Rapid tuberculosis and mycobacteria phenotypic drug susceptibility testing with scattering imaging and particle tracking

Abstract

Disclosed herein are methods based on a large volume scattering imaging (LVSim) platform for detection of mycobacterial heteroresistance and conducting drug susceptibility testing (DST). In one aspect, images collected from LVSim are processed using single particle tracking (SPT), which enable tracking of clumping characteristics. The LVSim-SPT approach enables phenotypic detection of a 1% heteroresistant mycobacterium subpopulation in 5 replication cycles and may be used for testing drug susceptibility. In another aspect, images collected from LVSim are analyzed for scattering intensity and scattering count, and comparison of these image characteristics of a test sample with an untreated control sample identifies drug susceptibility.

Description

RAPID TUBERCULOSIS AND MYCOBACTERIA PHENOTYPICDRUG SUSCEPTIBILITY TESTING WITH SCATTERING IMAGINGAND PARTICLE TRACKINGRELATED APPLICATIONS

[0001] This application claims priority to and the benefit of U.S. provisional patent application 63 / 663,062, filed June 21, 2024 to Haydel et al., titled “RAPID TUBERCULOSIS AND MYCOBACTERIA PHENOTYPIC DRUG SUSCEPTIBILITY TESTING WITH SCATTERING IMAGING AND PARTICLE TRACKING,” the entirety of the disclosure of which is hereby incorporated by reference.SEQUENCE LISTING

[0002] In accordance with 37 C.F.R. § 1.831, the present specification makes reference to a Sequence Listing submitted electronically in the form of an XML file (entitled “187WO- PCTSeqList.xml”, created on June 16, 2025, 6,451 bytes in size). The entire contents of the Sequence Listing are herein incorporated by reference in their entirety, with the intention that, upon publication (including issuance), this incorporated Sequence Listing will be inserted in the published document immediately before the claims.TECHNICAL FIELD

[0003] This document relates to pathogen testing apparatus and methods.BACKGROUND

[0004] Mycobacterium tuberculosis, the bacterium responsible for tuberculosis (TB), stands as the 13 th leading cause of death globally, claiming 1.5 million lives each year. In fact, TB causes over 10 million illnesses a year and is estimated to account for 31.8 million deaths between 2020 and 2050 with associated $17.5 trillion in economic impact. TB, while preventable and treatable, poses significant challenges due to the emergence of multidrug-resistant TB (MDR- TB), extensively drug-resistant TB (XDR-TB), and pre-XDR-TB. These strains defy conventional treatment protocols, with MDR-TB being minimally resistant to the primary first line drugs isoniazid and rifampin. Pre-XDR-TB strains further resist fluoroquinolones in addition to isoniazid and rifampin, while XDR-TB, fulfilling pre-XDR-TB criteria, showcases resistance to at least one additional Group A drug. The morbidity and mortality rates associatedwith MDR-, pre-XDR-, and XDR-TB surpass those of antibiotic-susceptible TB, demanding prolonged treatment courses with expensive and more toxic second-line drugs, leading to elevated healthcare costs.

[0005] Worryingly, XDR-TB has been detected in over 90% of countries globally, comprising a notable portion of MDR-TB cases. Mycobacterial drug resistance is a growing global concern, as totally drug-resistant M. tuberculosis strains circulate in both endemic and nonendemic regions around the world. Exacerbating the global health issue, drug-resistant subpopulations, or heteroresistance, of M. tuberculosis are found in up to 30% of tuberculosis patients. This underscores the urgent need for novel methods allowing for accurate and rapid detection of bacterial drug susceptibility profiles and highlights the global inaccessibility of treatment, with only 43% of individuals diagnosed with MDR-TB accessing or enrolling in treatment programs between 2018-2021.

[0006] Accurate and timely detection of heteroresistance in AT. tuberculosis infections has been a long-standing scientific goal, but drug susceptibility testing (DST) for mycobacteria remains a major clinical challenge due to the slow growth and intrinsic drug resistance of many pathogenic species. Conventional phenotypic methods can take up to four weeks post-isolation to yield actionable susceptibility profiles, delaying effective treatment and increasing the risk of resistance development. Additionally, current methods allow only for detection of known molecular markers, which often fails to consistently detect heteroresistance at the therapeutic failure limit (of 1%), or are unable to detect resistance caused by mutations outside of previously characterized genomic regions directly associated with drug resistance.

[0007] The current gold standard for detection of mycobacterial heteroresistance remains agar based, requires up to six weeks, and is labor intensive. Advanced, commercially available drug susceptibility testing platforms, such as the Bactec MGIT 960 and the GeneXpert system, have been adapted and repurposed for the detection of mycobacterial drug heteroresistance more rapidly. Yet, the Bactec MGIT 960 system only allows for a 2-fold decrease in time to result, while the GeneXpert system can only be used to detect known and characterized genetic markers of resistance. Both systems require expensive equipment and instrumentation, unsuitable for low income, endemic regions. Thus, more rapid and accurate methods of drug susceptible testing and detection of heteroresistance are much needed. A rapid and simple platform for accurate TB and mycobacteria phenotypic detection of heteroresistance is paramount for closing the global TB diagnosis and treatment gap and addressing the global health threat of MDR- and XDR-TB.SUMMARY

[0008] Described herein is a method of identifying drug resistance in a population of mycobacteria comprising providing a culture that includes the population of mycobacteria, wherein the culture is split into a test sample and a control sample. An antimicrobial drug is administered to the test sample, and the growth of the mycobacteria is evaluated after drug administration. This evaluation includes recording images of the culture using a large volume scattering imaging (LVSim) platform for a period of at least 1-10 seconds at regular intervals over at least two doubling times of the population of mycobacteria. The method further includes tracking the median intensity of selected pixels from the recorded images, where the selected pixels depict the growth of a single mycobacterium. Drug resistance is detected when the selected pixels depict small cell aggregates instead of a single cell, or medium-sized and / or large-sized mycobacterial clumps.

[0009] Particular embodiments may comprise one or more of the following features. The culture may be split into a plurality of test samples and a control sample, with different concentrations of the antimicrobial drug administered to each test sample. The culture may comprise a concentration of mycobacteria resulting in 10-5,000 particles per field-of-view from the LVSim platform. The selected pixels may depict small cell aggregates with diameters >7 pm that comprise viable cells, indicating heteroresistance. The small cell aggregates may comprise actively dividing and drug-resistant cells, also indicating heteroresistance. The selected pixels may depict medium-sized mycobacterial clumps with diameters of 24.5-49 pm and / or large-sized clumps with diameters >49 pm. The population of mycobacteria may include at least one species selected from the group consisting of Mycobacterium tuberculosis, Mycobacterium abscessus, Mycobacterium avium, Mycobacterium intracellulare, Mycobacterium kansasii, and Mycobacterium smegmatis. The step of tracking median intensity may include removing background noise from the recorded images to reveal underlying single mycobacterial cells or small cell aggregate scattering intensity. Removing background noise may comprise averaging pixel intensity from a plurality of recorded images to produce an afteraveraged image stack, generating a local minimum projection image stack from a plurality of recorded images, subtracting the local minimum from the after-averaged image stack to produce a minimum subtracted image stack, calculating the stack median from both stacks, and subtracting the stack median from the recorded images to remove background noise caused by mechanical drift. The method may further comprise removing local background noise aftermechanical drift noise is removed. The local background noise may be measured from 10 pixels. The plurality of recorded images used for the after-averaged image stack may be four images from adjacent time frames, and the plurality used for the local minimum projection image stack may be ten images from adjacent time frames.

[0010] Also described herein is a method of identifying drug susceptibility in a population of mycobacteria. The method includes providing a culture comprising the population of mycobacteria, wherein the culture is split into a test sample and a control sample. The growth of the population is evaluated after administration of an antimicrobial drug or a combination of antimicrobial drugs to the test sample. This evaluation involves recording images of the culture using a LVSim platform for a period of at least 1-10 seconds at regular intervals for at least two doubling times of the population, to produce a set of recorded images depicting both the test and control samples. In one aspect, scattering intensity is tracked from the recorded images, and the test sample is detected as being inhibited by the drug(s) when the images show decreased scattering intensity in the test sample compared to the control sample. In another aspect, scattering count is tracked from the recorded images, and the test sample is detected as being inhibited by the drug(s) when the images show decreased scattering count in the test sample compared to an initial normalized scattering count value. The initial normalized scattering count value is calculated from the control sample images. In some aspects, both scattering intensity and scattering count are tracked.

[0011] Particular embodiments may comprise one or more of the following features. The detection of inhibition may further comprise identifying that the scattering intensity of the test sample is at or below a species-specific threshold. The detection may further comprise identifying that the scattering count of the test sample is at or below a species-specific threshold. The culture may comprise a concentration of mycobacteria resulting in 10-5,000 particles per field-of-view from the LVSim platform. The population of mycobacteria may include at least one species selected from the group consisting of M. tuberculosis, M. abscessus, M. avium, M. intracellulare , M. kansasii, and AL smegmatis.

[0012] Further described herein is a method of testing drug susceptibility in a population of mycobacteria. The method includes providing a culture comprising the population of mycobacteria, wherein the culture is split into a test sample and a control sample. An antimicrobial drug is administered to the test sample. Images of the culture are recorded using a LVSim platform for a period of at least 1-10 seconds at regular intervals for at least twodoubling times of the population of mycobacteria. The method further includes counting medium-sized and / or large-sized mycobacterial clumps from the recorded images.

[0013] Particular embodiments may comprise one or more of the following features. Mediumsized mycobacterial clumps may have diameters of 24.5-49 pm and / or large-sized clumps may have diameters of >49 pm. The method may include detecting that the population is susceptible to the antimicrobial drug when the recorded images of the test sample show a suppressed number of medium-sized and / or large-sized clumps compared to the control sample. The culture may be split into a plurality of test samples and a control sample, with different concentrations of the antimicrobial drug administered to each test sample. The population of mycobacteria may include at least one species selected from the group consisting of M. tuberculosis, M. abscessus, M. avium, M. intracellulare , M. kansasii, and M. smegmatis. The culture may comprise a concentration of mycobacteria resulting in 10-5,000 particles per field- of-view from the LVSim platform. Medium-sized and / or large-sized clumps may be counted by tracking the median intensity of selected pixels from the recorded images, where the selected pixels depict the growth of a single mycobacterium. Tracking median intensity may include removing background noise from the recorded images to reveal the underlying scattered intensity of a single mycobacterial cell or small aggregate. Removing background noise may involve averaging pixel intensity from a plurality of recorded images to produce an afteraveraged image stack, generating a local minimum projection image stack, subtracting the local minimum from the after-averaged stack to produce a minimum subtracted image stack, calculating the stack median from both stacks, and subtracting the stack median from the recorded images to remove background noise caused by mechanical drift. The method may further include removing local background noise from the recorded images after mechanical drift noise is removed.

[0014] A method of identifying drug susceptibility or resistance within a population of mycobacteria utilizing machine learning and deep learning algorithms is also disclosed. The method comprises providing a culture comprising the population of mycobacteria, wherein the culture is split into a test sample and a control sample; obtaining a set of recorded images of the population of mycobacteria using the LVSim platform over a period of observation time after the test sample is administered an antimicrobial drug or a combination of antimicrobial drugs, wherein the set of recorded images comprises a set of images depicting the test sample and a set of images depicting the control sample; and processing the set of recorded images to extract the measurements related to cell growth, cell movement, cell morphology, intensity,and aggregation from the set of images depicting the test sample and the set of images depicting the control sample. The method next comprises predicting the drug susceptibility classification of the population of mycobacteria by providing the measurements related to cell growth, cell movement, cell morphology, intensity, and aggregation to one or more machine learning models as input; and outputting the drug susceptibility classification. The one or more machine learning models are generated by a training data set comprising measurements related to cell growth, cell movement, cell morphology, intensity, and aggregation of a different population of mycobacteria being inhibited by an antimicrobial drug and measurements related to cell growth, cell movement, cell morphology, intensity, and aggregation of a population of mycobacteria that has not administered an antimicrobial drug.

[0015] In some aspects, the outputted drug susceptibility classification includes susceptibility to the antimicrobial drug or combination of antimicrobial drugs, resistance to the antimicrobial drug or combination of antimicrobial drugs, intermediate category between susceptibility and resistance to the antimicrobial drug or combination of antimicrobial drugs, and heteroresistance to the antimicrobial drug or combination of antimicrobial drugs. In some implementations, the period of observation time is not more than two doubling times of the population of mycobacteria.

[0016] A LVSim platform is additionally described herein. The platform comprises a camera with a lens assembly, a linear translational stage, a beam-block, and a lighting unit. The linear translational stage comprises a motorized belt driven sample tray and a thermal control unit, wherein the translational stage linearly shifts samples in the sample tray. The thermal control unit is beneath the sample tray and provides precision temperature control to create an appropriate growth environment for mycobacterial cells. The camera, beam-block, lighting unit are placed in a line and the beam-block is placed in front of the light unit thereby preventing direct illumination of the sample tray and the cameraBRIEF DESCRIPTION OF THE DRAWINGS

[0017] The patent or application file contains at least one drawing executed in color. Copies of this patent or patent application publication with color drawing(s) will be provided by the Office upon request and payment of the necessary fee.

[0018] Implementations will hereinafter be described in conjunction with the appended and / or included DRAWINGS, where like designations denote like elements.

[0019] The features and advantages of the present disclosure, and the manner of attaining them, will become more apparent and the present disclosure will be better understood by reference to the description of the present disclosure taken in conjunction with the accompanying drawings, wherein:

[0020] FIGs. 1A-1C depict, in accordance with certain embodiments, the LVSim single particle tracking (SPT) principle of detecting 1% resistant mycobacteria subpopulations. FIG. 1A: Growth dynamics of heteroresistant mycobacteria populations. In the presence of antibiotics, the growth of susceptible mycobacterium spp. is largely inhibited, while resistant mycobacteria slowly grow and increase its proportion within the sample. FIG. IB: Ensembled analysis of heteroresistant sample (left) versus resistant-targeted tracking (middle). Targeted tracking of resistant species and unveiling its transient growth among the inhibited background reveals earlier growth compared to ensemble analysis (right). FIG. 1C: Representative LVSim image of M. tuberculosis cells. Despite the spatial resolution limit, single cell (~ 3-4 pm in length, left) can be differentiated from small cell aggregates (~ 12 pm in length, middle) based on particle intensity. By tracking population particle intensity distribution over time, growth dynamics of rare resistant subpopulations can be unveiled by identifying actively resistant cells based on particle intensity.

[0021] FIGs. 2A and 2B depict, in accordance with certain embodiments, LVSim-SPT rapid detection of 1% kanamycin-resistant Mycobacterium smegmatis. FIG. 2A: Ensemble LVSim tracking analysis of the whole heteroresistant population: 100% KanS:0% KanR M. smegmatis (black) and 99% KanS:l% KanR AL smegmatis (red) populations in the presence of Kan (8 pg / mL). Ensemble LVSim tracking detected 1% AL smegmatis heteroresistant cells in 28-32 hours. FIG. 2B: Targeted LVSim tracking of actively growing resistant cells: 100% KanS:0% KanR AL smegmatis (black) and 99% KanS: 1% KanR AL tuberculosis (red) populations in the presence of Kan (8 pg / mL). Targeted LVSim tracking detected 1% AL smegmatis heteroresistant cells in 23 hours. LVSim measurements were collected every 2-4 hours for 32 hours.

[0022] FIGs. 3A and 3B depict, in accordance with certain embodiments, LVSim-SPT rapid detection of 1% kanamycin-resistant AL tuberculosis. FIG. 3A: Ensemble LVSim intensity tracking analysis of the whole heteroresistant population: 100% KanS:0% KanR AL tuberculosis (black) and 99% KanS: 1% KanR AL tuberculosis (red) populations in the presence of Kan (8 pg / mL). Ensemble LVSim tracking detected 1% AL tuberculosis heteroresistant cells in 108 hours. FIG. 3B: Ensemble LVSim intensity tracking analysis of the wholeheteroresistant population: 100% KanS:0% KanR M. tuberculosis (black) and 98% KanS:2% KanR M. tuberculosis (red) populations in the presence of Kan (8 pg / mL). Ensemble LVSim tracking detected 2% AL tuberculosis heteroresistant cells in 108 hours.

[0023] FIGs. 4A-4C, depict, in accordance with certain embodiments, mycobacterial clumping dynamics revealed by LVSim. FIG. 4A: Classification of mycobacterial clumps based on LVSim sizes. Small-clumps: D < 24.5 pm; Medium -clumps: 24.5 pm < D < 49 pm; Large- clumps: D > 49 pm. FIG. 4B: LVSim results for M. tuberculosis without antibiotic treatment at different times. H: hour. Red arrows: Indicate the clumps shown in detail in FIG. 4A. FIG. 4C: Histogram of AL tuberculosis sample without (top) or with (bottom) antibiotic treatment at 0 hours (blue) and 40 hours (orange). Mycobacterial clumps within the imaging volume were grouped with bin-sizes of 3.5 pm, and categorized into small-, medium- and large-clumps based on their largest size calculated from their size profile during the recording period.

[0024] FIGs. 5A-5E depict, in accordance with certain embodiments, clumping dynamics of M. tuberculosis as a LVSim rapid phenotypic DST feature. FIG. 5A: Global fit of M. tuberculosis clump formation and dynamics without (left, black) or with (right, red) antibiotic treatment at 0 hours (hatched line) and 40 hours (solid line). Clumps were grouped into small (D < 24.5 pm), medium (24.5 pm < D < 49 pm), and large (D > 49 pm) clumps based on their largest size calculated from their size profile during the recording period. FIG. 5B: Calculated abundance (percentage) and increase of M. tuberculosis small, medium, and large clumps in an untreated control sample from 0 hours (top) to 40 hours (bottom). FIG. 5C: Relative population abundance and dynamics of small, medium, and large M. tuberculosis clumps in antibiotic untreated (top) and treated (bottom) sample at 0 hours (blue) and 40 hours (orange). FIG. 5D: Fold change of AL tuberculosis clump subpopulations from 0 hours to 40 hours for antibiotic untreated (blue) and treated (orange) samples. FIG. 5E: Normalized count of medium (top) and large (bottom) AL tuberculosis clumps over time in the absence (black) or presence (red) of an antibiotic.

[0025] FIGs. 6A and 6B depict, in accordance with certain embodiments, clumping dynamics of AL abscessus Rough as a LVSim rapid phenotypic DST feature. FIG. 6A: LVSim normalized count of medium AL abscessus myco-clumps over time in the absence (black) or presence (red) of an antibiotic or experimental compound. FIG. 6B: LVSim normalized count of large AL abscessus myco-clumps over time in the absence (black) or presence (red) of an antibiotic or experimental compound.

[0026] FIGs. 7A-7F depict, in accordance with certain embodiments, time-resolved measurement and LVSim ensemble scattering intensity tracking (FIGs. 7A, 7C, and 7E) and validation using endpoint microdilution CFU enumeration (FIGs. 7B, 7D, and 7F) of M. tuberculosis with first line (rifampin, FIGs 7A and 7B), group A (bedaquiline, FIGs. 7C and 7D), and experimental (diarylthiazole-48, FIGs. 7E and 7F) antibiotics.

[0027] FIGs. 8A-8E depict, in accordance with certain embodiments, utilization of time- resolved LVSim scattering intensity (FIGs. 8A-i, 8B-i, 8C-i, 8D-i, and 8E-i) and count (FIGs. 8A-ii, 8B-ii, 8C-ii, 8D-ii, and 8E-ii) tracking, and validation using endpoint microdilution CFU enumeration (FIGs. 8A-iii, 8B-iii, 8C-iii, 8D-iii, and 8E-iii) for AL abscessus Smooth DST with conventional and novel antibiotics.

[0028] FIGs. 9A-9D depict, in accordance with certain embodiments, utilization of time- resolved LVSim scattering intensity (FIGs. 9A-i, 9B-i, 9C-i, and 9D-i) and count (FIGs. 9A-ii, 9B-ii, 9C-ii, and 9D-ii) tracking, and validation using endpoint microdilution CFU enumeration (FIGs. 9A-iii, 9B-iii, 9C-iii, and 9D-iii) for M. abscessus Rough DST with conventional and novel antibiotics.

[0029] FIGs. 10A-10D depict, in accordance with certain embodiments, the LVSim multiplexed setup. FIG. 10A depicts the side view. FIG. 10B depicts the top view. FIG. 10C depicts a 3D view. FIG. 10D depicts the sample tray design. The system adopts a forward scattering geometry (i.e., illumination pathway and detection pathway were aligned on the same axis) to create a uniform illumination over the image volume. A beam-block is placed in front of the lighting unit (LED array) to prevent direct illuminating of the sample and the camera. The recorded dark field style images largely suppress background, while bacterial cells and similar particles appear as bright spots compared to the dark background, creating high- contrast images with high signal-to-noise ratios. A sample tray with eight cuvettes is carried by a stepper motor, which provides programmable and precise motion. A built-in thermal control unit beneath the stepper motor offers 0.1°C temperature control to create an appropriate growth environment for mycobacterial cells. Cuvettes with loaded mycobacterial cultures are inserted into the sample tray and are periodically and sequentially imaged. Recording intervals and total recording times depended on the experimental conditions required for specific mycobacteria.

[0030] FIGs. 11A-11D depict, in accordance with certain embodiments, the LVSim high- throughput multiplexed setup. The principle of LVSim high-throughput multiplexed setup resembles the LVSim multiplexed setup and includes an automated cuvette expansion bank fora high-throughput LVSim DST system. The cuvette bank holds multiple parallel cuvette racks, and the base can be translated horizontally so that it can position any one of the racks next to the imaging position. The cuvette bank can accommodate 2-12 racks. The rack being measured will be translated to the imaging position sequentially to measure all cuvettes on the rack. The mechanical translations are realized with long range linear sliders, which can provide up to 210 mm stoke with up to 200mm / s speed and 3 mm precision. Individual cuvette racks have temperature control function to maintain the cuvette at 35-37°C for optimal mycobacteria growth. Mixing function is implemented to the cuvette bank using linear back and forth agitation driven by the sliders. The logistics of the mixing and positioning will be controlled by the system software.

[0031] FIG. 12 depicts, in accordance with certain embodiments, the system software, which includes detection optics, temperature and sample control hardware, system control, userinterface, and data analysis software.

[0032] Those of ordinary skill in the art will understand that the compositions, methods, and systems specifically described herein and illustrated in the accompanying drawings are nonlimiting exemplary embodiments and that the scope of the various embodiments of the present disclosure is defined solely by the claims. The features illustrated or described in connection with one exemplary embodiment may be combined with the features of other embodiments. Such modifications and variations are intended to be included within the scope of the present disclosure.DETAILED DESCRIPTION

[0033] Detailed aspects and applications of the disclosure are described below in the following drawings and detailed description of the technology. Unless specifically noted, it is intended that the words and phrases in the specification and the claims be given their plain, ordinary, and accustomed meaning to those of ordinary skill in the applicable arts.

[0034] In the following description, and for the purposes of explanation, numerous specific details are set forth to provide a thorough understanding of the various aspects of the disclosure. It will be understood, however, by those skilled in the relevant arts, that the present disclosure may be practiced without these specific details. It should be noted that there are many different and alternative configurations, devices, and technologies to which the disclosed embodiments may be applied. The full scope of the disclosures is not limited to the examples that are described below.

[0035] The singular forms “a,” “an,” and “the” include plural referents unless the context clearly dictates otherwise. Thus, for example, reference to “a step” includes reference to one or more of such steps.

[0036] For the methods described herein, the culture comprising the population of mycobacteria may be biological sample collected directly from the patient or a cultured patient sample. As such, the culture comprising the population of mycobacteria may be a tissue biopsy from the lungs, liver, or bone marrow, sputum, blood, or aspiration. In another implementation, the culture comprising the population of mycobacteria is cultured from a patient’s biological sample.

[0037] “Deep learning” as used herein, is used in its generally accepted meaning as a class of machine learning algorithms using a cascade of many layers of nonlinear processing units, as for example neural networks and adaptive processors, that can be based on unsupervised or supervised learning, pattern analysis applications and the like.

[0038] Identified in more than 90% of countries worldwide, multi drug-resistant TB (MDR- TB) represents a serious public health concern. Gold standard TB phenotypic drug susceptibility testing (pDST) requires mycobacteriology reference lab processing and can take 28 - 42 days to obtain results. While molecular technologies rapidly detect drug resistance genes or DNA mutations, they only work for known genetic traits of drug resistance. Therefore, DNA tests to detect mycobacterial resistance to new drugs, such as bedaquiline (BDQ) and linezolid (LZD), are not yet available since information regarding the associated genetic resistance mechanisms is currently emerging. To address this significant need, disclosed herein is a phenotypic-based, label-free optical imaging system and method that simultaneously determines Mycobacterium species’ susceptibility and resistance to new TB drugs within 24 hours. The system and method enable rapid pDST with new drugs that have no or limited genetic-based testing methods available.

[0039] The optical imaging system is referenced herein as Large Volume Scattering imaging (LVSim) and rapid machine learning (ML) TB pDST (LVSim-TBDST) system. With turnaround times of 6 - 24 hours from puretuberculosis cultures, LVSim-TBDST rapidly inform clinical decisions and predict success or failure of MDR-TB drug treatment. The speed of the disclosed system and method in identifying drug susceptibility facilitate prompt, effective treatment of MDR-TB patients by rapidly determining phenotypic susceptibility or resistance to new TB drugs, TB drugs currently in clinical trials, or anti-TB compounds in preclinical development, all of which lack molecular or commercial assays for monitoringemerging resistance. This low-magnification solution scattering imaging system and real-time video-based object scattering intensity detection to track multiple intrinsic phenotypic features of individual TB cells and clusters for rapidly correlating and predicting TB growth and response to drug exposure. Without the need for any molecular labeling or biochemical tags, scattering optical imaging from M. tuberculosis in liquid culture is obtained by subtracting both spatial and temporal background from short videos.

[0040] The LVSim-TBDST system is a continuous monitoring platform and comprises a linear translational stage comprising a motorized belt driven sample tray and a thermal control unit (also described herein as a stepping motor carrying a sample tray). The sample tray is driven by the stepper motor that provides programmable, precise linear motion, enabling sequential imaging of multiple samples for determination of drug susceptibility, minimum inhibitory concentration (MIC), or resistance. The thermal control unit is beneath the sample tray and may be built-in beneath the stepping motor. In some aspects, the thermal control unit provides 0.1°C-precision temperature control to create an appropriate growth environment for mycobacterial cells at 37°C. The system further comprises a beam-block, a lighting unit, and a camera with a lens assembly. The beam-block is placed in front of the lighting unit (for example, an LED array) to prevent direct illuminating of the sample and the camera. Thus, the camera, beam-block, lighting unit are placed in a line. The camera capture dark-field style images, which have largely suppressed background. Thus, bacterial cells and similar particles appear as bright spots compared to the dark background, creating high-contrast images with high signal -to-noise ratios. Each cuvette is periodically and sequentially imaged for 10 to 80 seconds per cuvette. Recording intervals and total recording times depend on the experimental conditions required for specific mycobacteria.

[0041] The method of identifying drug resistance of a mycobacteria culture and the presence of heteroresistance in a mycobacteria culture relies on the analysis of single particle count and intensity emitted from individual bacterial particles and their change over time, allowing for classification of growing versus non-growing bacteria, even below the spatial resolution limit.

[0042] LVSim leverages the intrinsic clumping behavior of mycobacteria, a function of their mycolic acid-rich cell envelope and a recognized virulence trait as a dynamic biomarker of viability. In the absence of effective antibiotic treatment, growing mycobacteria form larger and more numerous clumps. Conversely, successful drug exposure suppresses this phenotype. Accordingly, clump detection and clump size tracking are novel phenotypic markers for drug susceptibility determination. By sensitively tracking clump size and number in real-time andapplying a quantitative threshold (for example, ~25 pm), LVSim enables phenotypic susceptibility determination across diverse antibiotic classes within 48 hours. The examples demonstrate the utility of this approach across M. tuberculosis and both smooth and rough morphotypes oiM. a scessus, including susceptibility profiling for investigational agents. The up to 63-fold reduction in turnaround time relative to solid-based phenotypic DST positions LVSim as a practical and scalable platform for precision-guided TB therapy.

[0043] The time profile of multiple M. tuberculosis phenotypic signatures and signature patterns, including cell count, proliferation, cell aggregation, and morphological intensity, obtained by optical tracking can be analyzed via ML and correlated with the mycobacterial growth rate and mycobacterial response to drug exposure. LVSim-TBDST generates a comprehensive report providing quantitative drug resistance and susceptibility information and patient-specific drug recommendations. In one example, LVSim-TBDST data revealed M. tuberculosis susceptibility to rifampin within 16 hours, which is within one TB doubling time (20-24 hours) and susceptibility to BDQ within 8 hours, approximately one-third of TB doubling time.

[0044] The LVSim-TBDST methodology may be generalized as a method of identifying drug susceptibility in a population of mycobacteria. The method is applicable to all mycobacterium species, thus population of mycobacteria may be inclusive of all mycobacterium species. In some aspects, the species in population of mycobacteria are from at least one mycobacterium species selected from the group consisting of: Mycobacterium tuberculosis, Mycobacterium abscessus, Mycobacterium avium, Mycobacterium intr acellular e, Mycobacterium kansasii, and Mycobacterium smegmatis.

[0045] In one aspect, the method comprises providing a culture comprising the population of mycobacteria, wherein the culture is split into a test sample and a control sample; and evaluating the growth of the population of mycobacteria after an antimicrobial drug is administered to the test sample. The test sample is administered the antimicrobial drug, while the control sample is untreated. The evaluation step comprises recording images of the culture (both the test sample and the control sample) through the LVSim platform for a period of at least 1-10 seconds at regular intervals for at least two doubling times of the population of mycobacteria to produce a set of recorded images depicting the test sample and the control sample; and tracking scattering intensity from the set of recorded images. In some implementations, the set of recorded images comprises images of the culture recorded through the LVSim platform for a period of at least 1-10 seconds at regular intervals for at least twodoubling times of the population of mycobacteria. The method further comprises detecting the test sample as being inhibited by the antimicrobial drug upon the set of recorded images depicting the test sample having decreased scattering intensity compared to the set of recorded images depicting the test sample and the untreated control sample. In some implementations, detection of the test sample as being inhibited by the antimicrobial drug further requires the scattering intensity of the set of recorded figures to be at or below a species-specific threshold 1.1.

[0046] In some embodiments, a combination of antimicrobial drugs is administered to the test sample. In some embodiments, the culture is split into multiple test samples. In such embodiments, each test sample is administered a different antimicrobial drug thus evaluating multiple antimicrobial drugs. Alternatively, each test sample is administered a different concentration of the antimicrobial drug if only one antimicrobial drug is evaluated. In yet another implementation, each test sample is administered a different combination of antimicrobial drugs, different concentrations of the same combination of antimicrobial drugs, or different concentrations of a variety of combinations of antimicrobial drugs. In certain implementations, the provided culture comprises a concentration of mycobacteria resulting in 10-5,000 particles per field-of-view from the LVSim platform.

[0047] In another aspect, the method comprises providing a culture comprising the population of mycobacteria, wherein the culture is split into a test sample and a control sample and evaluating the growth of the population of mycobacteria after the antimicrobial drug is administered to test sample. This evaluation step comprises recording images of the culture through the LVSim platform for a period of at least 1-10 seconds at regular intervals for at least two doubling times of the population of mycobacteria to produce a set of recorded images depicting the test sample and the control sample; and tracking scattering count from the set of recorded images, wherein an initial normalized scattering count value is calculated from the set of recorded images depicting the control sample. In some implementations, the set of recorded images comprises images of the culture recorded through the LVSim platform for a period of at least 1-10 seconds at regular intervals for at least two doubling times of the population of mycobacteria. The method further comprises detecting the test sample as being inhibited by the antimicrobial drug upon the set of recorded images depicting the test sample having decreased scattering count compared to the initial normalized scattering count value. In some implementations, detection of the test sample as being inhibited by the antimicrobial drugfurther requires the scattering count of the set of recorded figures to be at or below a speciesspecific threshold. In some aspects, the scattering count threshold is 1.15.

[0048] In some embodiments, a combination of antimicrobial drugs is administered to the test sample. In some embodiments, the culture is split into multiple test samples. In such embodiments, each test sample is administered a different antimicrobial drug thus evaluating multiple antimicrobial drugs. Alternatively, each test sample is administered a different concentration of the antimicrobial drug if only one antimicrobial drug is evaluated. In yet another implementation, each test sample is administered a different combination of antimicrobial drugs, different concentrations of the same combination of antimicrobial drugs, or different concentrations of a variety of combinations of antimicrobial drugs. In certain implementations, the provided culture comprises a concentration of mycobacteria resulting in 10-5,000 particles per field-of-view from the LVSim platform.

[0049] In some implementations of the method of identifying drug susceptibility in a population of mycobacteria, both scattering intensity and scattering count are evaluated. Thus, in certain embodiments, the method comprises providing a culture comprising the population of mycobacteria, wherein the culture is split into a test sample and a control sample; and evaluating the growth of the population of mycobacteria after the antimicrobial drug is administered to test sample. The evaluation step comprises recording images of the culture through the LVSim platform for a period of at least 1-10 seconds at regular intervals for at least two doubling times of the population of mycobacteria to produce a set of recorded images; and tracking scattering intensity and scattering count from the set of recorded images. In some implementations, the set of recorded images comprises images of the culture recorded through the LVSim platform for a period of at least 1-10 seconds at regular intervals for at least two doubling times of the population of mycobacteria. The method further comprises detecting the population of mycobacteria as being inhibited by the antimicrobial drug upon the set of recorded images depicting the test sample having decreased scattering intensity compared to that of the set of recorded images depicting the control sample and the set of recorded images depicting the test sample having decreased scattering count compared to the initial normalized scattering count value, which is calculated from the set of recorded images depicting the control sample.

[0050] The LVSim platform may also be used for phenotypic detection of mycobacterial heteroresistance at the 1% therapeutic failure limit - also referred to herein as the LVSim- SPT methodology. As shown in the examples, through single particle tracking, LVSim can be usedto phenotypically detected M. smegmatis 1% kanamycin heteroresistance within a mixed population, exceeding genotypic detection sensitivity. The examples demonstrate that LVSim pDST can be a transformative method in the detecting TB minor resistant subpopulations. Thus, the LVSim-SPT methodology may be generalized as a method of identifying drug resistance in a population of mycobacteria. In some embodiments, the species in population of mycobacteria are from at least one mycobacterium species, which includes but is not limited to Mycobacterium tuberculosis, Mycobacterium abscessus, Mycobacterium avium, Mycobacterium intr acellular e, Mycobacterium kansasii, and Mycobacterium smegmatis.

[0051] This method comprises providing a culture comprising the population of mycobacteria; administering an antimicrobial drug to the culture; and evaluating the growth of the population of mycobacteria after the antimicrobial drug is administered. The evaluation step comprises recording images of the culture through the LVSim platform for a period of at least 1-10 seconds at regular intervals for at least two doubling times of the population of mycobacteria; and tracking median intensity of selected pixels from the recorded images, wherein the selected pixels depict growth of a single mycobacterium. The method further comprises detecting drug resistance within the population of mycobacteria in response to exposure to the antimicrobial drug upon the selected pixels depicting clumping within the culture, for example, the detection of small cell aggregates instead of a single cell or medium-sized and / or large-sized mycobacterial clumps. The medium-sized clumps have diameters of 24.5 < 49 pm, while the large-sized clumps have diameters of > 49 pm. In some embodiments, two or more antimicrobial drugs are administered to the culture. In some embodiments, the culture is split into multiple test samples and a control sample, wherein each test sample is administered a different concentration of the antimicrobial drug, different combination of antimicrobial drugs, different concentrations of the same combination of antimicrobial drugs, or different concentrations of a variety of combinations of antimicrobial drugs. In certain implementations, images of the culture are recorded through the LVSim platform for a period of at least 1-10 seconds at regular intervals for at least two doubling times of the population of mycobacteria. In some aspects, the provided culture comprises a concentration of mycobacteria resulting in 10-5,000 particles per field-of-view from the LVSim platform.

[0052] In some implementations, the step of tracking median intensity of selected pixels from the recorded images comprises removing background noise from the recorded images thereby revealing an underlying single mycobacterial cell scattered intensity. In particularimplementations, the step of removing background noise involves removing background noise caused by mechanical drift and removing local background noise (for example, from 10 pixels).

[0053] The step of removing background noise caused by mechanical drift comprises averaging intensity of pixels from a plurality of recorded images to produce an after-averaged image stack; producing a local minimum projection from a plurality of recorded images at a pixel level to produce a local minimum projection image stack; producing a noise-free image stack subtracting after-averaged image stack by the temporal local minimum to produce a minimum subtracted image stack; and calculating the stack median from the after-averaged image stack and the minimum subtracted image stack. Background noise caused by mechanical drift is removed by subtracting the stack median from the intensity of a recorded image.

[0054] In some aspects, the averaged image stack is generated from intensities from four images from adjacent time frames. As the local minimum projection image stack is a stack of images representing a temporal local minimum, in some aspects, the local minimum projection image stack is produced from 10 images from adjacent time frames.

[0055] The single particle tracking methodology may also be used for evaluating clumping characteristics, which is useful for detecting drug susceptibility. The method of testing drug susceptibility in a population of mycobacteria comprises providing a culture comprising the population of mycobacteria, wherein the culture is split into a test sample and a control sample; administering an antimicrobial drug to the test sample; recording images through the LVSim platform for a period of at least 1-10 seconds at regular intervals for at least two doubling times of the population of mycobacteria; and counting medium-sized and / or large-sized mycobacterial clumps from the recorded images. In some aspects, medium-sized mycobacterial clumps have diameters of 24.5 < 49 pm. In some aspects, large-sized mycobacterial clumps have diameters of > 49 pm.

[0056] The method of testing drug susceptibility further comprises detecting the population of mycobacteria is susceptible to the antimicrobial drug upon the recorded images depicting the test sample having suppressed number of medium-sized and / or large-sized mycobacterial clumps compared to the recorded images depicting the control sample. In some embodiments, two or more antimicrobial drugs are administered to the culture. In some embodiments, the culture is split into multiple test samples and a control sample, wherein each test sample is administered a different concentration of the antimicrobial drug, different combination of antimicrobial drugs, different concentrations of the same combination of antimicrobial drugs, or different concentrations of a variety of combinations of antimicrobial drugs. In certainimplementations, images of the culture are recorded through the LVSim platform for a period of at least 1-10 seconds at regular intervals for at least two doubling times of the population of mycobacteria. In some aspects, the provided culture comprises a concentration of mycobacteria resulting in 10-5,000 particles per field-of-view from the LVSim platform.

[0057] In some implementations, the step of counting medium-sized and / or large-sized mycobacterial clumps from the recorded images comprises tracking median intensity of selected pixels from the recorded images, wherein the selected pixels depict growth of a single mycobacterium. In some aspects, the step of tracking median intensity of selected pixels from the recorded images comprises removing background noise from the recorded images thereby revealing an underlying single mycobacterial cell or small cell aggregate scattered intensity.

[0058] In particular implementations, the step of removing background noise involves removing background noise caused by mechanical drift and removing local background noise (for example, from 10 pixels).

[0059] The step of removing background noise caused by mechanical drift comprises averaging intensity of pixels from a plurality of recorded images to produce an after-averaged image stack; producing a local minimum projection from a plurality of recorded images at a pixel level to produce a local minimum projection image stack; producing a noise-free image stack subtracting after-averaged image stack by the temporal local minimum to produce a minimum subtracted image stack; and calculating the stack median from the after-averaged image stack and the minimum subtracted image stack. Background noise caused by mechanical drift is removed by subtracting the stack median from the intensity of a recorded image.

[0060] In some aspects, the averaged image stack is generated from intensities from four images from adjacent time frames. As the local minimum projection image stack is a stack of images representing a temporal local minimum, in some aspects, the local minimum projection image stack is produced from 10 images from adjacent time frames.

[0061] By integrating bacterial growth tracking, morphology tracking, and motion tracking, LVSim establishes a transformative, data-rich platform for rapid, predictive pDST. An advanced computational framework enhances the speed, accuracy, and automation of resistance detection, positioning LVSim as a next-generation platform for real-time, high- throughput mycobacteria pDST for any new, preclinical, or experimental drug that lacks a genotypic DST method. Accordingly, a method of identifying drug susceptibility in a population of mycobacteria utilizing machine learning and deep learning algorithms is also disclosed. This method tracks multiple phenotypic features, including, for example, growthdynamics, morphological changes, and movement patterns, to predict drug susceptibility without the need for defining and quantifying each feature. In one aspect, the method tracks cell division in real time thereby tracking bacterial growth. In another aspect, the method quantifies total scattering intensity and its temporal variation thereby providing information on bacterial size, shape, and physiological -associated variations in response to drug exposure. In yet another aspect, the method utilizes advanced particle motion analysis to differentiate bacteria based on sizes, shape, and activity. Data associated with the multiple phenotypic features (for example, cell growth, cell movement, cell morphology, intensity, and aggregation) are inputted into a training algorithm (for example, convolution neural network or long shortterm memory) to output a predictive phenotypic DST method.

[0062] Accordingly, this method of identifying drug susceptibility and / or drug resistance in a population of mycobacteria comprising providing a culture comprising the population of mycobacteria, wherein the culture is split into a test sample and a control sample; and obtaining a set of recorded images of the population of mycobacteria using the LVSim platform over a period of observation time after the test sample is administered an antimicrobial drug or a combination of antimicrobial drugs. The control sample is not administered the antimicrobial drug or combination of antimicrobial drugs. The set of recorded images comprises a set of images depicting the test sample and a set of images depicting the control sample. The method next comprises processing the set of recorded images to extract the measurements related to cell growth, cell movement, cell morphology, intensity, and aggregation; predicting the drug susceptibility classification of the population of mycobacteria by providing the measurements related to cell growth, cell movement, cell morphology, intensity, and aggregation to one or more machine learning models as input; and outputting the drug susceptibility classification. The one or more machine learning models are generated by a training data set comprising measurements related to cell growth, cell movement, cell morphology, intensity, and aggregation of a different population of mycobacteria being inhibited by an antimicrobial drug and measurements related to cell growth, cell movement, cell morphology, intensity, and aggregation of a population of mycobacteria that has not been administered an antimicrobial drug. The one or more machine learning models may be selected from convolution neural network and long short-term memory network. The outputted drug susceptibility classification includes susceptibility to the antimicrobial drug or combination of antimicrobial drugs, resistance to the antimicrobial drug or combination of antimicrobial drugs, intermediate category between susceptibility and resistance to the antimicrobial drug or combination ofantimicrobial drugs, and heteroresistance to the antimicrobial drug or combination of antimicrobial drugs.

[0063] In some aspects, the set of recorded images comprises a set of images depicting the test sample and a set of images depicting the control sample over a period of at least two doubling times of the population of mycobacteria, wherein the LVSim platform records an image for a period of at least 1-10 seconds at regular intervals during the period of observation time. In certain implementations, the period of observation time is not more than two doubling times of the population of mycobacteria.

[0064] More specifically, this disclosure, its aspects and embodiments, are not limited to the specific material types, components, methods, or other examples disclosed herein. Many additional material types, components, methods, and procedures known in the art are contemplated for use with particular implementations from this disclosure. Accordingly, for example, although particular implementations are disclosed, such implementations and implementing components may comprise any components, models, types, materials, versions, quantities, and / or the like as is known in the art for such systems and implementing components, consistent with the intended operation.EXAMPLES

[0065] The present disclosure is further illustrated by the following examples that should not be construed as limiting. The contents of all references, patents, and published patent applications cited throughout this application, as well as the Figures, are incorporated herein by reference in their entirety for all purposes.Example 1. Methods and Materials

[0066] LVSim multiplexed setup. The principle of LVSim multiplexed setup resembles dark field microscopy. The system adopts a forward scattering geometry (i.e., illumination pathway and detection pathway were aligned on the same axis) to create a uniform illumination over the image volume (FIGs. 10A and 10B). A beam-block was placed in front of the lighting unit (LED array) to prevent direct illuminating of the sample and the camera (FIG. 10A). The recorded dark field style images largely suppressed background, while bacterial cells and similar particles appear as bright spots compared to the dark background, creating high-contrast images with high signal-to-noise ratios. A sample tray with eight cuvettes was carried by a stepper motor, which provided programmable and precise motion (FIGs. 10C and 10D). Abuilt-in thermal control unit beneath the stepper motor offered 0.1 °C temperature control to create an appropriate growth environment for mycobacterial cells. Cuvettes with loaded mycobacterial cultures were inserted into the sample tray and were periodically and sequentially imaged for 80 seconds per cuvette. Recording intervals and total recording times depended on the experimental conditions required for specific mycobacteria.

[0067] LVSim high-throughput multiplexed setup. The principle of LVSim high- throughput multiplexed setup resembles the LVSim multiplexed setup and includes an automated cuvette expansion bank for a high-throughput LVSim-TBDST system. As illustrated in FIG. 11, the cuvette bank will have a base that can hold multiple parallel cuvette racks, and the base can be translated horizontally so that it can position any one of the racks next to the imaging position. The cuvette bank can accommodate 2-12 racks. The rack being measured will be translated to the imaging position sequentially to measure all the 8 cuvettes on the rack. The mechanical translations are realized with long range linear sliders, which can provide up to 210 mm stoke with up to 200mm / s speed and 3 mm precision. Individual cuvette racks have temperature control function to maintain the cuvette at 35-37°C for optimal mycobacteria growth. Mixing function is implemented to the cuvette bank using linear back and forth agitation driven by the sliders. The logistics of the mixing and positioning will be controlled by the system software. The system software will include detection optics, temperature and sample control hardware, system control, user-interface, and data analysis software (FIG. 12).

[0068] LVSim precision tracking of multiple phenotypic features of bacterial cells. Leveraging advanced machine learning and deep learning algorithms, LVSim enables precision tracking of multiple phenotypic features of bacterial cells, including growth dynamics, morphological changes, and movement patterns. Analyzing LVSim video data with machine learning and deep learning algorithms accurately quantifies cell number, spatial positioning, and intensity variations, allowing for high-resolution phenotypic characterization. Exemplary machine learning and deep learning algorithms useful for analyzing LVSim video data for phenotypic signatures include convolution neural network and long short-term memory network. Specific implementations of such algorithms to LVSim video data may be found in U.S. Patent No. 11,834,696; Iriya et al., Biosensors, 2024, 14(2): 89; Zhang et aJ, ACS Sensors, 2022, 7(8): 2262-2272; Iriya et al., IEEE Sensors Journal, 2020. 20(9): 4940-4950; and Yu et al., Anal. Chem., 2018, 90 (10): 6314-6322. Through extensive training and optimization, machine learning and deep learning algorithms extract LVSim-specific phenotypic signatures,enabling rapid, data-driven decision-making for predictive pDST without the need for defining and quantifying each feature. The signatures are bacterial growth tracking, morphology tracking, and motion tracking.(1) LVSim bacterial growth tracking: With single-cell resolution, LVSim technology enables real-time division tracking. In another application, this methodology achieved rapid pDST of Escherichia coli cultures within 60 min, significantly accelerating resistance detection.(2) LVSim morphology tracking: By quantifying total scattering intensity and its temporal variation, LVSim provides precise insights into bacterial size, shape, and physiological-associated variations. This intensity correlates with growth rate and antibiotic response, offering a powerful, label-free method for phenotypic resistance profiling.(3) LVSim motion tracking: Utilizing advanced particle motion analysis, LVSim accurately differentiates bacteria based on size, shape, and activity, enhancing the ability to distinguish drug-susceptible and drug-resistant phenotypes with unprecedented precision.

[0069] LVSim data analysis. Image processing of the LVSim recorded images consisted of four steps to remove the background noise and reveal the underlying single mycobacterial cell scattered intensity. 1) The raw images stack was averaged every four frames (1 min video, 2400 frames averaged to 600 frames) to enhance the signal-to-noise ratio. Motion drift induced blur was also reduced during this step. 2) Local minimum projection at pixel level was performed every 10 frames in a rolling manner (Frame 1 - 10, frame 2 - 11, etc.). The rolling local minimum images were then projected to a stack of images representing the temporal local minimum. Subtracting the after-averaged image by local minimum yielded a background drift and cuvette defects scattering induced noise-free images. 3) The after averaged and minimum subtracted image stack was then used to calculate the stack median by projecting the median intensity of each pixel in the whole stack. The projected local median image was then subtracted from the entire stack to remove dynamic noise caused by mechanical drift. 4) The remaining background was finally removed by eliminating the local background. The radius of background smoothing of 10 pixels was chosen to avoid any moving objects in the video being removed as background. These steps were compiled and automated in Imaged software.

[0070] LVSim detection of mycobacterial heteroresistance. Mycobacterium smegmatis me2155 (M. smegmatis KANS) was electrotransformed with the integrative pLJR962 plasmid (a gift from Dr. Sarah Fortune; Addgene plasmid #115162) to generate a recombinant M. smegmatis kanamycin resistant strain (M. smegmatis KANR). Mixed populations of M.smegmatis KANSand M. smegmatis KANRof 95%: 5% and 99%: 1%, respectively, were generated to simulate heteroresistance and were subsequently incubated with and without KAN (8 pg / mL). Each mixed culture was monitored by LVSim every 4 h for 32 h to detect growth and emergence of the M. smegmatis KANRresistant subpopulation.

[0071] LVSim mycobacterial heteroresistance validation. At each initial (Ti) and final time point (Tf) of the LVSim heteroresistance assay, mycobacterial cultures were serially diluted and plated onto M7H10 agar, supplemented with 50 pg / mL of kanamycin when necessary. Agar plates were sealed and incubated for 3 d. Viable colonies were enumerated, and CFU / ml was calculated. Susceptible and resistant mycobacterial subpopulations were calculated by subtracting the CFU count of drug-resistant mycobacteria (supplemented M7H10 agar + KAN50) from the CFU count of all viable bacteria (supplemented M7H10 agar) at Ti and Tf. All experiments were conducted in biological triplicate.

[0072] LVSim mycobacterial clump analyses. The tracking of mycobacterial spatiotemporal dynamics was achieved by ImageJ plug-in TrackMate. Each mycobacterial clump was detected with a thresholding filter with a defined intensity threshold. Next, spatiotemporal trajectory of individual bacterial clump was formed by connecting the clump in adjacent time frames with LAP tracker. Short tracks were filtered out to avoid false-tracking caused by particle-induced transient background variation. As LVSim can track the spatial movement of each clump within the recording time, and mycobacterial clumps were drifting freely inside the liquid medium, the resulting size profile of the mycobacterial clumps were highly scattered. The upper 10 percent of size values within the profile of each mycobacterial clump were averaged to represent the largest size of each clump captured within the recording period. The derived size from the size profile of each clump was then subjected to statistical analysis in Matlab. Mycobacterial clumps were grouped with bin size of 3.5 pm for later categorization.

[0073] Diarylthiazole-48 synthesis. For the synthesis of 2-bromo-l-(2-chloropyridin-4- yl)ethan-l-one, the starting material l-(2-chloropyridin-4-yl)ethan-l-one (1 equivalents, 0.039 mol, 6.06 g) was dissolved in acetic acid (AcOH) (100 mL) in a 250 mL round bottom flask equipped with a stir bar. Then 30% hydrobromic acid (HBr) in AcOH (8.36 mL, 0.047 mol, 1.2 equivalents) was added dropwise over 5 minutes at room temperature (RT) and a white precipitate formed during / after this addition. Elemental bromine (1.8 mL, 0.035 mol, 0.9 equivalents) was then added dropwise over 5 minutes at RT while stirring and the reaction turned orange with opaque precipitate. The reaction was then stirred overnight at RT and then filtered. The filtered solid was rinsed repeatedly with cold ether and allowed to dry fullyresulting in 2-bromo-l-(2-chloropyridin-4-yl)ethan-l-one as a light tan / off white solid in 86% yield (7.89 g).

[0074] For the synthesis of 4-(2-chloropyridin-4-yl)-2-(2-ethylpyridin-4-yl)thiazole, the starting material 2-bromo-l-(2-chloropyridin-4-yl)ethan-l-one (1) (1 equivalents, 0.039 mol, 8.04 g) and 2-ethylpyridine-4-carbothioamide (1 equivalents, 0.0343 mol, 5.70 g) was dissolved in ethanol (150 mL) in a 250 mL round bottom flask equipped with a stir bar and a condenser. The reaction was then heated to 135°C for 2 h then the heat was reduced to 90 °C for 3-5 h while stirring and an orange clump of precipitate formed in the center. The reaction was then cooled to RT and poured into a separatory funnel with saturated aqueous NaHCCh (250 mL) and extracted with ethyl acetate (100 mL x 3). The combined organic layers were washed with saturated NaCl solution, dried with Na2SO4, and excess solvent was removed using a rotary evaporator. The resulting crude solid was dry loaded onto silica in preparation for purification via flash column chromatography (9: 1 dichloromethane:acetone, Rf = 0.27) and isolated as a white solid.

[0075] For the synthesis of 2-(2-ethylpyridin-4-yl)-4-(2-methoxypyridin-4-yl)thiazole, the starting material 4-(2-chloropyridin-4-yl)-2-(2-ethylpyridin-4-yl)thiazole (1 equivalent, 9.94 mmol, 3.00 g) was dissolved in methanol (30 mL) in a 100 mL round bottom flask equipped with a stir bar. Then sodium hydroxide (NaOH) (30 equivalents, 0.3 mol, 11.93 g) was solvated in 25 mL of deionized water and the resulting aqueous NaOH solution was added to the reaction while stirring. The round bottom was then fitted with a condenser with a Teflon sleeve to prevent sticking, heated to 110°C, and stirred for 24 hours. The reaction was then cooled to RT, excess methanol was removed using a rotary evaporator and then extracted with ethyl acetate (25 mL x 2). The combined organic layers were then dried with sodium sulfate (Na- 2804) and dry loaded directly onto silica to be purified via flash column chromatography (4: 1 dichloromethane:acetonitrile) and isolated as a pale pink solid. The final product (>99% purity; liquid chromatography-mass spectrometry) was solubilized in dimethylsulfoxide (DMSO).

[0076] LVSim mycobacterial DST. Following adjustment, mycobacterial cultures were aseptically added to sterile semi-micro cuvettes (Millipore, USA) containing a single, UV- irradiated steel bead for imaging. Prior to imaging, cuvettes were removed from the LVSim platform and incubated at 37°C with aerobic agitation for 7 min at 100 rpm, followed by a 5- min static settlement period. Each cuvette was sequentially imaged and recorded for a total of 80 s. Drug inhibition threshold for LVSim-mediated DST was defined as <10% increase of scattering intensity and / or scattering count compared to the initial normalized scatteringintensity and / or scattering count values, respectively. The calculated drug inhibition threshold for LVSim normalized scattering intensity and scattering count measurements were 1.1 and 1.15, respectively. Drug susceptibility classification required the drug-treated samples must remain below the drug inhibition threshold and statistically significant divergences and differences in the normalized scattering intensity and / or scattering count of drug-treated and untreated samples.

[0077] LVSim mycobacterial DST validation. At each initial (Ti) and final time point (Tf) of the LVSim DST assay, mycobacterial cultures were serially diluted and plated onto M7H10 agar. Agar plates were sealed and incubated for 3 d (M. smegmatis, M. abscessus) or 14-21 d (M. tuberculosis). Viable colonies were enumerated, and CFU / mL was calculated. All experiments were conducted in biological triplicate.

[0078] RNA isolation and quantitative reverse transcriptase PCR (qRT-PCR). M. abscessus cultures were grown aerobically to mid-logarithmic phase (ODeoo 0.6-0.8) prior to being diluted to an ODeoo of 0.1. Diluted cultures were exposed to clarithromycin (CLR) (8 pg / mL) or rifampin (RIF) (2 pg / mL). Cultures were harvested at 0 h, 12 h, and 72 h by centrifugation at 3000 x g for 10 min at 4°C. The supernatant was discarded, and the cell pellet was resuspended in 1 mL TRIzol reagent (Invitrogen, Waltham, MA, USA). The resuspended pellets were transferred to 2 mL screw cap tubes containing -800 mg of zirconia beads. Cells were disrupted thrice by bead beating (BioSpec Products, Bartlesville, OK, USA). The resulting lysates were incubated at room temperature for 5 min and centrifuged at 13,000 x g for 1 min. The resulting aqueous phase was transferred to a microcentrifuge tube containing chloroform (200 pL) and samples were vortexed for 15 s followed by 5 min incubation at 4°C and centrifugation at 13,000 x g for 15 min at 4°C. The upper, aqueous phase was transferred to a new microcentrifuge tube containing isopropanol (500 pL), and RNA was precipitated overnight at 4°C. The resulting precipitated RNA was pelleted by centrifugation at 13,000 x g for 15 min at 4°C and washed twice with 70% ethanol. After evaporation of residual ethanol, the isolated RNA was resuspended in 100 pL nuclease-free H2O. Total RNA (30 pg) was treated with DNase I (New England Biolabs, USA) for 20 min at 37°C to degrade contaminating genomic DNA. RNA samples were purified using the RNeasy Mini Kit (Qiagen, Hilden, Germany) and eluted in 50 pL nuclease-free H2O. RNA concentrations were quantified by Nanodrop (Thermo Scientific, Waltham, MA, USA), and quality was assessed by agarose gel electrophoresis. Cleaned RNA samples were reverse transcribed into cDNA using the iScript cDNA Synthesis Kit (BioRad, Hercules, CA, USA), according to the manufacturer’sinstructions. Primer efficiency was validated against 10-fold dilution standard curves with acceptable amplification efficiencies of 90-110% and determination coefficient of >0.99. Relative gene expression was calculated using the 2'AACtmethod, using the M. abscessus sigA gene as reference.Table 1. qRT-PCR primers used for measurement of AL abscessus and arr expression.

[0079] Statistical analysis. Statistical analyses were performed using GraphPad Prism 9 (GraphPad Software, San Diego, CA, USA), and p values of <0.05 were considered statistically significant. Unpaired t-tests were used to assess bacterial growth differences in LVSim validation experiments.

[0080] Bacterial media and culture conditions. All mycobacteria were cultured in Middlebrook 7H9 medium (Becton Dickinson and Company Limited, USA) supplemented with 0.05% Tyloxapol, 0.5% glycerol, and 10% ADC (bovine serum albumin, dextrose, catalase) (hereafter referred to as M7H9 broth) and grown on Middlebrook 7H10 agar (Becton Dickson and Company Limited, USA) supplemented with 0.5% glycerol and 10% ADC (hereafter referred to as M7H10 agar). Cultures were grown to mid-logarithmic phase (ODeoo = 0.6-0.8), washed, resuspended in media, and diluted 1000-fold. Following, cultures were further adjusted to yield -500-1000 particles per field-of-view. Kanamycin (KAN) was added to the growth medium when required at a concentration of 8 pg / mL. All media and antibiotics were filtered using a 0.22 pM filter. All liquid cultures were grown at 37°C with aerobic agitation.Example 2. Rapid phenotypic detection of a 1% heteroresistant M. smegmatis subpopulation

[0081] Detection of mycobacterial heteroresistance at the clinically relevant 1% resistance level remains difficult for current phenotypic and genotypic methods and is a primary contributor to the emergence of antibiotic resistance through incorrect or insufficient antibiotic treatment. To combat the rise in antibiotic resistance, we leveraged our LVSim platform for the detection of a 1% drug-resistant M. smegmatis subpopulation, by way of rapidly and accurately determining intra-population drug heteroresistance. LVSim achieves direct visualization of IxlO3cells and single particle-tracking of multiple phenotypic features simultaneously to distinguish between actively growing (FIG. 1A, red cells) and growth- inhibited (FIG. 1A, cyan cells) bacteria. Moreover, LVSim-SPT allows for targeted tracking and analysis of individual particles and does not rely on time intensive ensembled analysis (FIG. IB). The method of classifying growing versus non-growing bacterial cells with LVSim- SPT is based on the intensity of single particles. Although LVSim has a low spatial resolution limit (12 pm, ~4 mycobacterial cells), particles at or below this spatial resolution threshold will appear as spots of similar size (FIG. 1C). However, changes in particle intensity correlate directly with increased number of bacteria, even below the spatial resolution threshold, and positive changes in intensity over time directly correlate with bacterial growth (FIG. 1C). Further, subdivision and binning of particles based on intensity, rather than size, and tracking of intensity changes of individual particles simultaneously over time allows for accurate and rapid classification of growing versus non-growing bacteria. To demonstrate the utility, accuracy, and rapidity of our novel approach and methodology, the LVSim-SPT classification principles was applied to a 1% kanamycin heteroresistant population of M. smegmatis and achieved accurate detection of a 1% kanamycin resistant AL smegmatis subpopulation within 15 hours (five replication cycles), a significant improvement from currently employed phenotypic detection methods and surpassing genotypic methods in accuracy.

[0082] Specifically, the disclosed method with the adapted LVSim-SPT platform successfully identified <1% kanamycin-resistant mycobacterial subpopulations within the theoretical detection limit of just five division cycles, 15 hours for AL smegmatis (FIGs. 2A and 2B) and 108 hours for AL tuberculosis (FIGs. 3 A and 3B). This represents a significant advancement in phenotypic diagnostics, enabling rapid, sequencing-free detection of heteroresistance at clinically actionable thresholds. LVSim bridges the sensitivity of molecular techniques with the scalability of phenotypic assays, providing a powerful tool for early resistance detection, personalized antimicrobial therapy, and antimicrobial stewardship.

[0083] The LVSim-SPT platform is adapted from the two-channel continuous monitoring platform for urinary tract infection diagnosis. A stepping motor carrying a sample tray that can withhold up to eight cuvettes is applied for linearly shift sample with various antibiotic treatments (e.g., type or dosage) for susceptibility or resistance detection or minimum inhibitory concentration (MIC) determination. To facilitate the detection of heteroresistant subpopulation in a major susceptible background, a 100% susceptible M. smegmatis strain population and mixture of 1% M. smegmatis resistant strain:99% M. smegmatis susceptible strain populations, each with three biological replicates, are prepared and treated with kanamycin, a second-line antibiotic for TB treatment. All six samples are measured in the LVSim platform for up to 2 days and LVSim images is continuously recorded at 10 frames per second while stepping motor transitions the cuvettes sequentially. Each sample condition is recorded for one minute at 1.5 hours interval. LVSim images for each sample condition at each time point is then sectioned to 600 LVSim images stacks and processed to remove background noises via customized algorithms. Individual bacterial temporal intensity profiles are extracted by performing single particle tracking using Trackmate, an Imaged plugin function, from the processed LVSim image stacks. The detection of resistant subpopulation at therapeutic failure limit (1%) is extremely challenging, as for a starting count of 400 bacterial particles in LVSim field of view, < 4 of the observed particles are potentially heteroresistant cells. To enable the sensitive capture of these minor subgroups, leveraging the growth difference between susceptible and resistant mycobacterial strains in the presence of an antibiotic is crucial. As only resistant cells can survive and grow during the antibiotic treatment, growth induced particle intensity change should only be observed for resistant cells. Thus, a specific size threshold was utilized for bacterial cell detection and consequent tracking. Such diameter threshold should not be too small, which leads to indistinguishable particles as the area of interest only focus on the central brightest area of each particle. Threshold should not be too large, which also results indiscernible particles and the included background neutralizes the difference originated from a non-growing / growing cell. Thus, a diameter of 7 pixels is determined for bacterial cell detection. For actively growing mycobacterial cells, it would overfill the defined area over time, leading to increasing detected single particle intensity. While for growth-inhibited mycobacterial cells, single particle intensity would stay relatively consistent over time with slight changes or fluctuations or exhibit decreasing trend due to the death / disintegration of mycobacterial cell clusters. As all mycobacterial cells / clusters are freely drifting in liquid medium as a result of Brownian motion and liquid convection, in LVSimwhich adopts forward scattering as illumination pathway, the scattered intensity of individual mycobacterial cells varies corresponding to the cellular movements and rotations. To better capture the representative size of each mycobacterial cell, the top 10% intensity values within the extracted intensity profile of each cell is averaged and repeated for all cells within the image stack. As such, we apply a histogram to illustrate the single particle intensity distribution for each image stack at different conditions and time points. To achieve rapid detection of the 1% resistant subpopulation, each bin within the histogram is inspected over time as to pinpoint the range of single particle intensity which best characterizes the growing resistant populations. However, the initial histogram is laid out using manually determined, fixed bin size and bin number, which could miss the most sensitive intensity range originating from active mycobacterial growth, we establish a hill-climbing machine learning algorithm to exploit the intensity distribution through all conditions and time points to identify the most sensitive intensity range for rapid heterogeneity detection.Example 3. Tracking of spatiotemporal clumping dynamics in L VSim.

[0084] Cellular clumping of pathogenic mycobacterial species is a phenomenon associated with increased virulence but remains unused in the clinical and diagnostic detection of mycobacterial drug susceptibility. Traditional agar proportion methods for mycobacterial drug susceptibility neglects spatiotemporal clumping dynamics, as colony forming units (CFUs) are enumerated at a predetermined time point, irrespective of initial clump size or final CFU size. Using LVSim and its dark-field microscope-like nature, we leveraged a liquid culture approach to provide a dynamic growth environment, allowing mycobacterial clumps to naturally form and grow spontaneously and enabling sensitive tracking of clump-forming dynamics over time (FIG. 4A). In the absence of antibiotic treatment, mycobacteria naturally and spontaneously aggregate after dividing and form clumps of varying sizes due to their hydrophobic surfaces (FIG. 4B). To precisely determine mycobacterial clump sizes, we defined an intensity threshold and performed binary segmentation. Pixels with intensity values greater than the threshold were assigned a value of “1” as foreground compared to background values of “0”. Mycobacterial clumps were segmented by connecting the boundary region with the sharpest value decrease from 1 to 0. Mycobacterial clump sizes were determined after forming the contour of each clump to prevent miscounting single large mycobacterial clumps as several cells or clumps when using a size threshold of a fixed size. To analyze clumping dynamics and clumping size differentials, we first grouped the mycobacterial clumps with a bin size of 3.5 pm, the averagesize of a M. tuberculosis cell (FIG. 1C). By applying a binary segmentation classification system, we further partitioned the mycobacterial clumps into three classes: small (< 24.5 pm), medium (24.5 < 49 pm), and large (> 49 pm diameter) clumps (FIG. 4A). To improve representation of the clump size distributions, we plotted the clump size histograms and applied Gaussian curve fitting to the resultant distribution. As shown in FIGs. 4B and 4C, in small-, medium- and large-clump regions, each class in each condition exhibited different trends over the 40-hour testing time. In control conditions, a universal increase was observed in small, medium, and large clump regions, whereas addition of the FiFo-ATP synthase inhibitor bedaquiline largely inhibited formation of medium and large clumps over the observed time span (FIG. 4C). Thus, quantifiable, drug-responsive mycobacterial clumping behavior and dynamics is a new phenotypic feature to perform DST.Example 4. Leveraging TB clumps as a novel phenotypical DST feature

[0085] To incorporate clumping dynamics as a phenotypic feature, we first investigated the count change of each clump tercile over time with different antibiotics. As shown in Figure 5 A (left), after 40 h of tuberculosis growth in liquid culture, clump counts universally increased among all the terciles. However, when M. tuberculosis is treated with an antibiotic (e.g., bedaquiline) (FIG. 5 A, right), counts for both medium and large clump terciles drastically and significantly decreased, while the count for small clumps increased significantly. Formation of medium and large mycobacterial clumps is largely suppressed in the presence of antibiotics. In the presence of antibiotic, increased numbers of small tercile clumps potentially originate from the disintegration of existing medium and large clumps or M. tuberculosis inhibition or death, leading to antibiotic-induced malformations and / or mycobacterial debris. To further analyze differential changes, the subpopulation constitution, i.e., the percentage of each clump tercile among the whole population at given time for M. tuberculosis cultures with and without antibiotic exposure, was analyzed (FIG. 5B). For instance, with normal growth, the total number of small M. tuberculosis clumps increased over time, yet their relative abundance within the population decreased (FIGs. 5B and 5C). This phenomenon can be attributed to the concomitant formation of medium and large clumps during growth. The inverse trend was observed during M. tuberculosis antibiotic exposure, where the relative abundance of small clumps increased over time, while the relative abundance of medium and large clumps decreased. Fold-change analysis of the subpopulation percentages (FIG. 5C) revealed that antibiotic exposure leads to inversely proportional dynamics for all M. tuberculosis clumpterciles (FIG. 4D). Closer analysis of the dynamics and formation of large clumps showed considerable variation in count over time, while medium clump counts increased more consistently (FIG. 4E). The variation detected in the large clump count may originate from the significant difference in large to medium clump abundance (FIG. 4A, left, medium- and large- clumps region). While the large clump count surpassed the background noise level, the sole use of large clumps as a diagnostic marker for antibiotic susceptibility, or lack thereof, is not optimal due to their limited abundance. By plotting the spatiotemporal count dynamics of the medium and large clump terciles, antibiotic-treated, susceptible M. tuberculosis can clearly and rapidly be distinguished from an untreated M. tuberculosis control in 6-18 hours (FIG. 4E). Thus, the count change of medium and large clumps is a new phenotypic feature for M. tuberculosis DST, and this example demonstrates the promising potential of utilizing mycobacterial clumping dynamics as a novel phenotypic feature for rapid mycobacterial DST.Example 5. Mycobacterial clumping dynamics for M. abscessus Rough DST

[0086] Similar to M. tuberculosis, M. abscessus Rough morphological cells readily aggregate after dividing and clump together due to their hydrophobic surfaces. Therefore, we analyzed the count change of medium and large clumps as a new phenotypic feature for M. abscessus Rough DST with the experimental compound DAT-48. As shown in FIG. 6A, medium clump counts allowed for rapid, definitive susceptibility determination of AL abscessus Rough against DAT -48 within 3 h (1 replication cycle). Large clump counts also allowed for M. abscessus Rough DST with DAT-48 within 6-8 h (2 replication cycles) (FIG. 6B). Herewith we show that LVSim mycobacteria clumping dynamics serves as a supremely reliable phenotypic feature for rapid susceptibility determination.Example 6. LVSim phenotypic DST determined M. tuberculosis RIF, BDQ, and DAT-48 susceptibility results in 18 h, 8 h, and 44 h, respectively.

[0087] To establish LVSim as a rapid phenotypic DST platform, we tested M. tuberculosis H37Ra against rifampin (RIF; 0.125 pg / mL), bedaquiline (BDQ; 0.125 pg / mL), and the investigational agent DAT-48 (8 pg / mL) at 2* MIC concentrations and validated M. tuberculosis susceptibilities using gold-standard plating (FIG. 6). LVSim scattering intensity rapidly identified susceptibility to RIF within 16 h, a 28-fold reduction in time to result (TTR) compared to conventional plating (FIGs. 7A and 7B). For BDQ, LVSim scattering intensities for control and BDQ-exposed M. tuberculosis cultures diverged as early as 8 h post-treatmentand maintained separation throughout the 48-h observation period (FIG. 7C), enabling a 63- fold TTR reduction relative to standard DST. BDQ-exposed M. tuberculosis scattering intensity remained consistently below the inhibition threshold, indicating robust susceptibility determination under LVSim conditions (FIG. 7C). With the experimental compound DAT-48, LVSim scattering intensity enabled susceptibility classification within 44 h, an 11-fold TTR reduction (FIG. 7E). Although scattering intensity values exceeded the inhibition threshold (FIG. 7E), subsequent M7H10 agar plating confirmed susceptibility (FIG. 7F), suggesting DAT -48 induces growth inhibition and increases particle intensity within the LVSim detection window. Overall, LVSim reliably determined phenotypic susceptibility profiles for first-line, Group A, and experimental TB drugs within <48 h (Table 2), substantially reducing TTR compared to gold-standard methods. These results establish LVSim as a rapid and accurate tool for phenotypic DST across diverse drug classes.aSuccessful determinations are indicated by Y (Y es) or N (No) for standard pDST, LVSim scattering intensity, and LVSim scattering count.bpDST, phenotypic drug susceptibility testing. Expected results are indicated by S (Susceptible) or R (Resistant).cTTR, time to result.Example 7. LVSim M. abscessus Smooth morphotype phenotypic DST results in less than 360 minutes and phenotypic resistance results in 720 minutes.

[0088] Using LVSim, we achieved rapid phenotypic susceptibility testing of M. abscessusSmooth morphotype against clarithromycin (CLR; 16 pg / mL) and bedaquiline (BDQ; 0.5pg / mL) at 2 / MIC concentrations and validated M. abscessus Smooth morphotype susceptibilities using gold-standard plating (FIGs. 7A-7F). LVSim scattering intensity based susceptibility to CLR was determined within <360 min and to BDQ within <90 min, corresponding to 12-fold and 48-fold reductions in time to result (TTR), respectively, compared to gold-standard DST (FIG. 7A, B; Table 3). In contrast, normalized M. abscessus Smooth morphotype scattering count values for CLR- and BDQ-treated cultures remained near or below the inhibition threshold and failed to discriminate susceptibility profiles (FIG. 7A, B). To demonstrate the fidelity of LVSim in detecting concentration-dependent susceptibility and resistance, we subjected M. abscessus Smooth morphotype to rifampin (RIF), to which M. abscessus is susceptible at high concentration (256 pg / mL) but resistant at low concentration (2 pg / mL), and diarylthiazole-48 (DAT-48), a pre-clinical anti-TB drug. At high RIF concentration, LVSim intensity and count measurements for control and RIF -treated cultures diverged and maintained separation as early as 180 min (Fig. 7C; Table 3); RIF -treated cultures remained stagnant throughout the observation period, corresponding to a 24-fold reduction in TTR, and susceptibility was validated by gold-standard plating. Low RIF concentration similarly yielded divergence between control and RIF-treated cultures; however, consistent LVSim scattering intensity and count increases were observed for both groups throughout the 12 h observation period, and resistance was validated by gold-standard plating (FIG. 7D). LVSim further determined a lack of activity of DAT-48 (128 pg / mL) againstabscessus Smooth morphotype by recording overlapping LVSim intensity and count metrics, with resistance confirmed by gold-standard plating (FIG. 7E). Together, these results highlight the superiority of scattering intensity over count metrics for rapid phenotypic DST of AL abscessus Smooth under LVSim conditions and showcase the fidelity of LVSim-based DST.Table 3. Summary of AL abscessus ATCC19977 Smooth morphotype LVSim phenotypicDST to CLR, BDQ, RIF, and DAT-48,aSuccessful determinations are indicated by Y (Y es) or N (No) for standard pDST, LVSim scattering intensity, and LVSim scattering count.bpDST, phenotypic drug susceptibility testing. Expected results are indicated by S (Susceptible) or R (Resistant).cTTR, time to result.References

[0089] The teachings of the following publications are incorporated herein in their entirety by this reference.• Tao et al., “Antibiotic susceptibility testing with large-volume light scattering imaging and deep learning video microscopy,” US patent 11,834,696.• Iriya et al., “Deep Learning-Based Culture-Free Bacteria Detection in Urine Using Large- Volume Microscopy,” Biosensors, 2024, 14(2): 89.• Zhang et al., “Rapid Detection of Urinary Tract Infection in 10 Minutes by Tracking Multiple Phenotypic Features in a 30-Second Large Volume Scattering Video of Urine Microscopy,” ACS Sensors, 2022, 7(8): 2262-2272.• Iriya et al., “Rapid antibiotic susceptibility testing based on bacterial motion patterns with long short-term memory neural networks,” IEEE Sensors Journal, 2020, 20(9): 4940-4950.• Hui Yu, Wenwen Jing, Rafael Iriya, Yunze Yang, Karan Syal, Manni Mo, Thomas E Grys, Shelley E Haydel, Shaopeng Wang, Nongjian Tao, Phenotypic antimicrobial susceptibility testing with deep learning video microscopy, Anal. Chem., 2018, 90 (10), pp 6314-6322, DOI: 10.1021 / acs.analchem.8b01128

Claims

CLAIMSWe Claim:

1. A method of identifying drug resistance in a population of mycobacteria, the method comprising: providing a culture comprising the population of mycobacteria, wherein the culture is split into a test sample and a control sample; administering an antimicrobial drug to the test sample; evaluating the growth of the population of mycobacteria after the antimicrobial drug is administered, wherein the step comprises: recording images of the culture through a large volume scattering imaging (LVSim) platform for a period of at least 1-10 seconds at regular intervals for at least two doubling times of the population of mycobacteria; and tracking median intensity of selected pixels from the recorded images, wherein the selected pixels depict growth of a single mycobacterium; and detecting drug resistance within the population of mycobacteria in response to exposure to the antimicrobial drug upon the selected pixels depicting: small cell aggregates instead of a single cell; or medium-sized and / or large-sized mycobacterial clumps.

2. The method of claim 1, wherein the culture is split into a plurality of test samples and a control sample, different concentrations of the antimicrobial drug is administered to each test sample prior to evaluating the growth of the population.

3. The method of claim 1, wherein the provided culture comprises a concentration of mycobacteria resulting in 10-5,000 particles per field-of-view from the LVSim platform.

4. The method of any one of claims 1-3, wherein the selected pixels depict small cell aggregates of diameters >7 pm and the small cell aggregates comprise viable cells, the population of mycobacteria exhibits heteroresistance.

5. The method of any one of claims 1-3, wherein the small cell aggregates comprise actively dividing and drug-resistant cells, the population of mycobacteria exhibits heteroresistance.

6. The method of any one of claims 1-3, wherein the selected pixels depict medium-sized mycobacterial clumps have diameters of 24.5 < 49 pm and / or large-sized mycobacterial clumps have diameters of > 49 pm.

7. The method of any one of claims 1-3, wherein the population of mycobacteria comprises at least one species selected from the group consisting of: Mycobacterium tuberculosis, Mycobacterium abscessus, Mycobacterium avium, Mycobacterium intracellulare, Mycobacterium kansasii, and Mycobacterium smegmatis.

8. The method of any one of claims 1-3, wherein the step of tracking median intensity of selected pixels from the recorded images comprises removing background noise from the recorded images thereby revealing an underlying single mycobacterial cell or small cell aggregate scattered intensity.

9. The method of claim 8, wherein the step of removing background noise from the recorded images comprises: averaging intensity of pixels from a plurality of recorded images to produce an afteraveraged image stack; producing a local minimum projection from a plurality of recorded images at a pixel level to produce a local minimum projection image stack, wherein the local minimum projection image stack is a stack of images representing a temporal local minimum; producing a noise-free image stack subtracting after-averaged image stack by the temporal local minimum to produce a minimum subtracted image stack; calculating the stack median from the after-averaged image stack and the minimum subtracted image stack; and subtracting the stack median from the recorded images to remove background noise caused by mechanical drift.

10. The method of claim 9, further comprising removing local background noise from the recorded images after background noise caused by mechanical drift is removed.

11. The method of 10, wherein the local background noise is measured from 10 pixels.

12. The method of claim 9, wherein the plurality of recorded images for producing the after-averaged image stack is four images from adjacent time frames and then plurality of recorded images for producing the local minimum projection image stack is 10 images from adjacent time frames.

13. A method of identifying drug susceptibility in a population of mycobacteria, the method comprising: providing a culture comprising the population of mycobacteria, wherein the culture is split into a test sample and a control sample; and evaluating the growth of the population of mycobacteria after an antimicrobial drug or a combination of antimicrobial drugs is administered to the test sample, wherein the step comprises: recording images of the culture through a large volume scattering imaging (LVSim) platform for a period of at least 1-10 seconds at regular intervals for at least two doubling times of the population of mycobacteria to produce a set of recorded images depicting the test sample and the control sample; and tracking scattering intensity from the set of recorded images; and detecting the test sample as being inhibited by the antimicrobial drug or the combination of antimicrobial drugs upon the set of recorded images depicting the test sample displaying decreased scattering intensity compared to the set of recorded images depicting the control sample.

14. The method of claim 13, wherein the step of detecting the test sample as being inhibited by the antimicrobial drug further comprises detecting the scattering intensity of the set of recorded figures is at or below a species-specific threshold.

15. The method of claim 13 or 14, further comprising: calculating an initial normalized scattering count value from the set of recorded images depicting the control sample; and tracking scattering count from the set of recorded images depicting the test sample; wherein the step of detecting the test sample as being inhibited by the antimicrobial drug or the combination of antimicrobial drugs upon the set of recorded images depicting thetest sample having decreased scattering count compared to initial normalized scattering count value.

16. The method of claim 15, wherein the step of detecting the test sample as being inhibited by the antimicrobial drug further comprises detecting the scattering count of the set of recorded figures depicting the test sample is at or below a species-specific threshold.

17. A method of identifying drug susceptibility in a population of mycobacteria, the method comprising: providing a culture comprising the population of mycobacteria, wherein the culture is split into a test sample and a control sample; and evaluating the growth of the population of mycobacteria after an antimicrobial drug or a combination of antimicrobial drugs is administered to culture, wherein the step comprises: recording images of the culture through the LVSim platform for a period of at least 1-10 seconds at regular intervals for at least two doubling times of the population of mycobacteria to produce a set of recorded images depicting the test sample and the control sample; and tracking scattering count from the set of recorded images, wherein an initial normalized scattering count value is calculated from the set of recorded images depicting the control sample; and detecting the test sample as being inhibited by the antimicrobial drug or combination of antimicrobial drugs upon the set of recorded images depicting the test sample having decreased scattering counts compared to the initial normalized scattering count value.

18. The method of claim 17, wherein the step of detecting the test sample as being inhibited by the antimicrobial drug or combination of antimicrobial drugs further comprises detecting the scattering count of the second set of recorded figures depicting the test sample is at or below a species-specific threshold.

19. The method of claim 17 or 18, further comprising: calculating an initial normalized scattering intensity value from the set of recorded images depicting the control sample; and tracking scattering intensity from the set of recorded images depicting the test sample;wherein the step of detecting the test as being inhibited by the antimicrobial drug or the combination of antimicrobial drugs further comprises detecting the set of recorded images depicting the test sample displaying decreased scattering intensity compared to the set of recorded images depicting the control sample.

20. The method of claim 19, wherein the step of detecting the test sample as being inhibited by the antimicrobial drug or combination of antimicrobial drugs further comprises detecting the scattering intensity of the set of recorded figures depicting the test sample is at or below a species-specific threshold and displaying decreased scattering count than the control sample.

21. The method of any one of claims 13, 14, 17, and 18, wherein the provided culture comprises a concentration of mycobacteria resulting in 10-5,000 particles per field-of-view from the LVSim platform.

22. The method of any one of claims 13, 14, 17, and 18, wherein the population of mycobacteria comprises at least one species selected from the group consisting of: Mycobacterium tuberculosis, Mycobacterium abscessus, Mycobacterium avium, Mycobacterium intr acellular e, Mycobacterium kansasii, and Mycobacterium smegmatis.

23. A large volume scattering imaging (LVSim) platform comprising: a camera with a lens assembly; a linear translational stage comprising a motorized belt driven sample tray and a thermal control unit; a beam-block; and a lighting unit; wherein: the camera, beam-block, lighting unit are placed in a line and the beam-block is placed in front of the light unit thereby preventing direct illumination of the sample tray and the camera; the translational stage linearly shifts samples in the sample tray; and the thermal control unit is beneath the sample tray and provides precision temperature control to create an appropriate growth environment for mycobacterial cells.

24. A method of testing drug susceptibility in a population of mycobacteria, the method comprising: providing a culture comprising the population of mycobacteria, wherein the culture is split into a test sample and a control sample; administering an antimicrobial drug to the test sample; recording images of the culture through a large volume scattering imaging (LVSim) platform for a period of at least 1-10 seconds at regular intervals for at least two doubling times of the population of mycobacteria; and counting medium-sized and / or large-sized mycobacterial clumps from the recorded images.

25. The method of claim 24, wherein medium-sized mycobacterial clumps have diameters of 24.5 < 49 pm and / or large-sized mycobacterial clumps have diameters of > 49 pm.

26. The method of claim 24, further comprising detecting the population of mycobacteria is susceptible to the antimicrobial drug upon the recorded images depicting the test sample having suppressed number of medium-sized and / or large-sized mycobacterial clumps compared to the recorded images depicting the control sample.

27. The method of any one of claims 24-26, wherein the culture is split into a plurality of test samples and a control sample, different concentrations of the antimicrobial drug are administered to each test sample prior to evaluating the growth of the population.

28. The method of any one of claims 24-26, wherein the population of mycobacteria comprises at least one species selected from the group consisting of: Mycobacterium tuberculosis, Mycobacterium abscessus, Mycobacterium avium, Mycobacterium intracellulare, Mycobacterium kansasii, and Mycobacterium smegmatis.

29. The method of any one of claims 24-26, wherein the provided culture comprises a concentration of mycobacteria resulting in 10-5,000 particles per field-of-view from the LVSim platform.

30. The method of any one of claims 24-26, wherein medium-sized and / or large-sized mycobacterial clumps are counted from the recorded images by tracking median intensity of selected pixels from the recorded images, wherein the selected pixels depict growth of a single mycobacterium.

31. The method of claim 30, wherein the step of tracking median intensity of selected pixels from the recorded images comprises removing background noise from the recorded images thereby revealing an underlying single mycobacterial cell or small cell aggregate scattered intensity.

32. The method of claim 31, wherein the step of removing background noise from the recorded images comprises: averaging intensity of pixels from a plurality of recorded images to produce an afteraveraged image stack; producing a local minimum projection from a plurality of recorded images at a pixel level to produce a local minimum projection image stack, wherein the local minimum projection image stack is a stack of images representing a temporal local minimum; producing a noise-free image stack subtracting after-averaged image stack by the temporal local minimum to produce a minimum subtracted image stack; calculating the stack median from the after-averaged image stack and the minimum subtracted image stack; and subtracting the stack median from the recorded images to remove background noise caused by mechanical drift.

33. The method of claim 32, further comprising removing local background noise from the recorded images after background noise caused by mechanical drift is removed.

34. A method compri sing : providing a culture comprising the population of mycobacteria, wherein the culture is split into a test sample and a control sample; obtaining a set of recorded images of the population of mycobacteria using the LVSim platform over a period of observation time after the test sample is administered an antimicrobialdrug or a combination of antimicrobial drugs, wherein the set of recorded images comprises a set of images depicting the test sample and a set of images depicting the control sample; processing the set of recorded images to extract the measurements related to cell growth, cell movement, cell morphology, intensity, and aggregation from the set of images depicting the test sample and the set of images depicting the control sample; predicting the drug susceptibility classification of the population of mycobacteria by providing the measurements related to cell growth, cell movement, cell morphology, intensity, and aggregation to one or more machine learning models as input, wherein the one or more machine learning models are generated by a training data set comprising measurements related to cell growth, cell movement, cell morphology, intensity, and aggregation of a different population of mycobacteria being inhibited by an antimicrobial drug and measurements related to cell growth, cell movement, cell morphology, intensity, and aggregation of a population of mycobacteria that has not administered an antimicrobial drug; and outputting the drug susceptibility classification.

35. The method of claim 34, wherein one or more machine learning models are selected from convolution neural network and long short-term memory network.

36. The method of claim 33 or 34, wherein the outputted drug susceptibility classification is selected from susceptibility to the antimicrobial drug or combination of antimicrobial drugs, resistance to the antimicrobial drug or combination of antimicrobial drugs, intermediate category between susceptibility and resistance to the antimicrobial drug or combination of antimicrobial drugs, and heteroresistance to the antimicrobial drug or combination of antimicrobial drugs.

37. The method of claim 33 or 34, wherein the period of observation time is at least two doubling times of the population of mycobacteria.

Images