Platform for antimicrobial susceptibility testing and bacterial identification

Abstract

Disclosed is a device for performing growth detection, identification, and antimicrobial susceptibility testing (AST) functions for microbial species to achieve a sample-to-answer result. Methods for performing microbial growth, microbial species identification, and AST using such a device are also disclosed. The device may include a housing configured to receive a sample plate, a sample port configured to receive a sample, a fluid distribution system constructed and arranged to introduce the sample into one or more sample wells of the sample plate, a sample plate imaging system, and a controller configured to collect data from the sample plate imaging system. The controller may detect growth of the microbial species, identify the microbial species, and perform AST on the microbial species once the data is collected.

Description

[Technical Field]

[0001] Technology field Aspects generally relate to the rapid detection of bacteremia and fungemia (bloodstream infections) in patients exhibiting signs and symptoms that warrant rapid detection, identification, and direct antimicrobial susceptibility testing (AST) of the organisms associated with the infection. Testing is performed directly from blood samples or other human, animal, or environmental samples without the need for standard microbial culture. Results are available within eight hours from the time of blood collection. Rapid organism growth detection and identification can also be performed from various pharmaceutical ingredients and final products, providing significant advantages to the pharmaceutical industry for rapid identification and quantification of biological bioburden. [Background technology]

[0002] background Bloodstream infections are responsible for one-fifth of all deaths worldwide, including one-third of all hospital deaths in the United States, and nearly one-quarter of inpatient mortality, accounting for more than 11 million deaths annually. Bloodstream infections are also the single most costly inpatient condition, costing $42 billion annually in the United States alone. Survival depends on patients receiving effective antibiotic treatment as soon as possible. Rapid detection of the causative organism, along with early availability of AST results, can positively impact patient care and management, reducing complications and potential mortality.

[0003] Survival depends on treating with the appropriate antibiotic as quickly as possible, and this requires knowing the bacterial identity to a clinically usable level of specificity. In the United States, the primary organization that sets these usability levels is called the CLSI. Therefore, the most urgent diagnostic goal is to determine a CLSI-level bacterial ID. Despite many newer techniques, microscopy and Gram staining still play an important role in ID. However, the eye is limited in what it can see microscopically, making these other techniques necessary. Unfortunately, they are usually complex, expensive, and difficult to obtain, a significant problem, especially with the rise of antimicrobial resistance. Summary of the Invention [Means for solving the problem]

[0004] Summary of the Invention According to certain embodiments, a device configured to perform growth detection, identification, and antimicrobial susceptibility testing (AST) on a microbial species is provided. The device may include a housing configured to receive a sample plate. The device may include a sample port configured to receive a sample suspected of containing the microbial species. The device may further include a fluid distribution system constructed and arranged to introduce the sample into one or more sample wells of the sample plate. The device may include a sample plate imaging system. The device may further include a controller configured to collect data from the sample plate imaging system. Once the data is collected, the controller can detect growth of the microbial species, identify the microbial species, and perform AST on the microbial species.

[0005] In some embodiments, the fluid distribution system may include a gantry, a fluid dispensing head operably coupled to the gantry, and a pump fluidly connected to the fluid dispensing head, the pump including a stepper motor shaft having an absolute position magnetic encoder. The fluid distribution system may be constructed and arranged to facilitate digital growth detection.

[0006] In some embodiments, the sample plate imaging system may include a camera configured with an optical system, the camera being connected to the fluid dispensing head, a light source, and a detector array. In certain embodiments, the detector array may be a photodetector, such as a charge-coupled device (CCD), a complementary metal-oxide semiconductor (CMOS) detector, or a photodiode.

[0007] In further embodiments, the device may include a user interface, e.g., a screen, e.g., a touch screen. In some embodiments, the controller may be configured to transmit information regarding growth, identification, and / or AST to a user, e.g., via the user interface.

[0008] In a further embodiment, the device includes a heater, for example, located below the sample plate or between the sample plate and the optics, to facilitate sample preparation.

[0009] In some embodiments, the device may be further constructed and configured to allow for Gram staining of the sample.

[0010] In some embodiments, the device may be constructed and engineered to perform all three functions in less than about 8 hours. In certain embodiments, the device may be constructed and engineered to perform all three functions in less than about 6 hours. For example, the device may perform AST in less than about 1 hour. In some embodiments, the device may use the same technique for both growth detection and AST. In some cases, identification of microbial species is based on a stain-free technique, for example, without the use of Gram staining.

[0011] In some embodiments, the microbial species may include bacterial species. In some embodiments, the microbial species may be selected from the genera Acinetobacter, Escherichia, Klebsiella, Pseudomonas, Enterococcus, Streptococcus, and Staphylococcus. In some embodiments, the microbial species may belong to one of the following bacterial groups: Enterococci, Staphylococci, Enterobacterales, Acinetobacter baumannii, and Pseudomonas aeruginosa. In further embodiments, the microbial species may include non-bacterial species of fungi, mycobacteria, fecal parasites, blood parasites, or tissue parasites.

[0012] In some embodiments, the sample may be a whole blood sample of a subject. In some embodiments, the sample may be a blood component of a subject, such as plasma. In some embodiments, the sample may be a non-blood biological fluid sample of a subject. In further embodiments, the sample may relate to a pharmaceutical manufacturing component or a finished end product.

[0013] In some embodiments, the device may be configured to perform one or more of growth detection, identification and AST on a second microbial species.

[0014] According to certain embodiments, kits are provided for performing growth detection, identification, and antimicrobial susceptibility testing (AST) on microbial species. The kits may include any of the devices and sample plates disclosed herein.

[0015] In some embodiments, the sample plate may include a first portion for growth detection and a second portion for AST. In the separate sample plate, one or more wells in the second portion of the sample plate may be preloaded with lyophilized antibiotics. In some embodiments, the antibiotics may be preloaded according to a serial dilution scheme for AST.

[0016] In some embodiments, the sample plate may contain an area for sample smearing. In some embodiments, one or more wells of the sample plate may be defined by a geometric shape selected to promote sensitivity. In some embodiments, the sample plate may contain 384 or 1536 wells.

[0017] In further embodiments, the kit may include a source of growth medium or a detection amplifier. In further embodiments, the kit may include a source of Gram stain. In further embodiments, the kit may include a sample bottle.

[0018] According to certain embodiments, a method for detecting growth of a microbial species is provided. The method for detecting growth of a microbial species may include providing a sample suspected of containing the microbial species. The method for detecting growth of a microbial species may include acquiring a time series of images of the sample. The method for detecting growth of a microbial species may further include detecting growth of the microbial species by a time-dependent change in at least one of light and color over the time series of images.

[0019] In some embodiments, detecting growth may include detecting growth using an RGB-sensitive imaging sensor in a device, such as a device disclosed herein. Growth of the microbial species may be interpreted as a statistically significant change from baseline in at least one of light and color. In some embodiments, a decrease in light and / or a change in color over a time series of images may be interpreted as growth of the microbial species.

[0020] In some embodiments, the method of detecting growth of a microbial species may include acquiring a time series of images of each well of a sample plate and detecting growth in each well. In some embodiments, detecting growth may further include comparing one or more of the time series of images between wells, between a well and a control, or between a well and a reference time series. In some embodiments, the method of detecting growth of a microbial species may include comparing data across multiple wells to increase the sensitivity of growth detection.

[0021] In some embodiments, growth detection of a microbial species may be achieved in a duration of less than 5 hours. In some embodiments, several wells of a sample plate, e.g., at least one well, with observable growth may be used to determine several replicative microbial cells.

[0022] In some embodiments, the method of detecting growth of a microbial species may further include quantifying the bioburden of the sample based on the growth detection. In some embodiments, the method of detecting growth of a microbial species may further include one or more remedial actions based on the bioburden determination.

[0023] In some embodiments, the method of detecting growth of a microbial species may further comprise identifying and / or subjecting the microbial species to AST upon detecting growth.

[0024] In certain embodiments, the method of detecting growth of microbial species may further include detecting polymicrobial infection by comparing time series images between wells of a sample plate. The presence of polymicrobial infection may be assessed based on differences in color or doubling time across wells of a sample plate, e.g., a well plate, e.g., a microwell plate.

[0025] In some embodiments, the method of detecting growth of a microbial species may further include processing the time series of images to enhance their quality to facilitate the earliest possible detection of growth.

[0026] In further embodiments, the method of detecting growth of a microbial species may include adding a growth medium or detection amplification agent to a sample, which may be any suitable sample, such as a whole blood sample, a blood component sample, other bodily fluid, a product related to a pharmaceutical manufacturing component or a finished end product, a filtration membrane, or any other sample from which a microbial species may be harvested and analyzed.

[0027] In some embodiments, the method may be characterized by digital growth detection.

[0028] According to one aspect, a method for identifying a microbial species is provided. The method for identifying a microbial species may include imaging a sample having a plurality of microorganisms of a microbial species to obtain a series of multi-microbial images. The method for identifying a microbial species may include segmenting the multi-microbial images. The method for identifying a microbial species may further include measuring parameters of each segmented microorganism to obtain a multi-dimensional distribution of the measured parameters. The method for identifying a microbial species may further include classifying the microbial species based on the multi-dimensional distribution of the measured parameters.

[0029] In some embodiments, the measured parameters may relate to one or more of the size, shape, intrinsic color, arrangement, and other morphological characteristics of the microbial species. For example, at least one of the one or more measured parameters may be selected from the group consisting of width, length, interior density, membrane thickness, color heterogeneity, color intensity, curvature, tapering, aspect ratio, and concavity. In certain embodiments, the method for identifying a microbial species may include intrinsic color data associated with the microbial species without staining, i.e., without Gram staining.

[0030] In some embodiments, classification of microbial species may involve the use of machine learning, e.g., a machine learning algorithm, trained on at least one of the following modalities: (i) unstained slides imaged in direct light, (ii) unstained slides imaged in indirect light, e.g., dark-field microscopy, and (iii) pairings of pre- and post-stained images. In certain embodiments, classifying microbial species may involve probability distributions. In certain embodiments, classifying may be performed by a hierarchical approach. In certain embodiments, classification may be performed by majority voting, decision trees, or the relative entropy of the observed tally and a reference distribution of known microbial species.

[0031] In some embodiments, a method for identifying a microbial species may include distinguishing between Gram-positive and Gram-negative bacteria with a first confidence score. Using this first distinction, each bacterial species may then be identified with a second confidence score.

[0032] In further embodiments, the method for identifying a microbial species may include filtering the images, for example, using a colored filter. For example, a blue filter may be applied to all collected images to provide image sharpening. In some embodiments, the method for identifying a microbial species may include collecting images of each sample at multiple focal lengths of the optical system. As a non-limiting example, images of each sample are taken at a series of small or fine focal length intervals. In some embodiments, the fine focal length interval may be approximately 0.1 μm to 2 μm. In certain embodiments, the focal length interval may be 0.5 μm. In some embodiments, the method for identifying a microbial species may include acquiring multiple images at each focal length and combining the images.

[0033] In further embodiments, the method for identifying a microbial species may include selecting an image having a maximum value of one or more quality metrics for measurement. The quality of the selected image may be improved by removing or smoothing numerical noise before segmentation. In some embodiments, the method for identifying a microbial species may include performing dimensionality reduction, such as PCA or UMAP, on the multidimensional distribution of the measured parameters.

[0034] In some embodiments, measuring 100 randomly selected segmented microorganisms may be sufficient to identify a microbial species with approximately 93-97% confidence.

[0035] In some embodiments, the microbial species sample may be from a growth detection study, which may be performed using a device disclosed herein or the same or a different sample plate.

[0036] In a further embodiment, the method of identifying a microbial species may include identifying a second microbial species.

[0037] According to certain embodiments, methods are provided for performing growth detection, identification, and antimicrobial susceptibility testing (AST) on microbial species. The disclosed methods may be configured to perform all three functions, i.e., growth detection, identification, and antimicrobial susceptibility testing, in less than about 8 hours.

[0038] In some embodiments, one or both of the detection of growth of a microbial species and the identification of a microbial species may be performed using the methods disclosed herein.

[0039] In some embodiments, AST may be based on, for example, a differential proliferation detection method as disclosed herein. For example, AST may include dilution time modeling (DTM). Using the methods disclosed herein, AST may achieve categorical agreement with a reference method in less than 1 hour, for example, in less than 1 hour. In some embodiments, the reference method for AST comparison may include broth microdilution.

[0040] In some embodiments, AST may be performed for an antibiotic selected from cefepime, meropenem, ciprofloxacin, and gentamicin.

[0041] In some embodiments, the AST results may be reported to an operator, a laboratory information system, and / or an electronic medical record, for example, transmitted by a device disclosed herein, or displayed using a user interface on a device disclosed herein.

[0042] In some embodiments, AST may be performed once the microorganism has been identified at the category level and before the microorganism has been identified at the species level. In some embodiments, growth detection, identification, and AST may each be performed on a single device, such as a device disclosed herein.

[0043] According to certain aspects, a quality control method for a pharmaceutical manufacturing process is provided, which may include assessing the bioburden of a sample containing a pharmaceutical ingredient or a finished end product by performing a method for detecting the growth of microbial species, for example, using any of the devices or methods disclosed herein.

[0044] In some embodiments, a quality control method for a pharmaceutical manufacturing process may include approving or rejecting a pharmaceutical ingredient or finished end product based on a comparison of the assessed bioburden to a threshold value.

[0045] According to one embodiment, a single device configured to perform growth detection, identification, and antimicrobial susceptibility testing (AST) functions for a microbial species is provided.

[0046] In some embodiments, the device is configured to perform all three functions in less than about 8 hours.

[0047] In some embodiments, the device is configured to perform all three functions in less than about 6 hours.

[0048] In some embodiments, the device performs the AST in less than about 1 hour.

[0049] In some embodiments, the device uses the same technique for both proliferation detection and AST.

[0050] In some embodiments, the identification function is based on a machine learning stain-free approach.

[0051] In some embodiments, the microbial species comprises at least one species from the genera Acinetobacter, Escherichia, Klebsiella, Pseudomonas, Enterococcus, Streptococcus, and Staphylococcus.

[0052] According to one aspect, a method for identifying microbial species by machine learning is provided, the method comprising a stain-free technique.

[0053] According to one embodiment, a method for detecting growth of a microbial species is provided. The method may include acquiring a plurality of images of a microbial sample using an image acquisition system. The method may include sending or transmitting one or more of the plurality of images to an image analysis system including a non-transitory computer-readable medium storing a sequence of computer-executable instructions for determining growth of a microbial species from one or more of the plurality of images of the microbial sample by manipulating data corresponding to pixel intensities of one or more regions of one or more of the plurality of images into a hybrid model representing growth dynamics of the microbial species.

[0054] According to one aspect, a method for performing growth detection, identification, and antimicrobial susceptibility testing (AST) on a microbial species is provided, the method being configured to perform all three functions in less than about 8 hours.

[0055] The accompanying drawings are not intended to be drawn to scale. In the drawings, each identical or nearly identical component shown in various figures is represented by a like numeral. For clarity, not every component may be labeled in every figure. [Brief explanation of the drawings]

[0056] [Figure 1] FIG. 1 illustrates the time-to-answer direction from blood (as a representative immediate starting material) according to one or more embodiments.

[0057] [Figure 2A-B] 2A-2C show schematic diagrams of a device for AST, according to one or more embodiments: Fig. 2A shows a front view of the device; Fig. 2B shows a side cross-sectional view of the device; and Fig. 2C shows a front cross-sectional view of the device. [Figure 2C] Same as above.

[0058] [Figure 3] FIG. 3 illustrates the workflow of an AST according to one or more embodiments.

[0059] [Figure 4] 4A-4B show results from experiments demonstrating proliferation in less than 5 hours (FIG. 4A) and less than 4 hours (FIG. 4B) according to one or more embodiments.

[0060] [Figure 5A-B] Figures 5A-5C show the results of growth of five different bacterial species and reduced dimensionality visualization to aid in the identification of unknown samples. Figure 5A shows segmented cells for each of the five different bacterial species. Figure 5B shows the reduced dimensionality relationship of the five different bacterial species using UMAP. Figure 5C shows the confidence in the identification of the unknown sample. [Figure 5C] Same as above.

[0061] [Figure 6A] Figures 6A-6C show a comparison of AST performed using the reference method and the method disclosed herein. Figure 6A is a 3D plot of E. coli growth as a function of meropenem concentration. Figure 6B is a 3D plot of E. coli growth as a function of cefepime concentration. Figure 6C shows the equation used to model growth. [Figure 6B-C] Same as above.

[0062] [Figure 7] FIG. 7 shows the AST of several organisms on standard antibiotics using the methods disclosed herein.

[0063] [Figure 8A-B] 8A-8C illustrate the use of a fluid handling system according to one or more embodiments. Figure 8A illustrates dispensing fluid into a standard well plate. Figure 8B illustrates filling a well plate with 10 μL of liquid. Figure 8C illustrates the standard deviation and standard error of the fluid handling system across different dispense volumes. [Figure 8C] Same as above.

[0064] [Figure 9] FIG. 9 illustrates a bacterial identification technique according to one or more embodiments.

[0065] [Figure 10A-B] 10A-10E illustrate a proposed AST workflow according to one or more embodiments. FIG. 10A illustrates image acquisition. FIG. 10B illustrates the first step of image processing. FIG. 10C illustrates segmentation of individual cells within the image. FIG. 10D illustrates fingerprint analysis of segmented cells. FIG. 10E illustrates statistical comparison with a reference image for identification. [Figure 10C-D] Same as above. [Figure 10E] Same as above. DETAILED DESCRIPTION OF THE INVENTION

[0066] Detailed Description According to one or more embodiments, devices and methods for both antibiotic susceptibility testing and microbial identification are disclosed. In some aspects, both objectives can be accomplished with a single, relatively inexpensive and compact device. Beneficially, the disclosed systems and methods can achieve both objectives in a relatively short time, estimated at eight hours or less. Workflow can be significantly streamlined and errors reduced. Whole blood or other biological specimens may be directly introduced into the disclosed devices, and pathogenic organisms are identified and AST results are generated within four to eight hours. Mortality rates associated with ineffective treatments can be halved.

[0067] Different antibiotics are used to treat different bacteria. For example, vancomycin is used empirically to cover S. aureus but not Escherichia coli or other Enterobacterales, which are commonly resistant. Conversely, broad-spectrum β-lactam antibiotics empirically cover Enterobacterales but not S. aureus. P. aeruginosa and Acinetobacter baumannii still require other antibiotics. Bacterial identification allows for narrowing treatment, tailoring it to the causative bacterial species or group. Targeted therapy reduces the risk of life-threatening side effects, such as superinfection with Clostridioides difficile, due to the disruption of the body's normal bacterial flora by broad-spectrum antibiotics.

[0068] Antimicrobial susceptibility testing (AST) is the process of determining which antibiotics bacteria will respond to. Typically, to perform AST, bacteria are grown at different antibiotic concentrations to determine the lowest concentration that inhibits growth, known as the minimum inhibitory concentration, or MIC. If the MIC is below a certain breakpoint (set by CLSI or a similar standard-setting organization), the organism is considered susceptible, meaning the antibiotic can be used. The gold-standard AST method is phenotypic AST, in which organisms are tested for how well they grow in the presence of an antibiotic. Phenotypic AST is preferred to genotypic AST, which is an indirect method in which bacterial genetic material is tested for the presence of DNA sequences that correlate with phenotypic behavior. Phenotypic AST is the mainstay of care.

[0069] A critical factor in patient survival is the time it takes from blood draw to AST result: sample-to-answer time (S2A). Mortality increases by 7.6% for every hour a patient is given the wrong antibiotic. Even today, the median S2A is 2 days, as shown in Figure 1, with a long tail that can extend to weeks for increasing multidrug-resistant organisms. Considering that the average inpatient stay for a bloodstream infection is 7.1 days, a long S2A can render the AST meaningless and could mean patient death.

[0070] Mortality rates double for each day a patient receives ineffective treatment, so a reduction of such magnitude is of great importance. International guidelines require that patients receive broad-spectrum empirical antibiotics immediately after cultures are drawn. The next dose of antibiotic is typically administered 6-8 hours later. According to one or more embodiments, AST results may be available before the next dose is administered, i.e., within 6-8 hours of culture collection. According to at least some embodiments, S2A does not exceed a standard 8-hour work shift, thus avoiding delays and errors associated with shift changes. A meaningful target S2A may be ≤6-8 hours. The devices and methods described herein meet this goal.

[0071] Embodiments can detect bacteria in pharmaceutical ingredients throughout the manufacturing process. Current methods can take up to 48-72 hours to detect potential bacterial contamination in pharmaceutical ingredients and finished products. Meaningful targets for bacterial (bioburden) detection can result in more efficient manufacturing, saving companies time and money.

[0072] According to one or more embodiments, computer vision, signal processing, and other technologies may be adapted to operate S2A on the same shift. Computer vision is the application of artificial intelligence / machine learning to image analysis and other optical signals to achieve superhuman speed and / or accuracy. Signal processing, as used herein, includes cleaning up time series to detect the earliest possible signs of growth.

[0073] According to one or more embodiments, AST may be achieved in less than one hour starting from a positive blood culture. In some embodiments, differential growth detection for AST may be implemented, including detecting subtle differences in well darkening as bacteria grow, depending on the antibiotic and its concentration.

[0074] According to one or more embodiments, the AST may be in accordance with the techniques disclosed in co-pending International (PCT) Application No. PCT / US2022 / 042509, which is hereby incorporated by reference in its entirety for all purposes. According to one or more embodiments, more sensitive sensors and algorithms may be implemented.

[0075] Traditionally, it takes a median of 22 hours for blood cultures to become positive, after which the organism must be subcultured for organism identification and AST, requiring an additional 12-24 hours. According to one or more embodiments described herein, positive blood cultures can be detected earlier and S2A reduced.

[0076] According to one or more embodiments, digital growth detection is disclosed. "Digital" refers to the ability to count the number of viable organisms in a sample in a multi-well format, since each positive well begins with only a single pathogen cell. For example, a microtiter plate format allows for digital growth detection, resulting in pure microcultures, because each such microculture begins with a single bacterial cell. In at least some embodiments, highly sensitive ab initio growth detection may be achieved. In some non-limiting embodiments, AST techniques (such as those disclosed in the incorporated co-pending applications above) may be similarly used for the growth detection discussed herein, in many humans, where a solution matrix performs pathogen or other organism identification (ID), such as bacterial ID. In the gold standard version of phenotypic AST, bacteria are grown in serial dilutions of antibiotics—1 μg / mL, 2 μg / mL, 4 μg / mL, etc.—to find the lowest concentration at which bacteria cannot grow. This lowest concentration is the antibiotic's minimum inhibitory concentration (MIC). Generally, a low MIC in the laboratory means that the drug is effective in patients, but the cutoff or breakpoint for susceptibility versus resistance varies by ID. In some cases, bacteria may share the same breakpoint. Current laboratory standards require bacterial identification before reporting AST results, and organism identification has become critical to the clinical process of identifying the causative agent of infection. Because ID is critical for AST reporting, it is essential that bacterial identification be included in the S2A device.

[0077] While bacterial ID can be performed rapidly, taking only minutes to hours with the appropriate combination of mass spectrometry, PCR, and / or biochemical techniques, these techniques not only require the purchase and maintenance of hundreds of thousands of dollars worth of equipment, but also require the growth of pure cultures, which in most cases can take 12–24 hours.

[0078] According to one or more embodiments, computer vision may be used for ID, and the associated architecture and pipeline may be tailored according to one or more embodiments rather than simply borrowing from existing (i.e., non-medical) neural networks.

[0079] According to one or more embodiments, color data of unstained organisms from the sensor (i.e., without Gram stain or other staining) may be used for identification. For example, Pseudomonas aeruginosa and E. coli are both Gram-negative rods, but Pseudomonas aeruginosa notoriously has a green-purple appearance, while E. coli is typically gray-white to yellow. Just as colony color and bacterial cell morphology have long aided microbiologists in presumptive bacterial ID, they can aid computer vision ID more specifically and faster (as fast as the device can prepare slides; just a few minutes) without the need for Gram stain or any other staining. On-device bacterial ID is an innovation that enables rapid S2A.

[0080] Gram staining has been a mainstay of microbiology for 140 years, allowing microscopic examination to identify bacteria. Two stains are used in the Gram staining procedure: one that makes Gram-positive organisms appear purple and one that makes Gram-negative organisms appear pink, helping to make the morphology of individual bacterial cells (e.g., their size and shape) and their arrangement relative to one another visible to the human eye using a microscope. By definition, they also help distinguish Gram-positive from Gram-negative organisms. Staining differences reflect useful biological differences between organisms that affect potential antibiotic treatment options. However, there are two important tradeoffs. First, any information related to the intrinsic (i.e., unstained / pre-stained) color of the organisms is lost when they are Gram-stained. It is well documented that biological colonies can have distinctive colors that can be informative for identification.

[0081] The described embodiments may use color as part of a machine learning approach to identify organisms. Second, and importantly, while Gram staining is a simple procedure, it involves several staining and washing steps using several chemicals and requires interpretation by a trained clinical microbiologist. Operator interaction is a well-documented problem. Therefore, it would be advantageous to be able to perform ID without the need for staining.

[0082] Unstained cells have a characteristic color. Gram staining may actually conceal some of the valuable information it provides. From a macroscopic perspective, bacterial colony color has been a key indicator of identification since the Gram stain was first described. Microscopically, the color of unstained bacteria can be visualized by computer. Therefore, omitting the Gram stain upgrades the Gram stain's binary purple-pink color to a richer, more intrinsic palette that should be more informative for bacterial identification. The Gram stain itself provides largely redundant information; most clinically important Gram-negative bacteria are rod-shaped, and most Gram-positive bacteria are cocci; exceptions differ in sequence (e.g., Neisseria) and / or shape and size (e.g., Clostridia). Without being bound by any particular theory, it is believed that the same bacterial cell wall physiology that produces purple versus pink coloration upon Gram staining is detectable from optical measurements. For example, low-magnification electron microscopy (EM) has been shown to predict Gram staining. Machine learning has further been used to predict complex hematoxylin and eosin (H&E) staining patterns from unstained human tissue, suggesting that optical measurements can similarly predict Gram stain patterns from unstained images.

[0083] According to one or more embodiments, photomicrographs of unstained slides may be used for identification. Deep networks are adept at noticing subtle differences in light and color, allowing them to identify bacteria from unstained slides. In some embodiments, three (related) modalities may be used for training: (i) unstained slides imaged in direct light, (ii) unstained slides imaged in indirect light, a method known as dark-field microscopy, and (iii) pairing of pre- and post-stain images, a method used in pathology to predict staining patterns but not for machine learning of staining indications, and not used for bacterial, fungal, or other microbiological identification.

[0084] According to one or more embodiments, identification may be achieved from unstained slides, particularly utilizing both the faint color of individual bacteria or other cells on the slide and color as determined by non-slide-based methods, e.g., the color of organisms growing in broth (liquid medium).

[0085] Individual bacteria on a Gram stain (or other stain) are described by size, shape, and color. A typical broth culture slide contains thousands of organisms. This abundance provides an opportunity for statistical learning based on bacterial size, shape, and other morphological parameters. Starting with an image, the overall approach is to use image segmentation to identify and isolate each bacterial (or other) cell in each image. (Non-cellular debris from blood, urine, or other matrices can also be segmented.)

[0086] According to one or more embodiments, individual bacteria may be measured. With regard to bacterial measurement in general, existing measurements that exist as part of descriptions in the research literature are in a wide range (e.g., 0.5-1 μm by 0.8-1.5 μm, a two-fold range in all dimensions), vary by source, are unreferenced, and generally reflect conventional wisdom dating back to textbooks a century before high-throughput measurements became available.

[0087] For the specific measurements assumed as part of this disclosure, bacteria are generally convex in shape, classically (mostly) somewhere between rods (bacilli) and spheres (cocci). They differ in such things as the curvature of their edges (boxcar), whether and / or to what extent they taper (coryneform), and their aspect ratios (coccobacillus, fusiform, filamentous). Formulas exist to interpolate between all these shapes parametrically, i.e., smoothly. The Cassini oval and the Cartesian oval are common examples. Harmonic functions are another type of formula used for shape interpolation. Different forms of f(x) (e.g., f(x) = ) for different k are possible.

number

number

[0088] According to one or more embodiments, each bacterium may be fitted using one or more of these equations, and parameters recorded. Parameters reflecting not just simple width and length but also the internal complexity and density differences of the cell, membrane thickness (by dark-field microscopy), and the color intensity and / or heterogeneity of each of these may be measured.

[0089] According to one or more embodiments, a distribution or histogram of these parameters may be created for each slide. Each slide is thereby converted into a multidimensional distribution of parameters, one dimension for each parameter measured. Instead of "E. coli is 0.5-1 μm by 0.8-1.5 μm," there is a multidimensional distribution that describes exactly how the length and width, as well as the other parameters mentioned above, are distributed for hundreds or thousands of individual cells on a slide.

[0090] According to one or more embodiments, a robust representation of each species and strain may be trained. Most simply, training consists of a database that maps each species (or strain) to its multidimensional distribution. The multidimensional distribution can be thought of as a point cloud, where each point is a single cell and the point's location is given by its measurement. Each species has a cloud of a different shape with regions of different density. Given an unknown organism, steps 1-3 disclosed herein are performed to create its cloud, find the most similar cloud in the database, and assign the corresponding bacterium identity.

[0091] Another implementation involves learning a representation or model of the cloud, e.g., using deep learning, maximum entropy modeling, or some other technique, and then using the resulting model to assign identities to unknowns. In such an implementation, the key is for the model to output the probability that a given bacterial cell belongs to the cloud; the set of probabilities for all cells can, in some embodiments, be used for an aggregate or "total" probability to assign identities or determine whether more than one identity exists. Thus, given a slide of an unknown organism, for example, each cell is segmented and measured, and then for each cell, a model of, e.g., E. coli is asked what the probability is that the unknown cell is found in the E. coli cloud. The result is a set of probabilities, i.e., a probability distribution, for all cells present that the unknown is E. coli. Such a distribution is generated for each organism in the database—E. coli, P. aeruginosa, S. aureus, etc.—and whichever organism's model gives the highest probability is assigned as the identity of the unknown.

[0092] According to one or more embodiments, a machine learning approach to microbial identification is disclosed that does not require Gram staining or other stains.

[0093] According to one or more embodiments, identity may be machine-learned from (i) unstained light microscopy images and / or (ii) dark-field images and / or (iii) color combinations in growth experiments. Dark-field microscopy images may still have color information that can be used.

[0094] According to one or more embodiments, non-bacterial organisms may be identified, which may include fungi such as yeast and Aspergillus, mycobacteria such as Mycobacterium tuberculosis, fecal parasites such as pinworms, blood parasites such as malaria and Babesia, and tissue parasites.

[0095] According to one or more embodiments, instead of deep learning, for example, parameterized learning may be implemented, in which a set of defined parameters can be varied.

[0096] According to one or more embodiments, individual organisms in a multi-microbial image may be segmented and classified.

[0097] According to one or more embodiments, color from microtiter growth experiments may be measured. Intensity can be considered a monochromatic measurement. So can optical density at a specific wavelength, e.g., 600 nm (yellow) is the most commonly used wavelength at which E. coli has its peak absorbance. The embodiments disclosed herein use color (e.g., R, G, B), measured, for example, by combining signals from specific sensors. The use of color in the disclosed embodiments also includes color change over time, which may be represented, for example, as a color wheel, where color is represented circumferentially by the angle around the wheel, time is represented by the radius, and the center indicates the beginning (i.e., t=0). As the organism grows over time, the color change of the broth is represented by a line, not necessarily a straight line, moving from the center of the color wheel to the edge. The final color may refer to where the line ends when it hits the edge of the color wheel at the end of whatever period growth is observed. Color trends may also be observed, with the broth starting out as its default color before turning green, coinciding with the onset of pigment production, the idea being that patterns over time are useful. Statistics and computer / mathematical aided amplification of color / color changes imperceptible to humans are used to assign colors at each point in time.

[0098] According to one or more embodiments, computer vision techniques that enable AST directly from blood in less than one hour without the need for culture are enhanced to achieve rapid growth detection and identification. Computer vision platform technology simplifies engineering, enabling all three steps in a single, concise solution, thereby enabling an affordable end product in the fight against bloodstream infections. The ability to rapidly detect and identify bacteria provides an important solution to the pharmaceutical industry as an improved method for determining bioburden.

[0099] According to one or more embodiments, analysis of unstained images of bacterial species, e.g., collected using microscopy, involves three steps: a cleanup step, a segmentation step, and a fingerprinting step. During image cleanup, one or more machine learning algorithms are used to enhance the quality of the collected images, such as by removing or smoothing numerical noise in the collected pixels. In the segmentation step, individual bacterial cells in the images are identified using image processing that upsamples each pixel to reduce the contribution of debris in each image, effectively segmenting each identified bacterial cell from the rest of the image. "Segmentation" is a term in image processing that refers to isolating pixels corresponding to an object from all other pixels in the image (i.e., from the background and all other objects). Then, in the fingerprinting step, quantitative measurements of cell size, shape, intrinsic color, and sequence are performed on each segmented bacterial cell to identify the bacterial species. device

[0100] According to one aspect, a device for bacterial growth detection and antimicrobial susceptibility testing (AST) is provided. An embodiment of the device is shown in FIGS. 2A-2C. With reference to FIGS. 2A-2C, the device includes a sample entry port 1, a user interface 2, and a housing 3. A door 4 is incorporated into the housing 3, which opens to allow a user or operator to insert a standard well plate into the device. While the sample entry port 1 is designed to accept a standard phlebotomy sample bottle, other types of sample entry ports, such as syringes, are within the scope of this disclosure. As shown, the user interface 2 is a touchscreen, but it may also be any type of suitable display output, such as a non-capacitive LCD / LED screen with dedicated controls, e.g., a keypad, keyboard, or mouse, a cathode ray tube (CRT) screen, or any other suitable display. Alternatively, the user interface may be an external display that connects to the device using a standard display connection, e.g., Universal Serial Bus (USB), DVI, HDMI, VGA, DisplayPort, or any other suitable display standard. If the user interface is external, the device may include connections to external controls such as a keyboard and / or mouse.

[0101] Continuing with reference to FIGS. 2A-2C, the device includes a gantry 5, a fluid dispensing head 6, and a pump 7. In some embodiments, the device may also include fluid reservoirs for reagents and / or waste reservoirs. The gantry 5 generally includes a frame adapted to act as a track along which the fluid dispensing head 6 moves in a Cartesian plane. The gantry 5 also includes a motor used to actuate the fluid dispensing head 6 along its axis. The fluid dispensing head includes a fluid tubing fluidly connected to the pump and a movable dispensing needle. The tubing and movable dispensing needle may be any tubing and needle suitable for use in studying bacterial species, types of which are known in the art. Typically, the tubing is non-stick and non-toxic to reduce clogging of the fluid dispensing head 6 and ensure the viability of the bacterial species during various fluid transfers. A camera configured with appropriate optics to enable imaging of the bacterial species in each well of the sample plate is operably coupled to the fluid dispensing head 6. During operation, the gantry 5 moves the associated fluid dispensing head 6 and its camera to dispense fluid into each well of the sample plate and image each well. A pump 7 directs fluid from the sample inlet 1 through tubing 9 to the fluid dispensing head 6 for distribution to each well. The pump 7 can aspirate and remove fluid from each well so that the fluid can be collected for further analysis. The pump 7 can be any suitable pump for moving biological fluids. As shown in Figures 2A-2C, the pump 7 is a peristaltic pump, but this is merely an example; any suitable pump can be used. In a further embodiment, the device includes a heater to maintain samples within the device at a temperature conducive to analysis. The heater can also be used for sample preparation, such as by drying solvents or fixing samples to microscope slides. The heater can be located below the sample plate or between the sample plate and the optics, but can be located anywhere within the device.

[0102] 2A-2C, once door 4 is opened, well plate 8 can be inserted into plate holder 12 located within the space in the lower portion of housing 3. Once inserted, well plate 8 is positioned beneath light source 11 and associated fluid dispensing head 6 and its camera. Below well plate 8 and plate holder 12 is detector array 13, which captures light passing through sample plate 8 when illuminated by light source 11. Detector array 13 can be any suitable photodetector, such as a charge-coupled device (CCD), a complementary metal-oxide semiconductor (CMOS) detector, and a photodiode.

[0103] The device includes a controller 10 for performing one or more operations within the device. The controller 10 includes a processor for executing instructions, a non-transitory computer-readable medium for storing instructions executed by the processor 10 and for storing collected images and processed results, and a communication module for providing connectivity to the Internet for data movement and communication within the environment in which the device is located, such as a hospital, clinic, research facility, or industrial workplace. The controller 10 may be implemented using one or more computer systems. The computer system may be, for example, a general-purpose computer based on an Intel CORE® type processor, an Intel XEON® type processor, an Intel CELERON® type processor, an AMD FX type processor, an AMD RYZEN® type processor, an AMD EPYC® type processor, an AMD R-series or G-series processor, or any other type of processor or combination thereof. Alternatively, the computer system may include a programmable logic controller (PLC), specially programmed, application-specific hardware, such as an application-specific integrated circuit (ASIC), or a controller for an analytical system. In some embodiments, the controller C may be operatively connected or connectable to a user interface constructed and arranged to allow a user or operator to view relevant operating parameters of the systems disclosed herein, adjust said operating parameters, and / or shut down operation of the system as needed. The user interface may include a graphical user interface (GUI) including a display configured to be interacted with by a user or service provider and to output status information for the system.

[0104] The controller 10 may include one or more processors typically connected to one or more memory devices, which may include, for example, any one or more of disk drive memory, flash memory devices, RAM memory devices, or other devices for storing data. The one or more memory devices may be used to store programs and data during operation of the odor control system and / or control subsystem. For example, the memory devices may be used to store historical data, operational data, etc. Programming code implementing embodiments of the present invention, i.e., software including the deep learning algorithms used in the various steps of image processing and statistical fitting disclosed herein, may be stored on a computer-readable and / or writable non-volatile recording medium and then typically copied to one or more memory devices, where it may then be executed by one or more processors. Such programming code may be written in any of a number of programming languages, such as ladder logic, Python, Java, Visual Basic, C, C#, or C++, Fortran, Pascal, Eiffel, Basic, COBOL, or various combinations thereof.

[0105] The communications module of the controller 10 may include a wired communications connection via an industry standard connection such as a broadband Internet connection, e.g., a local area network (LAN) or wide area network (WAN) using USB, RS-232, RJ-11, RJ-45, Ethernet, or another wired standard. Alternatively or additionally, the communications module may include wireless connectivity via a wireless transmission standard, e.g., Wi-Fi, BLUETOOTH®, 5G NR FR2, LTE Cat 1, LTE Cat M1, or Cat NB1 standard. In some embodiments, the controller 10 has a communications module that includes both wired and wireless communications features. kit

[0106] The present disclosure further provides kits including a device described herein and one or more sample plates, e.g., well plates, e.g., 384- or 1546-well plates. The sample plate provided by the kit may be divided to provide a first portion of the plate used for growth detection and a second portion used for AST. For example, for the wells in the second portion of the plate, one or more of these wells may be pre-filled or pre-loaded with a block or volume of antibiotic, e.g., a lyophilized pellet of antibiotic. If the wells in the second portion of the sample plate contain antibiotics for AST, the antibiotic may be formulated as a serial dilution scheme, i.e., increasing or decreasing antibiotic concentrations at regular volume intervals. In some embodiments, the sample plate included as part of the kit disclosed herein may include an area of ​​the plate, i.e., a portion of the well, for performing the sample smears disclosed herein.

[0107] The kits disclosed herein may include one or more reagents useful for bacterial species growth detection, identification, or AST. In some embodiments, the kits disclosed herein may include a growth medium source or a detection amplification factor. In some embodiments, the kits disclosed herein may include a Gram stain dye. These reagents may be included with the kit in any suitable packaging. The kits disclosed herein may also include one or more additional components, such as sample bottles, e.g., vacutainers or other reagent-specific blood collection tubes. method

[0108] According to one embodiment, a method for detecting growth of a microbial species is disclosed. The method for detecting growth of a microbial species includes providing a sample suspected of containing the microbial species. The method for detecting growth of a microbial species includes acquiring a time series of images of the sample. The method for detecting growth of a microbial species further includes detecting growth of the microbial species by a time-dependent change in at least one of light and color over the time series of images.

[0109] In some embodiments, detecting growth includes detecting growth using an RGB-sensitive imaging sensor in the device. Growth of the microbial species is interpreted as a statistically significant change from a baseline in at least one of light and color. For example, a decrease in light and / or a change in color over a time series of images is interpreted as growth of the microbial species.

[0110] In some embodiments, the method of detecting growth of a microbial species includes acquiring a time series of images of each well of a sample plate and detecting growth in each well. In some embodiments, detecting growth further includes comparing one or more of the time series of images between wells, between a well and a control, or between a well and a reference time series. In some embodiments, the method of detecting growth of a microbial species includes comparing data across multiple wells to increase the sensitivity of growth detection.

[0111] Using the methods disclosed herein, growth detection of microbial species is achieved in a duration of less than 5 hours, e.g., less than 5 hours, less than 4 hours, less than 3 hours, less than 2 hours, or less than 1 hour. In some embodiments, several wells of a sample plate with observable growth are used to determine the number of replication-competent microbial cells. Without being bound by any particular theory, the number of wells divided by the total volume of sample placed on the plate generally equals the number of replication-competent cells, i.e., colony-forming units (CFU) per unit volume, in mL volume.

[0112] In some embodiments, the method for detecting the growth of microbial species further comprises quantifying the bioburden of the sample based on the growth detection. As used herein, "bioburden" refers to the amount and type of natural bacterial and fungal flora present on or within a device, substrate, or chemical. Bioburden plays a major role in determining what is needed to achieve sterility in a given environment. In some embodiments, the method for detecting the growth of microbial species further comprises one or more remedial actions based on the quantification of bioburden.

[0113] In some embodiments, the method of detecting growth of a microbial species further comprises identifying and / or subjecting the microbial species to AST upon detecting growth.

[0114] In certain embodiments, the method for detecting growth of microbial species further comprises detecting a polymicrobial infection by comparing a time series of images between wells of a sample plate, wherein the presence of a polymicrobial infection is assessed based on differences in color or doubling time across wells of a sample plate, e.g., a well plate, e.g., a microwell plate.

[0115] In some embodiments, the method of detecting growth of a microbial species further comprises processing the time series of images to enhance their quality to facilitate the earliest possible detection of growth. As disclosed herein, detection of growth of a microbial species is achieved in a duration of less than 5 hours, e.g., less than 5 hours, less than 4 hours, less than 3 hours, less than 2 hours, or less than 1 hour.

[0116] In a further embodiment, the method of detecting growth of a microbial species comprises adding a growth medium or detection amplification agent to a sample, which may be any suitable sample, such as a whole blood sample, a blood component sample, other bodily fluid, a product related to a pharmaceutical manufacturing component or a finished end product, a filtration membrane, or any other sample from which a microbial species may be harvested and analyzed.

[0117] In some embodiments, the method is characterized by digital proliferation detection.

[0118] According to one aspect, a method for identifying a microbial species is disclosed. The method for identifying a microbial species includes imaging a sample containing a plurality of microorganisms of a microbial species to obtain a series of multi-microbial images. The method for identifying a microbial species includes segmenting the multi-microbial images. The method for identifying a microbial species further includes measuring parameters of each segmented microorganism to obtain a multi-dimensional distribution of the measured parameters. The method for identifying a microbial species further includes classifying the microbial species based on the multi-dimensional distribution of the measured parameters.

[0119] In some embodiments, the measured parameters relate to one or more of the size, shape, intrinsic color, arrangement, and other morphological characteristics of the microbial species. For example, at least one of the one or more measured parameters is selected from the group consisting of width, length, internal density, membrane thickness, color heterogeneity, color intensity, curvature, tapering, aspect ratio, and concavity. In certain embodiments, the method for identifying microbial species includes intrinsic color data associated with the microbial species without staining, i.e., without Gram staining.

[0120] In some embodiments, classification of microbial species involves the use of machine learning, e.g., a machine learning algorithm, trained on at least one of the following modalities: (i) unstained slides imaged in direct light, (ii) unstained slides imaged in indirect light, e.g., dark-field microscopy, and (iii) pre- and post-stain image pairings. In certain embodiments, classifying the microbial species involves a probability distribution. In certain embodiments, classifying is performed using a hierarchical approach. In certain embodiments, classification is performed using majority voting, decision trees, or the relative entropy of the observed tally and a reference distribution of known microbial species.

[0121] In some embodiments, a method for identifying a microbial species includes distinguishing between Gram-positive and Gram-negative bacteria with a first confidence score, and using this first distinction, each bacterial species is then identified with a second confidence score.

[0122] In further embodiments, the method for identifying microbial species includes filtering the images, for example, using a colored filter. For example, a blue filter can be applied to all collected images to provide image sharpening. In some embodiments, the method for identifying microbial species includes collecting images of each sample at multiple focal lengths of the optical system. As a non-limiting example, images of each sample are taken at a series of small or fine focal length intervals. In some embodiments, the fine focal length interval may be between about 0.1 μm and 2 μm, e.g., between about 0.1 μm and 2 μm, between about 0.2 μm and 1.8 μm, between about 0.3 μm and 1.6 μm, between about 0.4 μm and 1.4 μm, between about 0.5 μm and 1.2 μm, or between about 0.6 μm and 1 μm, e.g., between about 0.1 μm, about 0.2 μm, about 0.3 μm, about 0.4 μm, about 0.5 μm, about 0.6 μm, about 0.7 μm, about 0.8 μm, about 0.9 μm, about 1 μm, about 1.1 μm, about 1.2 μm, about 1.3 μm, about 1.4 μm, about 1.5 μm, about 1.6 μm, about 1.7 μm, about 1.8 μm, about 1.9 μm, or about 2 μm. In certain embodiments, the focal length interval is 0.5 μm. In some embodiments, the method of identifying a microbial species comprises acquiring multiple images at each focal length and combining the images.

[0123] In further embodiments, the method for identifying microbial species includes selecting an image having a maximum value of one or more quality metrics for measurement. The quality of the selected image can be improved by removing or smoothing numerical noise before segmentation. In some embodiments, the method for identifying microbial species includes performing dimensionality reduction, such as PCA or UMAP, on the multidimensional distribution of the measured parameters.

[0124] Using the identification methods disclosed herein, measuring 100 randomly selected segmented microorganisms is sufficient to identify a microbial species with approximately 93-97% confidence.

[0125] In some embodiments, the microbial species sample is from a growth detection study, which may be performed using a device disclosed herein or the same or a different sample plate.

[0126] In a further embodiment, the method of identifying a microbial species includes identifying a second microbial species.

[0127] According to one embodiment, a method for performing growth detection, identification, and antimicrobial susceptibility testing (AST) on a microbial species is disclosed. The disclosed method is configured to perform all three functions in less than about 8 hours.

[0128] In some embodiments, one or both of the detection of growth of a microbial species and the identification of a microbial species are performed using the methods disclosed herein.

[0129] In some embodiments, the AST is based on a differential proliferation detection method, such as those disclosed herein. For example, the AST includes dilution time modeling (DTM). Using the methods disclosed herein, the AST achieves categorical agreement with the reference method in less than 1 hour, for example, less than 1 hour, less than 55 minutes, less than 50 minutes, less than 45 minutes, less than 40 minutes, less than 35 minutes, less than 30 minutes, less than 25 minutes, less than 20 minutes, less than 15 minutes, less than 10 minutes, or less than 5 minutes. In some embodiments, the reference method for AST comparison includes broth microdilution.

[0130] In some embodiments, AST is performed with an antibiotic selected from cefepime, meropenem, ciprofloxacin, and gentamicin. Any antibiotic may be used in the AST disclosed herein, and the selection of antibiotic correlates with the bacterial identification performed prior to AST. The disclosed method is in no way limited by the selection of antibiotic used for AST.

[0131] In some embodiments, the AST results are reported to an operator, a laboratory information system, and / or an electronic medical record, e.g., transmitted by a device disclosed herein, or displayed using a user interface on a device disclosed herein.

[0132] In some embodiments, once a microorganism has been identified at the category level, AST may be performed before the microorganism is identified at the species level. In some embodiments, growth detection, identification, and AST are each performed on a single device, such as a device disclosed herein.

[0133] According to certain aspects, a quality control method for a pharmaceutical manufacturing process is disclosed, which includes subjecting a sample containing a pharmaceutical ingredient or a finished end product to a method for detecting the growth of microbial species to assess its bioburden, for example, using any of the devices or methods disclosed herein.

[0134] In some embodiments, a quality control method for a pharmaceutical manufacturing process includes approving or rejecting a pharmaceutical ingredient or finished end product based on a comparison of the assessed bioburden to a threshold value. [Example]

[0135] The function and advantages of these and other embodiments can be better understood from the following examples, which are intended to be illustrative in nature and are not to be construed as limiting the scope of the invention. Example 1

[0136] This example demonstrates the growth detection component of an embodiment of the disclosed system. In particular, it demonstrates growth detection directly from whole blood in less than five hours. Figure 3 shows a typical experimental workflow. In the workflow shown, human blood was spiked with E. coli at a dose typical of a sepsis patient (e.g., 1–10 CFU / mL), mixed with growth medium, and automatically dispensed into wells of a microwell plate. The number of wells was large, typically 384 or 1,536, so that each well contained purely single bacteria via Poisson filling. This demonstrated that growth detection was digital. As a result, the number of wells with growth was proportional to the number of replicative bacterial (or other) cells (i.e., CFU). Thus, this system enabled quantification of bioburden. Each well was monitored for growth using an automated RGB high-sensitivity imaging sensor. The sensor data from each well was processed to generate a time series. Each time series was processed, and statistically significant changes from baseline were interpreted as growth. Processing included, but was not limited to, accounting for background changes in blood over time and excluding extreme values. Figures 4A-4B show results from two experiments demonstrating growth in less than 5 hours and less than 4 hours, respectively. In certain embodiments, data from multiple wells can be compared to increase sensitivity. Differences in time series between wells (e.g., color or doubling time) can be used to indicate a suspected polymicrobial infection, which can be further investigated by identification. Example 2

[0137] This example demonstrates bacterial identification directly from unstained smears. "Unstained," as used herein, refers to material imaged without the addition of a staining agent, such as Gram stain, or other reagent-mediated enhancement procedures. Bacteria were grown as described in Example 1. Smears were made from material recovered from wells after growth detection as described in Example 1. Making the smears consisted of placing the liquid material on a glass slide, spreading the material across the slide, and then heating and drying the material to fix it to the slide.

[0138] The slides were then imaged at 40x magnification. A blue filter was used to sharpen the images. Indeed, the filter altered the color profile in a predictable manner while preserving useful color information. Images were taken at multiple focal lengths and at a small / fine focal length interval of 0.5 μm. This distance was chosen because it is smaller than the diameter of a bacterial cell, ensuring that at least one focal length would result in a focused image. Multiple images were automatically taken at each focal length. These images were then combined to virtually eliminate imaging noise. Multiple images were taken without the slide, and additional images were taken with the slide but without illumination. These were used to generate "clean" combined images from each focal length, controlling for potential background artifacts. Comparison of the clean combined images from different focal lengths allowed the most focused and clean combined image to be selected. Potential debris, such as bacterial cells and remnants of lysed red blood cells, collectively referred to here as "objects," were clearly visible in these images despite the absence of staining, as shown in Figures 5A-5C. At this point, further cleanup was carried out using super-resolution techniques to sharpen the image, for example using deep learning.

[0139] Each object in the microscope field was segmented. Figure 5A shows the segmentation on cells from different bacterial species and groups. There were typically hundreds of objects per field, the majority of which were bacterial cells. Because the wells from which the smears were made began with a single bacterial cell, the cells were known to belong to the same bacterial species. Therefore, it was unlikely that the cells segmented from a given smear could have come from multiple different microbial strains or species.

[0140] Next, some specific measurements were made for each segmented object. In this example, these measurements were made by counting the number of pixels within the object and, knowing the dimensions of the pixel, converting it into standard area units, e.g., μm 2The measurements included the size of the object, calculated by converting it into units of length, e.g., μm, the circumference of each object, also calculated in units of length, e.g., μm; the average color of the object, calculated as both a red-green-blue value and a hue-brightness-saturation value, each value between 0 and 255; and the number of neighbors; each object was corrected for the total number of objects in the field of view, since a denser field of view would simply result in more neighbors by chance. The number of neighbors and related metrics provided a quantitative method of assessing bacterial cell arrangement, useful for identification. For example, Staphylococci clustered, meaning that each Staphylococcus cell had many neighbors, while Enterococci did not cluster, meaning that each Enterococcus cell had fewer neighbors. Thus, the measurements associated each object with an ordered series of numbers. In this example, these associations corresponded to size, circumference, color, and relative arrangement. Mathematically, the technical term for this ordered series of numbers is a "vector," and the measurement step described in this paragraph can be said to assign a vector to each object or define a vector for each object, with the vector describing the object. Each object received its own vector, and thus the set of vectors for all objects segmented from the smear was a quantitative description of the smear.

[0141] In this example, as shown in Figures 5A-5C, smears were prepared, segmented, and measured as described herein for each of five bacterial species or groups: Staphylococcus aureus, Pseudomonas aeruginosa, Acinetobacter baumannii, Enterobacterales (the group containing E. coli), and Enterococci (the group containing Enterococcus faecium and Enterococcus faecalis). A set of vectors for each species or group defined that species or group. Each vector can be thought of as a point in space. When there were three measurements, this was a point in three-dimensional space, i.e., x, y, and z on a Cartesian plane. The first measurement provided the distance in the x direction, the second provided the distance in the y direction, and the third provided the distance in the z direction. The set of vectors thus corresponded to a "point cloud" in this space. Each point in the point cloud corresponded to a measurement from a single bacterial cell. When there were more than three measurements, as in this example, the result was a point cloud in a higher-dimensional space. To aid in visualization of high-dimensional spaces, mathematical tools known as "dimensionality reduction" techniques (e.g., principal component analysis (PCA) and uniform manifold approximation and projection (UMAP)) were used to allow the higher dimensions to be "flattened" down to two dimensions while preserving the overall shape and relative arrangement of the point clouds. Figures 5A-5C further illustrate the results of dimensionality reduction using the UMAP technique for the five species and groups of organisms mentioned above. The figures show that the corresponding point clouds had characteristic arrangements. These arrangements were the "fingerprints" of each species. For example, the S. aureus fingerprint was internalized, meaning that S. aureus cells were relatively homogeneous in their measurements. In contrast, the Enterococcus fingerprint had two parts: a larger part that was internalized and a smaller part that overlapped with the S. aureus point cloud. This arrangement indicated that the majority of Enterococcus cells appeared different from most S. aureus cells, but a few appeared identical to S. aureus.To demonstrate that a smear contained Enterococcus and not S. aureus, enough cells had to be measured so that some of them appeared in a larger portion of the S. aureus point cloud. Figures 5A-5C show that 100 cells were sufficient to identify an unknown with 93-97% confidence. In each panel of this figure, vectors of 100 cells were randomly selected from the smear and compared to the fingerprints in Figure 5B. The probability (×100) that they matched each fingerprint is shown as the height of the bar. In each case, a correct identification was made and is indicated by the highest bar, while an inaccurate identification was not nearby and is indicated by the next highest bar. Thus, this example demonstrated high-confidence identification from unstained cells in the context of the integrated three-part (growth detection / ID / AST) invention described in this disclosure. Example 3

[0142] This example demonstrated category agreement between an embodiment of the disclosed system and a reference method for AST in less than one hour. "Category agreement" is a term of art in AST that describes whether an organism is susceptible or resistant according to each of two methods, typically a reference method and an investigational method. The reference method is broth microdilution and is derived from CLSI (M100).

[0143] Figures 6A-6B show results from a typical experiment. Each experiment included a standard inoculum of bacteria grown at a concentration similar to that recovered from the growth detection step, which began with whole blood. The inoculum was automatically dispensed into wells of a microwell plate. The wells contained two-fold dilutions of standard antibiotics such as cefepime, meropenem, ciprofloxacin, and gentamicin (for Gram-negative bacteria). The large number of wells allowed for multiple replicates per condition, providing robustness against random shedding or growth fluctuations.

[0144] In each experiment, the growth and color of each well were monitored for growth using an automated imager. A series of images for each "bug-drug" combination was combined to create a unified picture of bacterial growth for each antibiotic, and MICs were measured to determine any correction factors required to maximize S / I / R categorical agreement with the reference MIC. "Unified picture" refers to dilution-time modeling, or DTM, which combines the Gompertz model of growth over time with the Hill model of growth as a function of antibiotic concentration to generate a comprehensive 3D picture of growth as a function of both time and concentration (Figures 6A-6B). The equation for this modeling is shown in Figure 6C. In the meropenem-susceptible strain of E. coli shown in Figure 6A, a decline in growth with increasing antibiotic concentration was clearly visible at 0.5 μg / mL. The MICs obtained by the disclosed method (red line) were categorically consistent with, and within a single 2-fold dilution of, the MICs of the reference method (blue line) accepted in clinical practice. This decline is observed with statistical confidence at 30 minutes in this example (black line). In contrast, in the cefepime-resistant strain of E. coli shown in Figure 6B, growth remains virtually unchanged as the antibiotic concentration increases until the concentration reaches 16 μg / mL, after which there is a decline, demonstrating the MIC from the method disclosed herein. Again, this is in category agreement with the MIC of the reference method, within a single two-fold dilution. In this example, the decline is observed with statistical confidence at 20 minutes.

[0145] Figure 7 shows a summary of 11 standard antibiotic results for organisms representing five key bacterial groups: Enterococci, Staphylococci, Enterobacterales, Acinetobacter baumannii, and Pseudomonas aeruginosa. These groups contain particularly important ESKAPE organisms, accounting for two-thirds to three-quarters of all bacterial bloodstream infections. Doubling times across multiple strains of Enterococci (including E. faecium and E. faecalis strains), Staphylococcus (including S. aureus), A. baumannii, P. aeruginosa, and Enterobacterales strains (including E. coli and Klebsiella) were typically 20-40 min, with doubling times typically 20-30 min. This is consistent with completing growth, ID, and AST in less than 8 h. Example 4

[0146] This example demonstrates automated fluid handling using a pipette, which was used in the examples disclosed herein for several purposes. These purposes included filling microwells with a mixture of whole blood (or other starting specimen) and growth medium (Example 1), obtaining material from one or more microwells for ID (Example 2), and filling microwells for AST (Example 3). The pipette was fast, filling at approximately 1 well / second, and accurate. The pipette included a movable tip held in place by a head. The tip was connected to tubing, as shown in Figures 8A-8B, through which the associated fluid flowed. Flow was controlled by a peristaltic pump. The pump used a two-lobe design with ball bearings to minimize friction with the tubing. Driving the pump was a stepper motor shaft with an absolute position magnetic encoder. This magnetic encoder provided the angular position of the lobe at any given time, which compensated for any position-related inhomogeneity in the flow, thus allowing for very low standard deviations in the dispensed volume per well. A motorized gantry system moved the head and needle from well to well. Figures 8A-8B show the use of the dispenser in filling either a 384-well or a 1,536-well microwell plate. The inset shows the dispensing of 10 μL of blood plus growth medium. It demonstrates both that the dispenser can handle this material without clogging and that the result is an accurate fill. The bar plots shown in Figure 8C demonstrate the accuracy of the dispenser. Specifically, these bar plots demonstrate excellent performance, with a standard deviation of less than 0.2 μL, corresponding to a standard error of 1.6%, for dispense volumes as small as 10 μL. Example 5

[0147] Training a deep neural network on a combination of Gram stain images and broth culture colors suggests that computer vision can be used to identify bacteria quickly enough to interpret AST results. Gram stain stains bacteria pink (Gram-negative) or purple (Gram-positive), making their shape, size, and relative arrangement visible under a microscope. Color is an attribute of macroscopic colonies on agar plates, and it has been demonstrated that color can be detected from broth cultures using the device disclosed herein in less than an hour. Gram stain and color aid in identification.

[0148] Deep learning, a mainstay of computer vision, has shown great promise for bacterial ID. Deep learning-based classifiers learn about bacteria by studying many images and finding unique patterns in each of a set of classes, which may be species (e.g., Enterococcus faecium), genus (Enterococci), or some other type of category (Gram-positive cocci). An existing dataset was released containing 660 Gram-stained images from pure colonies of 24 Gram-positive and 9 Gram-negative bacteria, including E. faecium, E. faecalis, S. aureus, A. baumannii, P. aeruginosa, and E. coli. Existing deep learning-based classifiers reached >99% accuracy on this dataset. Deep learning performs well on Gram stains from challenging real-world blood cultures, i.e., samples with background detritus, color and contrast variability, and crystallization artifacts, and recently achieved 94.9% accuracy in classification of clustered Gram-positive cocci (typical of Staphylococci), paired and chained Gram-positive cocci (Enterococci), and Gram-negative bacilli (A. baumannii, P. aeruginosa, and Enterobacterales).

[0149] A hierarchical approach that first classifies bacteria as Gram-positive or Gram-negative and then performs ID within those groups was thought to best translate to the real world. Deep networks are powerful, but can be fooled, and their decisions can be difficult to understand. To make them more reliable and interpretable, the compensation task was separated into stepwise tasks: first classification by Gram-stain appearance, then by identification. Each task had its own classifier and confidence score, making the decisions more accurate and interpretable.

[0150] It was further thought that identifying bacteria singly or in small groups and then aggregating the results would improve accuracy and interpretability. Classifiers often benefit from majority voting when the final decision depends on partial decisions, as is common in medical images. For identification, each Gram-stained image was divided into small (80 × 80-pixel) sections, each containing only approximately 1–20 bacterial cells (more than one was allowed to capture differences such as paired-and-chain versus cluster, parade arrangement, etc.). Each section was classified, and the winner of the vote was assigned as the final identification, as shown in Figure 9.

[0151] Gram stain heterogeneity is informative in clinical testing; for example, Gram-indefinite bacilli suggest Clostridia. Because each well started with a single cell, the experiments disclosed in this example ensured that heterogeneity was always characteristic of a single organism and not due to the presence of mixed organisms. When the majority rule proved insufficient, information within the aggregates was used to create more elaborate rules, including, for example, decision trees and / or the relative entropy of the observed aggregate and a reference distribution of known bacteria.

[0152] It was further believed that using the color of unstained bacteria would aid in identification, especially when organisms of different colors have similar Gram-stained appearances, as shown in the box inset of Figure 9. For example, P. aeruginosa and E. coli were essentially indistinguishable by eye on Gram staining. However, the green, blue, or purple colonies of P. aeruginosa were distinguishable from the clear to pale yellow colonies of E. coli. The device disclosed herein, in less than an hour, was shown to detect not only subtle changes in growth but also subtle changes in color that matched the known colors of the organisms shown in the center panel of Figure 9. These included the clear to yellow colors of E. coli and S. aureus and the purple to green color of P. aeruginosa due to pyocyanin and / or pyoverdin production. While not all strains were colored, the color provided discriminatory information for the identification of those that were colored.

[0153] In this example, a deep learning network was constructed and trained. Using hierarchical classification, color, and majority voting, the deep learning network demonstrated 100% Gram stain accuracy and 95% identification accuracy, including the set of internally stitched Gram stain images shown in Figure 9. A fixed objective 40x magnification microscope was fabricated using a lens that could be precisely positioned in front of the device's gantry-controlled camera in less than an hour. The camera was able to transition from viewing a well plate to viewing a Gram stain slide.

[0154] Each image was divided into small sections containing 1-10 bacterial cells each. The deep learning network disclosed herein demonstrated improved performance with color, as accuracy for E. coli and P. aeruginosa increased from 89% and 99% to x% and y%, respectively. Each classification took less than 1 second. The vote tally appeared to carry information about whether the ID was likely to be correct, suggesting that more refined rules would further improve performance. These results demonstrate the strength of the stepwise approach disclosed herein and support the use of machine vision and deep learning networks for bacterial identification as part of a single-shift S2A. Hypothetical Example 1

[0155] In one hypothetical embodiment, a device is designed to perform growth detection, ID, and AST on whole blood for the purpose of direct AST from whole blood in the setting of sepsis, as shown in Figures 2A-2C. Briefly, whole blood is collected into a container containing a proprietary or other bacterial growth medium. The container is scanned with a barcode scanner to collect and verify patient and sample information. The sample is inverted and inserted into the device, which is temperature-controlled and, optionally, an airtight enclosure. A disposable microwell plate containing an area for smearing is inserted through the front door of the device. The operator's hands-on time to perform these steps is approximately two minutes. Upon insertion of the container and plate, a needle or other collection device enters the container, and a peristaltic pump draws material through tubing. The material is dispensed into the microwells of the plate by a peristaltic pump-powered dispenser operated by a moving gantry controlled by a microcontroller. Some wells are left empty as controls. This process takes several minutes.

[0156] Step 1: Growth Detection: The plate is placed by a plate holder between an RGB illuminator and a highly sensitive kHz-MHz detector array containing a detector monitoring each well. The plate is then monitored for growth. The illuminator cycles between red, green, and blue light at a frequency of approximately single-digit Hz, allowing the detector to take many readings per cycle. These readings constitute a time series for each well. Program execution by the device's microprocessor integrates the measurements into clean color information. Generally, a decrease in light and / or a change in color is interpreted as growth. Comparison of time series between wells, between wells and controls, and between wells and reference time series allows for the subtraction of any background time-varying patterns, i.e., changes not related to growth. If necessary, clustering time series across wells can reveal two patterns that can be interpreted as growth and no growth, respectively. Such cross-well comparisons can also be used to increase the sensitivity of growth detection beyond that possible from a single well. In this context, increased sensitivity means detecting growth earlier by considering multiple wells than would be possible by considering each well separately. In this example, growth detection is expected to take less than or about 5 hours for the majority of clinically important bacteria. The touchscreen of the device allows the operator to monitor the process if or as desired.

[0157] Step 2: Identification: If growth is detected, the dispenser acts as an aspirator by reversing the pump direction, collecting positive material from one or more wells and dispensing it into the smear area, which is then dried by the enclosure's internal heating components. A magnification camera on a moving head is then positioned above the smear, and the smear is repeatedly imaged as directed by a program on the microprocessor. The head is moved up and down through a range that includes the focal plane shown in FIG. 10A. The microprocessor then processes the image, e.g., by deep learning-enabled super-resolution, with additional cleanup steps as needed, e.g., as described in Example 2 and shown in FIG. 10B. The object is then segmented, as shown in FIG. 10C, measured, and fingerprinted, as shown in FIG. 10D, and the fingerprint is compared to a reference fingerprint to reveal the identification, as shown in FIG. 10E. Note that the comparison shown in Figure 10E includes a comparison with fingerprints of blood debris (red blood cell ghosts, platelets, white blood cell nuclei, etc.) to avoid false identification. Identification is expected to take ≤15 minutes. The information is then reported to the operator via the device's user interface, which is depicted as a touchscreen. Optionally, a second identification may be performed if the time series shows evidence of a second organism in a different set of wells. The operator is notified of the presence of the second organism. Each sample may be returned to the operator at any time through the device's door.

[0158] Step 3: AST: Finally, using the dispenser disclosed herein, material from positive wells is aspirated, mixed with fresh medium if necessary, and dispensed into antibiotic-containing wells in different portions of the plate. The plate contains enough wells for replicates of 20 or more different antibiotics and / or antibiotic combinations, including combinations with, for example, beta-lactamase inhibitors, at multiple two-fold dilutions. As used herein, a "two-fold dilution" refers to a series of concentrations, such as 4 μg / mL, 2 μg / mL, and 1 μg / mL, with each subsequent dilution containing half the antibiotic concentration of the previous dilution. The plate is then repositioned between the device's illuminator and detector and imaged as described above. Thus, what is observed is differential growth detection, i.e., differences in growth in the presence of different antibiotics at different concentrations. Integrated photographs of growth are obtained as described in Example 3, MICs are determined for each antibiotic or antibiotic combination, and the results are reported to the operator via the device's user interface. In a hospital environment, results are reported via communication hardware to a laboratory information system (LIS) and also via that system to the hospital's electronic medical record (EMR), providing usable results to healthcare providers. Hypothetical Example 2 overview

[0159] This hypothetical example describes the planned development and testing of a device that performs growth detection, bacterial identification, and AST in a single hospital shift. Specific objectives are to develop each of these three components to specified target performance thresholds (Objectives 1–3), then combine them and test the resulting device (Objective 4). Objective 1. Demonstrate rapid proliferation detection with a photodiode-based device

[0160] Device. The initial prototype device disclosed herein is upgraded by moving remaining components inward, simplifying the circuitry to reduce noise, and replacing the white LED with red, green, and blue LEDs to measure color and improve sensitivity. Preliminary studies demonstrated that the prototype can be quickly assembled and upgraded, and rapid on- and off-LED switching can be operated. These preliminary studies also demonstrated the usefulness of color for bacterial identification. Growth detection experiments for Objective 1 proceed as in Example 1 using blood, but only 192 wells are filled (100 μL / well), with rapid cycling of the red, green, and blue LEDs to allow the photodiodes to record color changes. In addition to being useful for ID, color likely improves sensitivity, as color changes can precede intensity changes by as much as the doubling time observed in growth experiments using the device disclosed herein, which took less than an hour.

[0161] Microwell plate format. It is envisioned that the final device will contain a single 384 or 1,536 microwell plate, with half of the wells dedicated to proliferation detection and the other half dedicated to AST. Thus, our proliferation and AST experiments each use only half of the plate.

[0162] Experiments. Growth detection experiments are performed in triplicate with 100 extensively characterized strains from the BEI / ATCC and FDA-CDC Antimicrobial Resistance Bank (AR), the same strains we use. These include 20 strains each from five groups: Enterococci, Staphylococcus, Enterobacterales, A. baumannii, and P. aeruginosa, including methicillin-resistant S. aureus (MRSA; both linezolid-susceptible and -resistant strains), vancomycin-resistant Enterococci (VRE), extended-spectrum beta-lactamase (ESBL)-producing E. coli, carbapenem-resistant Enterobacterales (CRE), carbapenem-resistant A. baumannii (CRAB), and carbapenem-resistant P. aeruginosa (CRPA). Enterobacterales includes Enterobacter, Serratia, Citrobacter, Morganella, and Proteus, all common causes of severe BSI.

[0163] Performance Threshold. The minimum performance threshold is the ability to detect 14 doublings starting from a single bacterium. This is a change from 214 to approximately 16,000 bacteria / well, which translates to 4 hours and 40 minutes for a strain with a 20 minute doubling time. The sensitivity of the photodiode improves growth detection by reducing noise. Each step taken has been shown to reduce noise by 5- to 2-fold. A further 2-3x improvement from layout and circuit upgrades, as well as another 2x improvement from adding color, equates to two two-fold improvements, and is expected to be approximately 2x better than the improvement required to reach the above threshold.

[0164] Statistical rigor. Each experiment is performed in triplicate.

[0165] Pitfalls and Alternatives. Some bacteria divide slowly, and some blood contains growth inhibitors. Just as there are long tails beyond the median growth detection time (e.g., Cutibacterium), in current practice, long tails beyond our expected 4-6 hour median are expected. However, additional options are available to ensure that most bacterial strains are detected near the median. These include combining data across wells to improve sensitivity, modifying the broth to increase recovery, adding detection amplification factors such as in vivo dyes, and using deeper 1,536-well plates to concentrate the signal based on the general relationship that a similar volume and one-quarter area should result in approximately a 4x increase in sensitivity. Overall, these options are expected to result in greater than or equal to 8x sensitivity, further reducing the time from growth detection, which doubles in 20 minutes. Aim 2. Demonstrate rapid AST with a photodiode device for the same strains as in Aim 1.

[0166] Rationale: The sensitivity of the photodiode device should allow rapid AST even from a few bacteria available after rapid growth detection. 6 A sub-hour AST starting with bacteria / well has been demonstrated as disclosed herein. The sub-hour device disclosed herein achieves a frame rate of 5×10 5 ~1×10 6Changes in fewer bacteria / well were not detectable on a timescale of less than one hour. Growth detection is expected to peak at approximately 50 wells with 16,000 bacteria / well. Setting aside one well for identification (Objective 3) and dividing the remainder into 192 AST wells provides approximately 4,000 bacterial cells per AST well. The photodiode requires 2.3 doublings to detect growth starting from 4,000 cells and another doubling to detect the MIC, for a total of 3.3 doublings or slightly more than one hour for a 20-minute doubling time, for a total S2A of 5–6 hours. An S2A of ≤8 hours is expected to be applicable to most bacterial strains.

[0167] Performance Threshold: The threshold is an AST within 3 hours starting from this small inoculum.

[0168] Primary and Secondary Metrics. Percent categorical (S / I / R) agreement is important for treating patients and is therefore the primary metric. It is calculated in the standard FDA manner: number of agreements divided by total number tested. Secondary metrics include very large, major, and minor error rates, essential agreement (i.e., agreement ± one two-fold dilution), time to MIC, and k μm , MIC, and confidence level (mean ± sd of triplicates).

[0169] Ground truth. Gold standard reference MIC comparison criteria are determined by CLSI Reference Broth Microdilution (CLSI M07) and interpreted categorically according to CLSI guidelines (CLSI M100).

[0170] Testing and Validation: To determine rules for calling MICs, 40 strains are tested in triplicate from each of the five groups (200 total instead of 100 in Objective 1 to test a range of susceptibility patterns). These rules are then validated with an additional 200 strains.

[0171] Experiment. The experiment was as described in the Examples section of this specification, with four modifications. First, the experiment began with a smaller inoculum of approximately 4,000 bacteria per well. Second, the starting material was prepared in cation-adjusted Mueller-Hinton broth with 12% blood to mimic the material recovered from positive wells after growth detection. Third, only half of the wells on the plate were used. Fourth, there were 18 antibiotics instead of 11, with two replicates for each of five 2-fold dilutions, an improvement attributed to the improved sensitivity of the photodiode device. Concentrations are bracketed with the most common MICs across the most common bacteria. Twelve wells served as positive and negative controls (no antibiotics, broth only / empty wells, respectively).

[0172] Antibiotics. The antibiotics are the following CLSI first-line agents: ampicillin, clindamycin, daptomycin, doxycycline, linezolid, oxacillin, and vancomycin (for gram-positives); ampicillin / sulbactam, cefepime, ceftazidime, ceftazidime-avibactam, ceftriaxone, gentamicin, and meropenem (for gram-negatives); and cefazolin, cefoxitin, levofloxacin, and trimethoprim / sulfamethoxazole (both). These are preloaded onto plates and lyophilized.

[0173] Analysis. Using the DTM model disclosed herein, at the earliest time point, k μm Calculate and measure the MIC of each drug. μm It is measured as the concentration just above k. μm = 0.79 μg / mL (dilution between 0.5 and 1 μg / mL), the MIC is 1 μg / mL. μmIf the MIC is outside the tested range, the MIC is recorded as > or < the relevant limit (e.g., >64 μg / mL). Note that on modern platforms, raw MICs may require internal adjustment, e.g., eliminating one dilution for certain antibiotics, to match the reference MIC. If necessary, drug-, group-, and species-specific rules are developed to correct MICs and maximize categorical agreement.

[0174] Statistical rigor. 200 strains in triplicate is consistent with the FDA's 510(k) guidelines for statistical confidence. Bootstrapping was used to μm We obtain confidence intervals for . The time of detection is defined as the earliest time point at which we can detect growth in the positive control wells and see a monotonic dose-dependent trend that is robust to bootstrapping for strains that show dose dependence, or growth equivalent to the positive control across doses for strains with high levels of resistance.

[0175] Pitfalls and Alternatives: The current design detects mixed cultures, but requires a second plate to perform AST on them. In this situation, the original plate is used for AST on the additional organisms in the remaining wells. Objective 3. Demonstrate ID from micrographs and bacterial color.

[0176] Methodology. The plan is to continue developing hierarchical and majority-rule methods. Briefly, photomicrographs at 40x or 100x magnification are divided into small compartments containing bacteria. These are classified as Gram-positive vs. Gram-negative, cocci vs. bacilli vs. coccobacilli, etc. Within these categories, they are further classified using appropriate genus- or species-level categories, such as Enterobacterales vs. P. aeruginosa vs. A. baumannii for Gram-negative bacilli. Results across all compartments are combined by majority rule or similar rule to assign a final ID. Note that while identification to the species level has been demonstrated, AST interpretation requires identification only to the level of the CLSI category (e.g., Enterobacterales), which is often broader than a specific species.

[0177] Model Architecture. The following model architectures are tested: CNN EfficientNet, ConvNex, and RepLKNet, as well as Transformer Swin as part of a neural-based decision tree (NBDT). These CNNs and Transformers are currently the best performers for image classification tasks. NBDTs learn an entire hierarchy of labels at once (Gram-positive, cocci, clusters, yellow) instead of just the final label (S. aureus). They enable end-to-end learning while preserving hierarchy / multiple labels and human interpretability. NBDTs provide results for specific compartmentalization success, e.g., bacteria > degenerates: 99%, Gram-positive > Gram-negative: 99%, cocci > bacilli: 99%, clusters > paired-and-chains: 94%, yellow > white: 99%, and therefore S. aureus: 95%; many 95% S. aureus compartments → overall >99% S. aureus. Adding a layer to combine votes is also investigated.

[0178] Training and Validation Sets. In deep learning, a large, heterogeneous training set is key to generalizability. 200 strains are imaged in triplicate experiments as outlined in Objectives 1 and 2. Images are collected at 40x and 100x image fields. Each imaged field is expected to contain tens to hundreds of bacteria. To further heterogeneity, Gram stains are prepared in replicates, both manually and using our device system (Objective 4), and imaged with a microscope and a camera equipped with a photodiode as disclosed herein. Thus, this dataset contains heterogeneity across species, strains, cameras, magnifications, stains, and fields of view. Each Gram stain is manually reviewed for quality control and then labeled according to various strain characteristics. Standardized colors are added as HSV or RGB triples from growth detection and AST experiments. The dataset is divided into training and validation sets for each strain, with no strain appearing in both sets. Additionally, images from public datasets are added to the model validation set to further assess generalizability. In both training and validation, the partitions are generated automatically.

[0179] On-device Gram staining. An on-device Gram staining module is constructed that utilizes the fluid handling system disclosed herein, a heating element (to dry slides) disclosed herein, an existing camera with a photodiode detector (with autofocus), and an existing magnifying lens disclosed herein. Training is performed using a cloud-based service, such as Amazon Web Services, but classification (inference) is not computationally intensive and therefore occurs on the device's internal computer.

[0180] Statistical rigor. The experiment is expected to generate a large and diverse training set and a separate validation set. To test for robustness, each fit will be repeated 10-20 times and the Jaccard coefficient will be used to compare the stability of results between fits. Model performance will be measured by accuracy (correctly classified samples divided by total samples), both overall and for each category.

[0181] Performance Threshold. One goal of this experiment is 99% accuracy to the level of CLSI categories required to interpret AST. Note that even state-of-the-art ID systems have problems with certain pairs, e.g., E. coli vs. Salmonella strains with MALDI. Objective 4. Assemble an integrated photodiode device ready for pre-commercial field evaluation.

[0182] Components. The photodiode-based growth detector and fluid handler disclosed herein are integrated into a single device along with a Gram stain component for bacterial identification. Preliminary studies have either eliminated the risk of these components or demonstrated their full operability. Containers for blood culture bottles, mechanisms for removing / replacing the plate lids, and a touchscreen interface with the necessary GUI and software are also constructed and added. Expected Results

[0183] This proposal would result in a device that reduces the median S2A from 2 days to 6-8 hours. The anticipated outcome is therefore a meaningful step towards reducing mortality from bloodstream infections in people worldwide.

[0184] The phraseology and terminology used herein are for purposes of description and should not be regarded as limiting. As used herein, the term "plurality" refers to two or more items or components. The terms "comprising," "including," "carrying," "having," "containing," and "involving," whether in the specification, claims, or otherwise, are open-ended terms, i.e., meaning "including, but not limited to." Thus, the use of such terms is intended to encompass the items subsequently listed, and equivalents thereof, and additional items. Only the transitional phrases "consisting of" and "consisting essentially of" are closed or partially closed transitional phrases, respectively, with respect to the claims. The use of ordinal terms such as "first," "second," "third," etc. in the claims to modify claim elements does not, in itself, imply a priority, precedence, or ordering of one claim element over another, nor does it imply a chronological order in which the actions of a method are performed, but is merely used as a label to distinguish one claim element having a particular name from another element having the same name (except for the use of ordinal terms).

[0185] Having thus described several aspects of at least one embodiment, it will be understood that various changes, modifications, and improvements will readily occur to those skilled in the art. Any feature described in any embodiment may be included in, or substituted for, any feature of any other embodiment. Such changes, modifications, and improvements are intended to be part of this disclosure and are intended to be within the scope of the invention. Accordingly, the foregoing description and drawings are by way of example only.

[0186] Those skilled in the art should understand that the parameters and configurations described herein are exemplary and that the actual parameters and / or configurations will depend on the particular application in which the disclosed methods and materials are used. Those skilled in the art should also recognize or be able to ascertain using no more than routine experimentation equivalents to the specific embodiments disclosed.

Claims

1. 1. A device configured to perform growth detection, identification, and antimicrobial susceptibility testing (AST) on a microbial species, comprising: a housing configured to receive a sample plate; a sample port configured to receive a sample suspected of containing said microbial species; a fluid distribution system constructed and arranged to introduce said sample into one or more sample wells of said sample plate; a sample plate imaging system; and detecting growth of said microbial species; Identifying the microbial species; To perform AST on the microbial species, a controller configured to collect data from the sample plate imaging system and further configured to process the collected data; Devices that include:

2. 10. The device of claim 1, wherein the fluid distribution system comprises a gantry, a fluid dispensing head operably coupled to the gantry, and a pump fluidly connected to the fluid dispensing head and including a stepper motor shaft having an absolute position magnetic encoder.

3. 10. A device according to any preceding claim, wherein the fluid distribution system is constructed and arranged to facilitate digital growth detection.

4. the sample plate imaging system comprises: a camera configured with an optical system, the camera being connected to the fluid dispensing head; light source, and detector array 10. A device according to any preceding claim, comprising:

5. 10. A device according to any preceding claim, wherein the detector array comprises photodetectors, for example charge coupled devices (CCDs), complementary metal oxide semiconductor (CMOS) detectors, or photodiodes.

6. 10. A device according to any preceding claim, wherein the device further comprises a user interface.

7. 10. The device of any preceding claim, wherein the controller is configured to transmit information regarding growth, identification, and / or AST to a user.

8. 10. The device of any preceding claim, wherein the device further comprises a heater to facilitate sample preparation.

9. 10. The device of any preceding claim, wherein the device is further constructed and arranged to allow for Gram staining of the sample.

10. 10. A device according to any preceding claim, wherein the device is constructed and arranged to perform all three functions in less than about 8 hours.

11. 10. A device according to any preceding claim, wherein the device is constructed and arranged to perform all three functions in less than about six hours.

12. 10. The device of any preceding claim, wherein the device performs AST in less than about 1 hour.

13. 10. The device of any preceding claim, wherein the device is used for both proliferation detection and AST in the same manner.

14. 10. A device according to any preceding claim, wherein the identification of microbial species is based on a stain-free technique.

15. 10. The device of any preceding claim, wherein the microbial species comprises a bacterial species.

16. 10. A device according to any preceding claim, wherein the microbial species is selected from the genera Acinetobacter, Escherichia, Klebsiella, Pseudomonas, Enterococcus, Streptococcus, and Staphylococcus.

17. 10. A device according to any preceding claim, wherein the microbial species falls into one of the following groups of bacteria: Enterococci, Staphylococci, Enterobacterales, Acinetobacter baumannii, and Pseudomonas aeruginosa.

18. 10. A device according to any preceding claim, wherein the microbial species comprises non-bacterial species of fungi, mycobacteria, fecal parasites, blood parasites or tissue parasites.

19. 10. The device of any preceding claim, wherein the sample is a whole blood sample of the subject.

20. 10. A device according to any preceding claim, wherein the sample relates to a pharmaceutical manufacturing ingredient or a finished end product.

21. 10. The device of any preceding claim, wherein the device is configured to perform one or more of growth detection, identification and AST on a second microbial species.

22. A kit comprising a device according to any preceding claim and a sample plate.

23. 23. The kit of claim 22, wherein the sample plate comprises a first portion for proliferation detection and a second portion for AST.

24. 24. The kit of claim 23, wherein one or more wells of the second portion of the sample plate are preloaded with lyophilized antibiotic.

25. 25. The kit of claim 24, wherein the antibiotic is preloaded according to a serial dilution scheme for AST.

26. 26. The kit of any one of claims 22 to 25, wherein the sample plate contains an area for performing sample smears.

27. 27. The kit of any one of claims 22 to 26, wherein one or more wells of the sample plate are defined by a geometry selected to promote sensitivity.

28. 28. The kit of any one of claims 22 to 27, wherein the sample plate contains 384 or 1536 wells.

29. 29. The kit of any one of claims 22 to 28, further comprising a source of growth medium or a detection amplification factor.

30. 30. The kit of any one of claims 22 to 29, further comprising a Gram dye source.

31. 30. The kit of any one of claims 22 to 29, further comprising a sample bottle.

32. 1. A method for detecting the growth of a microbial species, comprising: providing a sample suspected of containing said microbial species; acquiring a time series of images of the sample; and detecting growth of said microbial species by time-dependent changes in at least one of light and color across said time series of images; A method comprising:

33. 33. The method of claim 32, wherein detecting proliferation comprises detecting proliferation using an RGB sensitive imaging sensor in the device.

34. 34. The method of claim 32 or 33, wherein a statistically significant change from baseline in at least one of light and color is interpreted as growth of said microbial species.

35. 10. A method according to any preceding claim, wherein a decrease in light and / or a change in colour over the time series of images is interpreted as growth of said microbial species.

36. 10. The method of any preceding claim, further comprising acquiring a time series of images of each well of the sample plate and detecting growth in each well.

37. 10. The method of any preceding claim, wherein detecting proliferation further comprises comparing one or more of the images of the time series between said wells, between said well and a control, or between a well and a reference time series.

38. 10. The method of any preceding claim, further comprising comparing data across multiple wells to increase sensitivity of proliferation detection.

39. 10. The method of any preceding claim, wherein proliferation detection is achieved in a duration of less than 5 hours.

40. 10. A method according to any preceding claim, wherein several wells of a sample plate with observable growth are used to determine the number of replication-competent microbial cells.

41. 10. The method of any preceding claim, further comprising quantifying the bioburden of the sample based on growth detection.

42. 41. The method of claim 40, further comprising taking remedial action based on the determination of bioburden.

43. 10. The method of any preceding claim, further comprising subjecting said microbial species to identification and / or AST upon detection of growth.

44. 10. The method of any preceding claim, further comprising detecting polymicrobial infections by comparing said time series of images between wells of a sample plate.

45. 44. The method of claim 43, wherein the presence of polymicrobial infection is assessed based on differences in color or doubling time across wells of a microwell plate.

46. 10. The method of any preceding claim, further comprising processing the time series of images to enhance their quality to facilitate the earliest possible detection of proliferation.

47. 10. The method of any preceding claim, further comprising adding a growth medium or a detection amplification factor to the sample.

48. 10. The method of any preceding claim, wherein the sample is a whole blood sample.

49. 10. The method of any preceding claim, wherein the sample relates to a pharmaceutical manufacturing ingredient or a finished end product.

50. 10. The method of any preceding claim, characterized by digital growth detection.

51. 1. A method for identifying a microbial species, comprising: imaging a sample containing a plurality of microorganisms of said microbial species to obtain a series of multi-microbial images; segmenting the multi-microbial image; measuring parameters of each segmented microorganism to obtain a multidimensional distribution of the measured parameters; and Classifying the microbial species based on the multidimensional distribution of measured parameters. A method comprising:

52. 52. The method of claim 51, wherein the measured parameters relate to size, shape, intrinsic color, arrangement and other morphological characteristics of the microbial species.

53. 53. The method of claim 51 or 52, wherein at least one of the measured parameters is selected from the group consisting of width, length, inner density, film thickness, color inhomogeneity, color strength, curvature, taper, aspect ratio, and concavity.

54. 54. The method of any of claims 51 to 53, comprising collecting unique color data associated with the microbial species without staining.

55. 55. The method of any of claims 51-54, wherein the classification comprises machine learning trained on at least one of the following modalities: (i) unstained slides imaged in direct light, (ii) unstained slides imaged in indirect light, and (iii) pre- and post-stain image pairings.

56. 56. The method of any of claims 51 to 55, wherein classifying the microbial species comprises a probability distribution.

57. 57. The method of any of claims 51 to 56, wherein the classifying is performed in a hierarchical manner.

58. 58. The method of any of claims 51 to 57, comprising distinguishing between gram-positive and gram-negative bacteria with a first confidence score.

59. 59. The method of claim 58, wherein each bacterial species is then identified with a second confidence score.

60. 60. The method of any of claims 51 to 59, wherein classification is performed by majority voting, decision trees, or the relative entropy of the observed tally and a reference distribution of known microbial species.

61. 61. The method of any of claims 51 to 60, further comprising filtering the image.

62. 62. The method of claim 61, wherein a blue filter is applied to the image.

63. 63. A method according to any one of claims 51 to 62, wherein images of each sample are taken at multiple focal lengths.

64. 64. A method according to any one of claims 51 to 63, wherein images of each sample are taken at a series of small or fine focal length intervals.

65. 65. The method of claim 64, wherein the focal length spacing is 0.5 μm.

66. 66. The method of any of claims 51 to 65, further comprising acquiring multiple images at each focal length and combining the images.

67. 67. The method of any of claims 51 to 66, further comprising selecting an image having a maximum value of one or more quality metrics for measurement.

68. 68. The method of any of claims 51 to 67, further comprising enhancing the quality of the selected images by removing or smoothing numerical noise before segmentation.

69. 69. The method of any of claims 51 to 68, further comprising performing dimensionality reduction on the multidimensional distribution of measured parameters.

70. 70. The method of any of claims 51 to 69, wherein measuring 100 randomly selected segmented microorganisms is sufficient to identify the microbial species with about 93-97% confidence.

71. 71. The method of any of claims 51 to 70, wherein the sample of the microbial species is derived from a growth detection study.

72. 72. The method of any of claims 51-71, further comprising identifying the second microbial species.

73. A method for performing growth detection, identification, and antimicrobial susceptibility testing (AST) on a microbial species, the method configured to perform all three functions in less than about 8 hours.

74. 74. The method of claim 73, wherein the detection of growth of the microbial species is performed according to any of the preceding claims.

75. 75. The method of claim 73 or 74, wherein the identification of the microbial species is carried out according to any of the preceding claims.

76. 76. The method of any of claims 73 to 75, wherein the AST is based on differential proliferation detection.

77. 77. The method of claim 76, wherein the AST comprises dilution time modeling (DTM).

78. 78. The method of claim 76 or 77, wherein the AST achieves categorical agreement with the reference method in less than one hour.

79. 79. The method of any of claims 76-78, wherein the reference method comprises broth microdilution. Single well: number of wells / total volume of sample plated (e.g., mL) = number of replication-competent cells (e.g., CFU, representing "colony forming units") per unit volume (e.g., CFU / mL).

80. 80. The method of any of claims 76 to 79, wherein the AST relates to an antibiotic selected from the group consisting of cefepime, meropenem, ciprofloxacin, and gentamicin.

81. 81. The method of any of claims 76-80, wherein the results for the AST are reported to an operator, a laboratory information system, and / or an electronic medical record.

82. 82. The method of any of claims 76 to 81, wherein once a microorganism has been identified at a category level, AST may be performed before the microorganism is identified at a species level.

83. 83. The method of any of claims 73 to 82, wherein the three functions are performed by a single device.

84. 1. A quality control method for a pharmaceutical manufacturing process, comprising:

10. A method of assessing the bioburden of a sample containing a pharmaceutical ingredient or a finished end product by subjecting the sample to a method for detecting the growth of a microbial species as defined in any of the preceding claims.

85. 85. The method of claim 84, further comprising approving or rejecting the pharmaceutical ingredient or finished end product based on a comparison of the assessed bioburden to a threshold value.

Images