurality of imaging times; (ii) providing the plurality of feature values ​​determined for the test mixture to a machine learning model, wherein the machine learning model comprises an ensemble comprising a plurality of decision trees, each decision tree having a plurality of leaf nodes, each leaf node being connected to a parent node by a branch indicating growth or inhibition; and (iii) obtaining a growth prediction for the test mixture as a machine learning output, comprising: predicting growth or inhibition based on whether there are more leaf nodes connected to the parent node by branches indicating growth or whether there are more leaf nodes connected to the parent node by branches indicating inhibition when multiple feature values ​​of the test mixture are provided to the machine learning model; as well as (d) the processor (51) of the computer system (49) generates a minimum inhibitory concentration (MIC) prediction for the first biological sample based on the plurality of machine learning outputs, comprising identifying a test mixture having a lowest concentration of the first antimicrobial agent that yields a prediction of growth inhibition as the machine learning output for the test mixture, wherein the machine learning model is a decision tree classifier, and in: (a) The plurality of parameters consists of the following: (i) microbial area; (ii) differences in microbial area; and (iii) microbial growth; (b) The plurality of characteristic values ​​include: (i) the microorganism area at the last imaging time from among the plurality of imaging times; (ii) the difference in microbial area from the second-to-last imaging time to the last imaging time; and (iii) Microbial growth at the last imaging time.

2. A method comprising: (a) incubating a first plurality of test mixtures (19) in a plurality of test wells (20) using an incubator system (7) of a biological test system (1), wherein: (i) inoculating each test mixture from the first plurality of test mixtures (19) with a first biological sample; (ii) each test mixture from the first plurality of test mixtures (19) has a concentration of a first antimicrobial agent; (iii) in each test mixture from the first plurality of test mixtures (19), a concentration of a first antimicrobial agent in the test mixture is different from a concentration of the first antimicrobial agent in each other test mixture from the first plurality of test mixtures (19); (iv) inoculating each test mixture from the first plurality of test mixtures (19) with the same first biological sample; and (v) the first plurality of test mixtures (19) comprises a growth mixture having a first antimicrobial agent concentration of zero; (b) capturing, for each test mixture from said first plurality of test mixtures (19), an image (119) of the test mixture using an antimicrobial susceptibility testing (AST) camera (35) at a plurality of imaging times, wherein each of said imaging times occurs after initiation of incubation; (c) for each test mixture from the first plurality of test mixtures (19) having a first antimicrobial agent concentration greater than zero, obtaining a plurality of machine learning outputs by a processor (51) of a computer system (49) performing the steps comprising: (i) determining a plurality of characteristic values ​​of the test mixture, wherein: (A) each eigenvalue corresponds to a parameter from a plurality of parameters; and (B) determining a plurality of characteristic values ​​of the test mixture based on images (119) of the test mixture captured at the plurality of imaging times; (ii) providing the plurality of feature values ​​determined for the test mixture to a machine learning model, wherein the machine learning model is a neural network having a plurality of output nodes, and each output node from the plurality of output nodes corresponds to a potential MIC; and (iii) obtaining an intermediate MIC prediction for the test mixture as an output of the machine learning, wherein the intermediate MIC prediction obtained as an output of the machine learning for the test mixture is a potential MIC corresponding to an output node having a highest value when a plurality of feature values ​​determined for the test mixture are provided to the neural network; as well as (d) the processor (51) of the computer system (49) generates a minimum inhibitory concentration (MIC) prediction for the first biological sample based on the plurality of machine learning outputs, the generating comprising providing the plurality of machine learning outputs to a MIC creation function, in: (a) The plurality of parameters consists of the following: (i) microbial area; (ii) differences in microbial area; and (iii) microbial growth; (b) The plurality of characteristic values ​​include: (i) the microorganism area at the last imaging time from among the plurality of imaging times; (ii) the difference in microbial area from the second-to-last imaging time to the last imaging time; and (iii) Microbial growth at the last imaging time.

3. The method of any one of claims 1 to 2, wherein: (a) the plurality of imaging times comprises an earliest imaging time separated from the start of incubation by a first duration; (b) each imaging time from the plurality of imaging times other than the earliest imaging time is separated from its previous imaging time by a second duration; and (c) The second duration is shorter than the first duration. The method of claim 3 , wherein the first duration is 90 minutes and the second duration is 30 minutes.

5. The method of any one of claims 1 to 2, wherein the plurality of imaging times comprises a latest imaging time that is four hours from the start of incubation.

6. The method according to any one of claims 1 to 2, wherein For each test mixture from the first plurality of test mixtures (19) having a first antimicrobial agent concentration greater than zero: (a) the plurality of characteristic values ​​of the test mixture comprises a first set of characteristic values ​​and a second set of characteristic values; (b) the first set of feature values ​​is based on an image captured at a first time from the plurality of imaging times; (c) the second set of feature values ​​is based on an image captured at a second time from the plurality of imaging times; and (d) the plurality of parameters comprises a parameter group, wherein each parameter from the parameter group corresponds to a value from the first set of feature values ​​and a value from the second set of feature values.

7. The method of claim 6, wherein: For each test mixture from the first plurality of test mixtures (19) having a first antimicrobial agent concentration greater than zero, the plurality of characteristic values ​​for the test mixture comprises, for each imaging time from the plurality of imaging times: a rate of change value for each parameter from the parameter set.

8. The method of claim 6, wherein: For each test mixture from the plurality of test mixtures (19) having a first antimicrobial agent concentration greater than zero: (a) the plurality of eigenvalues ​​comprises an eigenvalue growth group; (b) for each imaging time from the plurality of imaging times, each parameter from the parameter set corresponds to an eigenvalue from the eigenvalue growing set; and (c) Eigenvalues ​​from the eigenvalue growth set are based on images of the growth mixture captured at the plurality of imaging times.

9. The method of claim 6, wherein the set of parameters comprises microbial count.

10. The method of any one of claims 1 to 2, wherein: (a) The method further comprises: (i) incubating a second plurality of test mixtures, wherein: (A) inoculating each test mixture from the second plurality of test mixtures with a second biological sample; (B) each test mixture from the second plurality of test mixtures has a concentration of a second antimicrobial agent; and (C) in each test mixture from the second plurality of test mixtures, the concentration of the second antimicrobial agent in the test mixture is different from the concentration of the second antimicrobial agent in each other test mixture from the second plurality of test mixtures; (ii) for each test mixture from the second plurality of test mixtures having a second antimicrobial agent concentration greater than zero, obtaining a second plurality of machine learning outputs by performing the steps comprising: providing a plurality of feature values ​​determined for the test mixture to the machine learning model; and (b) for each test mixture from the first plurality of test mixtures having a first antimicrobial agent concentration greater than zero and each test mixture from the second plurality of test mixtures having a second antimicrobial agent concentration greater than zero, the machine learning model provided with the plurality of feature values ​​determined for the test mixture is the same machine learning model.

11. The method of claim 10, wherein the first antimicrobial agent is different from the second antimicrobial agent.

12. The method of claim 10, wherein the first antimicrobial agent and the second antimicrobial agent are each classified into the group consisting of: (a) β-lactam; (b) aminoglycosides; (c) Fluoroquinolones; (d) carbapenems; and (e) Tetracyclines.

13. The method of claim 10, wherein the first antimicrobial agent and the second antimicrobial agent are each selected from the group consisting of: (a) Cefepime; (b) cefotaxime; (c) ceftazidime; (d) gentamicin; (e) levofloxacin; (f) meropenem; and (g) Tetracycline.

14. The method of claim 10, wherein: (a) the first biological sample comprises a first microorganism; (b) the second biological sample comprises a second microorganism; and (c) the first microorganism is different from the second microorganism.

15. The method of claim 14, wherein the first microorganism and the second microorganism are both Gram-negative.

16. A non-transitory computer-readable medium for use in a biological testing system (1), having computer program instructions stored thereon, which, when executed by one or more processors (51), cause the one or more processors (51) to perform the method of any one of claims 1 to 15.

17. A computer system (49) configured with computer program instructions which, when executed by one or more processors (51) in a biological testing system (1), cause the one or more processors (51) to perform the method of any one of claims 1 to 15.

18. A biological testing system (1) comprising one or more processors (51) configured with a set of computer program instructions, which, when executed by the one or more processors (51), cause the one or more processors (51) to perform the method of any one of claims 1 to 15.

Images