AI-based cell classification method and system

Abstract

The present disclosure provides systems and methods for AI-based cell classification of cells from a patient sample to determine whether cells indicative of lung cancer are present. The systems and methods use 2D imaging to exclude cells unlikely to be indicative of lung cancer from subsequent 3D imaging, while 3D imaging is performed on cells likely to be indicative of lung cancer. The present disclosure further provides methods for training a 2D cell classifier for use in the systems and methods for AI-based cell classification.

Description

[Technical Field]

[0001] The present disclosure relates to systems and methods for artificial intelligence (AI)-based 2D classification of cells prior to 3D analysis to detect cells indicative of cancer or cells with characteristics of normal bronchial epithelial cells (BECs), and methods for training 2D cell classifiers for use in AI-based cell classification systems and methods. [Background technology]

[0002] Lung cancer is the deadliest cancer in the United States, with over 31 million patients at high risk of developing lung cancer. Early detection is the most reliable means of reducing lung cancer deaths, but many detection methods have low sensitivity and specificity, leading to missed diagnoses and resulting in higher mortality rates in addition to increased costs and unnecessary suffering from invasive procedures. Summary of the Invention [Means for solving the problem]

[0003] The present disclosure provides an artificial intelligence (AI)-based cell classification method, including: a) an optical tomography system generating representative 2D images of cells from a patient sample containing a plurality of cells; b) evaluating the representative 2D images with an abnormal cell 2D classifier to determine whether the cells have abnormal characteristics; c) evaluating the representative 2D images with a BEC 2D classifier to determine whether the cells have BEC-like characteristics; and d) if the cells are determined to have abnormal characteristics or BEC-like characteristics, the optical tomography system generating a 3D image of the cells.

[0004] The method has the following additional features: If the cell is determined to have no abnormal characteristics and no BEC-like characteristics, a 3D image of the cell is not generated. e) repeating steps a) through d) for a subset of cells within the plurality of cells to generate patient sample data reflecting a total number of enumerated analyzed cells having abnormal characteristics, a total number of enumerated analyzed cells having BEC-like characteristics, or a total number of enumerated analyzed cells; f) comparing the total number of enumerated analyzed cells having BEC-like characteristics to a threshold to determine whether the total number of enumerated analyzed cells having BEC-like characteristics is equal to or greater than the threshold and therefore sufficient for an accurate assay, or whether the total number of enumerated analyzed cells having BEC-like characteristics is less than the threshold and therefore insufficient for an accurate assay; repeating steps a) through f) until the total number of enumerated analyzed cells having BEC-like characteristics is sufficient to detect the presence of cells indicative of lung cancer in the patient sample with a preselected accuracy; the threshold being preselected by the number of enumerated cells having BEC-like characteristics that are statistically capable of detecting the presence of cells indicative of lung cancer in the patient sample with a preselected accuracy; the threshold is within the range of 250 to 2500 enumerated cells with BEC-like features, 400 to 1900 enumerated cells with BEC-like features, 600 to 1800 enumerated cells with BEC-like features, 800 to 1700 enumerated cells with BEC-like features, 1000 to 1600 enumerated cells with BEC-like features, 1200 to 1600 enumerated cells with BEC-like features, or 1300 to 1500 enumerated cells with BEC-like features; the threshold is at least 250 enumerated cells with BEC-like characteristics, at least 400 enumerated cells with BEC-like characteristics, at least 500 enumerated cells with BEC-like characteristics, at least 600 enumerated cells with BEC-like characteristics, at least 700 enumerated cells with BEC-like characteristics, at least 800 enumerated cells with BEC-like characteristics, at least 900 enumerated cells with BEC-like characteristics, at least 1000 enumerated cells with BEC-like characteristics, at least 1100 enumerated cells with BEC-like characteristics, at least 1200 enumerated cells with BEC-like characteristics, at least 1300 enumerated cells with BEC-like characteristics, at least 1400 enumerated cells with BEC-like characteristics, at least 1500 enumerated cells with BEC-like characteristics, at least 2000 enumerated cells with BEC-like characteristics, or at least 2500 enumerated cells with BEC-like characteristics; the method has a preselected 2D abnormal cell sensitivity of at least 90%; The method includes: having a preselected 2D abnormal cell specificity value of at least 65%; the method has a preselected 2D BEC sensitivity value of at least 85%; The patient sample is obtained from a sputum specimen; the patient sample is obtained by isolating and preserving a plurality of cells from a sputum specimen, the patient sample including abnormal cells, BECs, squamous epithelial cells, monocytes, lymphocytes, polymorphonuclear leukocytes, other white blood cells, debris, cell fragments, cell clusters, and any combination thereof; the abnormal cells are selected from the group consisting of cells exhibiting atypia, dysplastic cells, precancerous cells, pleomorphic dyskeratosis, type II pneumocyte abnormal squamous cells, adenocarcinoma cells, bronchioloalveolar carcinoma cells, abnormal neuroendocrine cells, small cell carcinoma cells, non-small cell carcinoma cells, tumor cells, neoplastic cells, bronchioloalveolar carcinoma cells, and any combination thereof; a), prior to generating a representative 2D image, evaluating the representative 2D image with an abnormal cell 2D classifier, or evaluating the representative 2D image with a BEC 2D classifier, pre-treating the patient sample to stain a plurality of cells with an agent that facilitates the staining; a) prior to enriching the patient sample for BECs or for cells likely to be indicative of cancer; a) before i) embedding a patient sample in an optical medium and injecting the sample-embedded optical medium into a capillary tube; and ii) mounting the capillary tube in an optical tomography system such that the capillary tube is between an illumination source and an objective lens of the optical tomography system; generating a 2D image of the cell by the optical tomography system includes: sweeping a focal plane of the optical tomography system across the single cell in 1 μm steps to generate single-plane 2D images of the single cell in a stepwise manner; compiling a set of single-plane 2D images of the single cell; and filtering the set of single-plane 2D images of the single cell to generate a representative 2D image of the cell; A representative 2D image of a cell is an image of the center of the cell. evaluating the representative 2D images with the abnormal cell 2D classifier and evaluating the representative 2D images with the BEC 2D classifier both include determining values for a plurality of cell feature measurements; The cellular feature measurements may include object shape features, cell shape features, cytoplasmic features, cell nuclear features, chromatin distribution, nuclear size features, nuclear texture features, other morphometric elements, or any combination thereof. in any combination.

[0005] The present disclosure provides a method of training an AI-based cell classification system, comprising: a) operating an optical tomography system to generate representative 2D images of cells; b) using the representative 2D images of such cells and known cell identification methods to assign a known identifier to each of a plurality of cells, the known identifier being: i) abnormal or having abnormal characteristics; ii) BEC or having BEC-like characteristics; or iii) normal or non-BEC or not having abnormal or BEC-like characteristics; and c) assigning a known identifier to each of the plurality of cells from the representative 2D images of such cells. Also provided is a method comprising calculating a plurality of cell feature measurements; d) performing a regression between the cell feature measurements and a known identifier for each cell to assign an abnormal 2D cell classification score and a BEC2D cell classification score for each cell; e) assigning a test identifier for each cell as: i) having an abnormal feature; ii) having a BEC-like feature; or iii) not having an abnormal feature or a BEC-like feature based on the cell's abnormal 2D cell classification score and BEC2D classification score; and f) comparing each test identifier with the known identifier for each cell to calculate the accuracy of the abnormal cell 2D classifier and the BEC2D classifier.

[0006] The method has the following additional features: comparing the accuracies of both of the 2D cell classifiers with a preselected accuracy requirement, and designating the 2D cell classifier as trained if the preselected accuracy requirement is met by the 2D cell classifier; comparing the accuracies of both 2D cell classifiers with a preselected accuracy requirement, and if the preselected accuracy requirement is not met by the 2D cell classifier, adjusting the parameters of the regression for the affected 2D cell classifier and repeating steps d) to f); The preselected accuracy requirement is a 2D abnormal cell sensitivity value of at least 90%; The preselected accuracy requirement is a 2D abnormal cell specificity of at least 65%; The preselected accuracy requirement is a 2D BEC sensitivity value of at least 85%; The preselected accuracy requirement is at least 60% 2D BEC singular values; the regression includes adaptively boosted logistic regression, random forest, decision tree, or any combination thereof; that known cell differentiation methods include cytological analysis and classification by a pathologist; generating a 3D image of a plurality of cells and using the 3D image in known cell identification methods; The cellular feature measurements may further include one or more of including object shape features, cell shape features, cytoplasmic features, cell nuclei features, chromatin distribution, nuclear size features, nuclear texture features, other morphometric elements, or any combination thereof, in any combination.

[0007] In addition, the training method may be used to train an AI, as described above or otherwise herein.

[0008] The present disclosure further provides a cell classification system comprising: an optical tomography system operable to generate representative 2D images of cells from a patient sample and, if the cells have abnormal or BEC-like characteristics, generate a 3D image of the cells; and a processor operable to compare the representative 2D images of the cells with an abnormal cell 2D classifier to determine if the cells have abnormal characteristics and to compare the representative 2D images of the cells with a BEC 2D classifier to determine if the cells have BEC-like characteristics.

[0009] In some embodiments, the system may be used in either or both of the cell classification or training methods described above or otherwise herein.

[0010] 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.

[0011] The present disclosure may be better understood by reference to the following detailed description in conjunction with the drawings, which are given by way of example only and in which like elements are designated by letters (e.g., 40a, 40b, 40c), in which: [Brief explanation of the drawings]

[0012] [Figure 1] FIG. 1 is a schematic diagram of an optical tomography system that may be used in the present disclosure. [Figure 2] FIG. 1 is a schematic diagram of an optical tomography system operated to acquire multiple 2D images of cells that can be used in the present disclosure. [Figure 3] 1 is a flowchart for the lung cancer detection method of the present disclosure. [Figure 4] 4 is a flowchart of the AI-based cell classification method of the present disclosure for use in the lung cancer detection method of FIG. [Figure 5] 1 is a flowchart of the disclosed 2D classifier training method. [Figure 6] 1 is an exemplary 2D cross-section of a 3D image of a cell that can be obtained using the optical tomography system of the present disclosure. [Figure 7] 1 is a cumulative probability density curve (CDF) plot, e.g., results of an abnormal cell 2D classifier according to the present disclosure, in which the green dashed line indicates the threshold for classifier operation, the red line indicates the trend for cells classified as having abnormal features ("abnormal"), the green line indicates the trend for cells classified as having BEC-like features ("epithelial"), the light blue line indicates the trend for miscellaneous cells ("remainder"), and the dark blue line indicates the trend for cells with BEC-like features and miscellaneous cells ("combined"). [Figure 8]1 is a CDF plot, e.g., results of a BEC2D classifier according to the present disclosure, where the dashed green line indicates the threshold for classifier operation, the green line indicates the trend for cells classified as having BEC-like features, and the light blue line indicates the trend for miscellaneous cells (normal). DETAILED DESCRIPTION OF THE INVENTION

[0013] The following description, along with the accompanying drawings, sets forth certain specific details to provide a thorough understanding of various disclosed embodiments. However, those skilled in the art will recognize that the disclosed embodiments may be practiced without one or more of these specific details, or with other methods, components, devices, materials, and the like, in various combinations. In other instances, well-known structures or components relevant to the environment of the present disclosure are not shown or described to avoid unnecessarily obscuring the description of the embodiments. Furthermore, various embodiments may be methods, systems, or devices.

[0014] The present disclosure relates to an AI-based cell classification method and system that can be used to detect cancer-indicating cells among multiple cells in a lung-related patient sample, such as sputum. Cancer-indicating cells may be cancerous, but they may also include cells with non-cancerous abnormal characteristics. In some embodiments, a sample can be designated as containing cancer-indicating cells (and thus positive for lung cancer) based solely on the detection of cells with non-cancerous abnormal characteristics. When a sample is designated as containing cancer-indicating cells, the patient can be referred for further testing.

[0015] The present system and method uses optical tomography, such as current optical tomography systems, to detect cells and generate 3D images of the cells. However, unlike existing optical tomography systems and methods, the disclosed system and method first generates a representative 2D image of the cell before generating the 3D image of the cell, and then analyzes the representative 2D image using an AI-based 2D cell classifier to determine whether a 3D image should be generated. Patient samples often contain a significant number of diverse cells that are unlikely to indicate lung cancer and therefore not useful for detecting lung cancer. 3D images of such cells are unnecessary, and generating them is a waste of resources. Preventing unnecessary 3D imaging of cells that are not useful for detecting lung cancer allows for more efficient sample processing.

[0016] The present disclosure also provides methods for training the AI of a 2D cell classifier for use in an AI-based cell classification system or method that uses cells with known identifiers.

[0017] In some embodiments, the lung cancer can be non-small cell lung cancer (NSCLC), particularly squamous cell carcinoma and adenocarcinoma. In some embodiments, the lung cancer can be small cell lung cancer (SCLC).

[0018] The term "cancer" refers to an excessive proliferation of cells resulting in unregulated growth, lack of differentiation, local tissue invasion, or metastasis.

[0019] In some embodiments, cells with abnormal characteristics may include cells exhibiting atypia, dysplastic cells, precancerous cells, pleomorphic dyskeratosis, type II pneumocyte aberrant squamous cells, adenocarcinoma cells, bronchioloalveolar carcinoma cells, abnormal neuroendocrine cells, small cell carcinoma cells, non-small cell carcinoma cells, tumor cells, neoplastic cells, bronchioloalveolar carcinoma cells, and any combination thereof.

[0020] In some embodiments, cells with BEC-like characteristics may include cells that typically line the airway lumen within a patient's lungs.

[0021] In some embodiments, the optical tomography system may be a CELL-CT® system (VisionGate, Inc., Washington, USA). AI-based cell classification system

[0022] 1 and 2, the AI-based cell classification system of the present disclosure may include an optical tomography system 100 that may be used to generate both 3D and 2D images of a cell 10 (e.g., cell 10a or cell 10b), or the AI-based cell classification method of the present disclosure may be performed using the optical tomography system 100. The operation of the optical tomography system is described for acquiring an image of one cell 10 within a volume of optical medium within the optical path of a high-magnification microscope, although images of multiple cells 10 within the same volume of optical medium may be acquired.

[0023] The optical tomography system 100 can include a cell imaging system 110 with an illumination source 120 optically coupled to an objective lens 130 such that the illumination passes through the microcapillary 30 and any intervening cells 10 before reaching the objective lens 130. The illumination then passes through the objective lens 130 to a beam splitter 140, which deflects a portion of the illumination onto a mirror 150 and reflects it back to the beam splitter 140 before transmitting it to a high-speed camera 160, while another portion of the illumination is transmitted through the beam splitter 140 directly to the high-speed camera 160 to generate a pseudo-projection image 40 of the cells 10 contained within the optical medium 20 in the microcapillary 30. During 3D imaging, at least one pseudo-projection image 40 of cell 10 is generated by scanning the volume occupied by cell 10 by oscillating mirror 150 in direction 60 (typically using an actuator (not shown), such as a piezoelectric motor), thus sweeping focal plane 50 through cell 10. The images are then merged to create a pseudo-projection image from a single perspective. Additional pseudo-projection images are obtained by rotating microcapillary 30. Each pseudo-projection image is a single image representing a sampled volume with an extent greater than the depth of field of objective lens 130. High-speed camera 160 generates multiple pseudo-projection images 40 for each cell 10, corresponding to multiple axial microcapillary rotational positions, examples of which are shown as 40a, 40b, and 40c in FIG. 1 . In some embodiments, 500 pseudo-projection images are generated as microcapillary 30 is rotated 360°.

[0024] In some embodiments, the optical tomography system 100 is communicatively coupled to a processor 170 operable to receive multiple pseudoprojection images 40 from the high-speed camera 160 and use the pseudoprojection images 40 to generate a 3D image (not shown) of the cell 10. Images before or after manipulation by the processor 170 may be stored in a communicatively coupled memory 180. The processor 170 may transmit data regarding the cell 10, or regarding or derived from multiple cells 10 contained within a patient sample, to a communicatively coupled output 190. The data transmitted to the output 190 may include a 2D or 3D image of one or more cells 10, or may include a summary of the cells in the patient sample analyzed by the optical tomography system 100, such as a graph or table, an indicator such as an anomaly indicator reflecting the likelihood that the patient has or is at high risk for developing lung cancer, or even a simple indication that the sample is positive for cells indicative of lung cancer.

[0025] In embodiments disclosed herein, the optical tomography system 100 can also be operated to generate representative 2D images of the cells 10. In such embodiments, the objective lens 130 has a focal plane 50 that moves as the objective lens 130 sweeps back and forth 60 across the microcapillary 30 and any cells 10 to generate multiple 2D images (not shown). This method of generating 2D images by moving the objective lens 130 differs from the method of generating 3D images using pseudo-projection images generated by oscillating the mirror 150. The processor 170 is operable to receive the multiple 2D images and determine a representative 2D image of the cells 10. In some embodiments, the representative 2D image of the cells 10 is an image of a central portion, such as the center of the cell. The processor 170 is then further operable to perform AI-based cell classification of the representative 2D images using a 2D cell classifier as described herein to determine whether the cells 10 have abnormal or BEC-like characteristics or whether they do not have abnormal or BEC-like characteristics. Only when the cell 10 is determined to have abnormal or BEC-like characteristics will the processor 170 direct the cell imaging system 110 to generate a pseudo-projection image 40 of the cell 10 .

[0026] A significant number of cells 10 in a patient's sputum sample or other sample are unlikely to indicate lung cancer, and therefore generating 3D images of these cells is not useful for detecting lung cancer. Many of these cells do not have the morphology of normal or abnormal cells of the bronchial epithelium, and therefore, they can be eliminated using 2D images alone. Removing such cells from a patient sample prior to AI-based cell classification is difficult, and therefore, in conventional cell classification methods, they are simply imaged along with cells likely to indicate lung cancer. The ability to use 2D images to first assess whether cells are likely to indicate lung cancer before capturing pseudoprojection images 40 allows AI-based cell classification to focus on cells 10 that provide meaningful information without wasting processing and analysis time on cells that are not useful in detecting lung cancer.

[0027] In some embodiments, the cell imaging system 110 includes an illumination source 120 , an objective lens 130 , a beam splitter 140 , a mirror 150 , and a high-speed camera 160 .

[0028] In some embodiments, the optical tomography system 100 further includes a processor 170, an optional communicatively coupled memory 180, and a communicatively coupled output 190.

[0029] In some specific embodiments, the optical tomography system 100 further includes a microcapillary 30, an optical medium 20, or one or more cells 10, while in other embodiments, the optical tomography system 100 does not include one or more of these potential components, although they may be provided for operation of the system.

[0030] In certain embodiments, the AI-based cell classification system 100 is operable to generate images in at least two different modes. One mode includes a cell search mode, during which 2D images of the cell 10 are generated and processed to generate a representative 2D image, which is then used to determine whether the cell 10 has abnormal or BEC-like characteristics. The 2D image in this mode is generated by moving the objective lens 130 in direction 60 to sweep the focal plane 50 through the cell. The second mode includes a projection image capture mode, during which if the cell 10 is determined to have abnormal or BEC-like characteristics during the cell search mode, a pseudo-projection image 40 is generated and used to generate a 3D image of the cell 10. The 3D image in this mode is generated by oscillating the mirror 150 in direction 60 to generate the pseudo-projection image 40. Lung cancer detection method

[0031] 3 and 4 illustrate a lung cancer detection method 200 that may be performed using an optical tomography system, such as optical tomography system 100. Elements of optical tomography system 100 are referenced by way of example in this description of lung cancer detection method 200. Similar components of different optical tomography systems may also be used in connection with lung cancer detection method 200.

[0032] The lung cancer detection method 200 includes a step 210 in which a lung cell sample is collected from a patient. For example, the lung cell sample may be sputum, although other sample types, such as a bronchoalveolar lavage (BAL), a nasal swab, or a sample obtained by biopsy, may also be used. In some embodiments, the sample may include urine or blood, particularly for the detection of other cancers or abnormal cells.

[0033] A sample can be subjected to the AI-based cell classification method 300 any time the cells have not been broken down to the point that limited cellular content remains. The duration of such time may depend on storage conditions, such as whether the sample is refrigerated or how the sample is or will be processed. In some embodiments, a sample may be subjected to the AI-based cell classification method 300 within 30 minutes, within 1 hour, within 2 hours, within 6 hours, within 12 hours, within 1 day, within 2 days, within 1 week, or within 2 weeks of collection.

[0034] In the case of sputum samples in particular, patient samples may contain many cell types that are unlikely to be indicative of lung cancer, such as white blood cells including polymorphonuclear leukocytes, monocytes, and lymphocytes, cell clusters, oral squamous cells, and squamous intermediate cells (depicted in Figure 6, top panel). Samples also typically contain significant amounts of non-cellular debris or cell fragments.

[0035] In some embodiments, the sample comprises abnormal cells, BECs, squamous cells, monocytes, lymphocytes, polymorphonuclear leukocytes, other white blood cells, debris, cell fragments, cell clusters, and any combination thereof.

[0036] In step 220, the sample is processed for analysis. Processing may optionally include staining or otherwise inspecting the cells with an agent to facilitate detection of any features, such as nuclei, of any of the plurality of cells 10 or any of the plurality of cells 10 using the optical tomography system 100, generating representative 2D images, evaluating the representative 2D images using an abnormal cell 2D classifier, evaluating the representative 2D images using a BEC 2D classifier, generating 3D images of the cells, or analyzing the 3D images of the cells. In a specific embodiment, the cells may be stained with hematoxylin.

[0037] Processing may optionally include enrichment of the sample for cells of interest, either alone or in combination with staining. For example, in some embodiments, the sample is enriched for BECs.

[0038] In one embodiment, samples can be enriched for BECs by staining cytoskeletal proteins and using fluorescence-activated cell sorting (FACS)-based enrichment.

[0039] In one embodiment, BEC enrichment can include treating a sample with at least one antibody, typically multiple antibodies, bearing a fluorescent conjugate that can be used for FACS-based enrichment. In particular, an antibody may bind to BECs, or the antibody may bind contaminating inflammatory cells such as neutrophils and macrophages. The antibody that binds contaminating inflammatory cells can include an anti-CD45 antibody. In some embodiments, the sample is treated with a combination of an antibody that binds BECs and an antibody that binds contaminating inflammatory cells, where the antibodies have separate fluorescent conjugates.

[0040] In one embodiment, cells may be stained with 4',6-diamidino-2-phenylindole (DAPI) alone or in combination with antibodies for the purpose of FACS-based cell enrichment.

[0041] Cells can be enriched by FACS in which gating is used to exclude DAPI-positive material (which tends to be doublet cells or debris), high side scatter objects, objects bound by anti-inflammatory cell antibodies, or any combination thereof. Cells can also be enriched by FACS in which gating is used to select cells bound by anti-BEC antibodies. In some embodiments, both exclusive and inclusive gating can be used and implemented simultaneously or sequentially.

[0042] Following any staining or enrichment, sample processing 220 involves placing the cells 10 contained in the sample into an optical medium 20 .

[0043] The optical medium 20 can be any medium reasonably expected to maintain the integrity of the cells 10 for the anticipated duration of time before and during AI-based cell sorting 300. The optical medium 20 can also have a viscosity that allows movement of the optical medium 20 through the microcapillary tubes 30. The optical medium 20 can also not interfere with image generation by the optical tomography system 100. In particular, the optical medium can have an optical index that matches the optical index of other components of the optical tomography system 100 and the microcapillary tubes 30 through which the light passes during image acquisition. Typically, the components of the optical tomography system 100 through which the light passes, and the microcapillary tubes 30, also have matching indices. For optimal optical tomography operation, any change in the movement of the light should be due to encountering the object being imaged, and not due to changes in the optical index of other components or objects in the light path.

[0044] In step 230, an optical medium 20 containing a plurality of cells 10 from a patient sample is injected into a microcapillary 30. In some embodiments, the microcapillary 30 may have an outer diameter of 500 μm or less, e.g., 30 μm to 500 μm. In some embodiments, the microcapillary 30 may have an inner diameter of 400 μm or less, e.g., 30 μm to 400 μm, e.g., 50 μm. In one embodiment, the entire portion of the patient sample to be analyzed is placed in one microcapillary 30. In another embodiment, the portions of the patient sample to be analyzed are placed in multiple microcapillaries 30 that can be sequentially subjected to the AI-based cell classification method 300. In yet another embodiment, the sample may be pumped through the microcapillary 30 from a sample reservoir.

[0045] In step 240, the micro-capillary tube 30 is mounted in the optical tomography system 100 so that the micro-capillary tube 30 is between the illumination source 120 and the objective lens 130.

[0046] In step 250, the optical medium 20 and any cells 10 contained therein are advanced into or through the microcapillary 30 (before the initial step 240) by applying pressure to one end of the microcapillary such that different volumes of the optical medium 20 carrying different portions of the patient sample are in the optical path of the tomography system 100 between the illumination source 120 and the objective lens 130. In some embodiments, a plunger (not shown) is used to advance the optical medium 20. For example, the plunger can be applied to a reservoir (not shown) of optical medium 20 and the patient sample connected to the microcapillary 30 to push additional optical medium 20 from the reservoir into the microcapillary 30.

[0047] The method then proceeds to AI-based cell classification 300, which includes a cell search mode 310 and a projection image capture mode 320, both of which are described in further detail in FIG.

[0048] In cell search mode 310, optical tomography system 100 generates 2D images of cell 10, which are further processed to generate a representative 2D image that is used to determine whether cell 10 has abnormal or BEC-like characteristics. In some embodiments, an algorithm selects a central image, such as the center of cell 10, as the representative 2D image. In some embodiments, in cell search mode 310, optical tomography system 100 sweeps focal plane 50 through cell 10 in direction 60 at 1 μm intervals to capture a series of 2D images. Multiple cells 10 can be identified within the same volume of optical medium 20 in the optical path of optical tomography system 100.

[0049] In projection image capture mode 320, each cell 10 identified as having abnormal or BEC-like characteristics is further imaged to generate multiple pseudo-projection images 40 that are used to generate a 3D image of the cell 10.

[0050] In step 260, it is determined whether a preselected number of cells with BEC-like characteristics have been 3D imaged. If the preselected number has been reached, the process moves to step 270. Otherwise, the process returns to step 250 to advance the sample within the microcapillary.

[0051] In step 270, images of cells selected for 3D imaging are provided to the user, for example, using output 190. Results may include results specific to individual cells 10, such as a 3D image of the cells 10. The results of AI-based cell classification may also include data based on the analysis of all or a portion of the patient sample or multiple cells 10, such as an abnormality index and / or a cancer threshold, as further described in FIG. 4. For example, the results of AI-based cell classification may include a list or other representation of individual cells 10 found to have abnormal or BEC-like features or otherwise marked according to preselected criteria for further scrutiny of the 3D images by the user. The results of AI-based cell classification may include graphical, statistical, or numerical information, such as the number of detected cells (typically the total number of analyzed cells enumerated), the number of cells for which 3D images were generated, the number of enumerated cells with abnormal features, the number of enumerated cells with BEC-like features, or the relative proportions of any of these cell groups. The results of AI-based cell classification may also include information regarding predicted accuracy. The results of AI-based cell classification may include an indicator, such as an abnormality indicator, that reflects the likelihood that the patient has or is at high risk of developing lung cancer, or even a simple indication that the sample is positive for cells indicative of lung cancer.

[0052] The sample identification data may also comprise the results of AI-based cell classification.

[0053] With particular reference to FIG. 4, the AI-based cell classification method 300 can include a cell search mode 310 and a projection image capture mode 320.

[0054] The cell search mode 310 may include a step 400 in which the optical tomography system 100 generates multiple single focal plane 2D images of a cross-sectional area of the microcapillary 30, each image at a different focal plane 50, by sweeping the focal plane 50 back and forth in direction 60 across the microcapillary (and any cells in any of the focal planes 50) by moving the objective lens 130. In one embodiment, the single focal plane 2D images may be obtained at 1 μm increments in direction 60.

[0055] In step 410, a representative 2D image of the cell 10 is generated. To do so, the processor 170 compiles and filters at least a portion of the multiple single focal plane 2D images to determine whether the 2D image contains features associated with the cell 10 or other solid objects within the optical medium 20, such as being dark relative to the optical medium 20. If the multiple single focal plane 2D images contain features associated with the cell 10 or other solid objects, the processor 170 analyzes at least a portion of the multiple single focal plane 2D images associated with the cell 10 (e.g., all or a portion of the 2D image containing dark regions) to determine which should be designated as a representative 2D image based on preselected criteria. In one embodiment, the processor determines which 2D image represents a central portion, such as the center, of the cell 10. This image representing the central portion of the cell 10 is designated as the representative 2D image of the cell 10.

[0056] If a representative 2D image is generated, the process proceeds to step 420. In some embodiments, rather than proceeding directly to step 250, the method determines whether there are additional cells 10 within the volume of optical medium and proceeds to perform step 410 (and, optionally, step 400 if additional 2D images are desired) for the other cells. This may be repeated for all, a fixed number, or a fixed percentage of the estimated cells 10 within the volume of optical medium 20 within the optical path of the optical tomography system 100.

[0057] In step 420, the processor 170 evaluates the representative 2D image using an abnormal cell 2D classifier to determine whether the cell 10 has abnormal characteristics. The processor can use an AI-trained or implemented algorithm in step 420. The abnormal cell 2D classifier can be trained as described in the training method 500 herein. Specifically, the abnormal cell 2D classifier can identify multiple features of the cell 10, perform cell feature measurements, and then compare at least a subset of such cell feature measurements with trained values for cell feature measurements associated with cells having abnormal characteristics and, optionally, cells without abnormal characteristics. Based on the comparison, the cell can be identified as having or not having abnormal characteristics. To facilitate the identification and measurement of multiple cell features, the abnormal cell 2D classifier can divide the representative 2D image into nuclear and non-nuclear cell portions. The cellular feature measurements may include object shape features, cell shape features, cytoplasmic features, cell nucleus features, chromatin distribution, nuclear size features, nuclear texture features, other morphometric elements, or any combination thereof.

[0058] The abnormal cell 2D classifier may have a preselected abnormal cell 2D accuracy, typically a preselected abnormal cell 2D sensitivity or a preselected abnormal cell 2D specificity.

[0059] If the cell 10 is determined to have anomalous characteristics in 2D, the method can proceed directly to projection image capture mode 320, as shown. However, in some embodiments, even if the cell 10 is determined to have anomalous characteristics in 2D, the method can still proceed to step 430 (not shown) to also determine whether the cell 10 has BEC-like characteristics in 2D.

[0060] In the illustrated embodiment, if the cell 10 is determined to have no abnormal characteristics, the method proceeds to step 430 .

[0061] Next, in step 430, the processor 170 evaluates the representative 2D images using a BEC2D classifier to determine whether the cells 10 have BEC-like features. The processor may use an AI-trained or implemented algorithm in step 430. The BEC2D classifier may be trained as described in training method 500 herein. Specifically, the BEC2D classifier may identify multiple features of the cells 10, perform cell feature measurements, and then compare at least one set of identified and measured cell feature measurements to trained values for cell feature measurements associated with cells having BEC-like features and, optionally, cells not having BEC-like features. Cells may be identified as having BEC-like features or not having BEC-like features based on the comparison. The cell features and cell feature measurements may be selected from the same group described above for the abnormal cell 2D classifier. In some embodiments, the identification and measurement of cell features may be performed separately for use with the abnormal cell 2D classifier and the BEC2D classifier. Cells can be designated as having BEC-like characteristics or not having BEC-like characteristics based on the comparison.

[0062] The BEC2D classifier may have a preselected 2D BEC accuracy, typically a preselected 2D BEC sensitivity or 2D BEC specificity.

[0063] Operating together, the abnormal cell 2D classifier and the BEC 2D classifier may have a preselected abnormal cell query rate or a preselected BEC query rate.

[0064] If the cell 10 is determined to have BEC-like characteristics, the method proceeds directly to projection image capture mode 320, as shown. If the cell 10 is determined not to have BEC-like characteristics, no 3D image is captured and the method proceeds to step 250. In some embodiments (not shown), rather than proceeding directly to step 250, the method proceeds to determine whether another cell 10 is present within the volume of optical medium and perform steps 410-430 (and optionally, step 400 if additional 2D images are desired) for the other cell. This may be repeated for all, a fixed number, or a fixed percentage of the estimated cells 10 within the volume of optical medium 20 within the optical path of the optical tomography system 100.

[0065] In projection image capture mode 320, optical tomography system 100 generates multiple pseudo-projection images 40 by oscillating mirror 150 in step 440. These images may be taken over a 360-degree rotation about cell 10. However, pseudo-projections 40 may be taken over as little as 180 degrees of rotation about cell 10.

[0066] Next, in step 450, processor 170 generates a 3D image of cell 10 using at least some of the pseudo-projection images 40. Typically, all of the pseudo-projection images 40 are used to generate the 3D image of cell 10. However, if it is determined that a pseudo-projection image 40 is of low quality or may contain errors, 3D imaging of the cell may be interrupted. Additionally, typically, pseudo-projection images 40 covering a 360-degree rotation around cell 10 are used to generate the 3D image, although pseudo-projection images covering a rotation of approximately 180 degrees around cell 10 may also be used.

[0067] In step 460, processor 170 uses the 3D classifier to determine whether the cells have abnormal or BEC-like characteristics. This information is used to determine whether the cells are likely to be indicative of lung cancer. The 3D classifier may also have a preselected accuracy value.

[0068] The process then returns to step 260 to determine whether a preselected number of BECs have been 3D imaged. If the preselected number has been reached, the process moves to step 270, which in some embodiments may include steps 470 and 480.

[0069] In step 470, the data regarding the cells is stored as patient sample data. The patient sample data may be used in determining an abnormality index. The patient sample data may also reflect a total number of enumerated analyzed cells having abnormal characteristics, a total number of enumerated analyzed cells having BEC-like characteristics, or a total number of enumerated analyzed cells, and steps 410-470, and optionally, step 400, if additional 2D images are required, may be repeated for additional cells until a preselected threshold for any or all of these totals is reached.

[0070] In step 480, the processor compares the total number of enumerated analyzed cells having abnormal characteristics to the total number of enumerated analyzed cells having BEC-like characteristics or the total number of enumerated analyzed cells to generate an abnormality index. For example, the abnormality index may be correlated with a percentage of either of these sums.

[0071] In step 490, the processor compares the abnormality index to a cancer threshold set to correspond to the patient sample being designated positive for cells indicative of lung cancer.

[0072] The entire method 300, including 2D and 3D cell sorting, may have a preselected assay accuracy, e.g., a preselected assay sensitivity or a preselected assay specificity. The cell sorting method 300 as described herein may be substantially faster than similar methods that do not use a 2D cell classifier, but instead generate 3D images of all identified or putative cells. In some embodiments, method 300 can reduce cell sorting time, such as for steps 400-460, by at least 25%, at least 35%, at least 40%, at least 45%, at least 50%, or within the range of 25% to 60%, 25% to 50%, 25% to 45%, 25% to 40%, 25% to 35%, 35% to 60%, 35% to 50%, 35% to 45%, 35% to 40%, 40% to 60%, 40% to 50%, 40% to 45%, 45% to 60%, or 45% to 50%.

[0073] In embodiments such as those described in cell classification method 300, 3D images are not generated for cells that do not have abnormal or BEC-like features as determined using 2D classification, although in other embodiments, 3D images may be generated for at least one cell or multiple cells that do not have abnormal or BEC-like features. Such images may be used, for example, to validate the performance or accuracy of an AI-based cell classification system or method and / or one or both of the component 2D cell classifiers. Generating such 3D images increases the total cell classification time unless 3D images are generated for all cells, but a time reduction should still be realized compared to methods in which all or nearly all cells are 3D imaged. In some embodiments, the total number of cells without abnormal or BEC-like features for which a 3D image is generated is less than or equal to 25%, 20%, 15%, 10%, 5%, 1%, or 0.1% to 25%, 0.1% to 20%, 0.1% to 30%, 0.1% to 40%, 0.1% to 50%, 0.1% to 60%, 0.1% to 70%, 0.1% to 80%, 0.1% to 90%, 0.1% to 100%, 0.1% to 120%, 0.1% to 140%, 0.1% to 200%, 0.1% to 250%, 0.1% to 200%, 0.1% to 140%, 0.1% to 20 ... The range of cell density may be limited to 0.1% to 15%, 0.1% to 10%, 0.1% to 5%, 0.1% to 1%, 1% to 25%, 1% to 20%, 1% to 15%, 1% to 10%, 1% to 5%, 5% to 25%, 5% to 20%, 5% to 15%, 5% to 10%, 10% to 25%, 10% to 20%, 10% to 15%, 15% to 15%, 15% to 20%, or 20% to 25%. Cells without abnormal or BEC-like features selected for 3D imaging may be selected randomly or based on preselected criteria, such as criteria that result in the generation of a 3D image of at least one of the most common cell types in the sample.

[0074] The AI-based cell classification method 300 as described herein may also substantially reduce the number of cells for which 3D images are generated compared to similar methods that do not use the abnormal cell 2D classifier and the BEC2D classifier (collectively, 2D cell classifiers) and instead generate 3D images of all identified or putative cells. In some embodiments, the method 300 may reduce the number of cells for which 3D images are generated by at least 30%, at least 40%, at least 50%, at least 55%, at least 60%, or within the range of 30% to 70%, 30% to 60%, 30% to 55%, 30% to 50%, 30% to 40%, 40% to 70%, 40% to 60%, 40% to 55%, 40% to 50%, 50% to 70%, 50% to 60%, 50% to 55%, 55% to 70%, 55% to 60%, or 60% to 70%.

[0075] The abnormal cell 2D classifier, the BEC2D classifier, and the overall AI-based cell classification method may have a pre-selected accuracy.

[0076] For any assay, the "sensitivity" of the assay is defined as the proportion of specimens (such as cells or the sample as a whole) that are actually positive for a characteristic that is also correctly identified as positive by the assay.

[0077] For any assay, the "specificity" of the assay is defined as the proportion of specimens (such as cells or the sample as a whole) that are actually negative for a characteristic that is also correctly identified as positive by the assay.

[0078] For the 2D cell classifiers used in cell classification method 300, sensitivity tends to be significantly more accurate than specificity, as only cells positive for abnormal or BEC-like features are 3D imaged. Thus, low sensitivity for either 2D cell classifier can result in cells likely to represent cancer not being 3D imaged and included in the lung cancer test results, potentially resulting in a false negative. In contrast, low specificity may result in more cells than necessary being 3D imaged, somewhat unnecessarily slowing the AI-based cell classification, but it does not result in any cells likely to represent cancer being missed and resulting in a false negative.

[0079] In the context of an abnormal cell 2D classifier, 2D abnormal cell sensitivity is the proportion of cells identified by the abnormal cell 2D classifier as having abnormal characteristics that are or would be determined to be abnormal using known discrimination methods, such as cytology (e.g., review by a pathologist). 2D abnormal cell specificity is the proportion of cells identified by the abnormal cell 2D classifier as not having abnormal characteristics that are or would be determined to be normal using known discrimination methods, such as cytology.

[0080] In some embodiments, the preselected 2D abnormal cell sensitivity can be at least 90%, at least 95%, at least 98%, or within the ranges of 90% to 100%, 90% to 99.9%, 90% to 99%, 90% to 98%, 90% to 95%, 95% to 100%, 95% to 99.9%, 95% to 98%, 98% to 100%, 98% to 99.9%, or 98% to 99%.

[0081] In other embodiments, the preselected 2D abnormal cell specificity may be at least 65%, at least 70%, at least 75%, or within the ranges of 65% to 100%, 65% to 99%, 65% to 90%, 65% to 80%, 65% to 75%, 65% to 70%, 70% to 100%, 70% to 99%, 70% to 90%, 70% to 80%, 70% to 75%, 75% to 100%, 75% to 99%, 75% to 90%, or 75% to 80%.

[0082] In the context of the BEC2D classifier, 2D BEC sensitivity is the proportion of cells identified by the BEC2D classifier as having BEC-like characteristics that are or would be determined to be BECs using known discrimination methods, such as cytology. 2D BEC specificity is the proportion of cells identified by the BEC2D classifier as not having BEC-like characteristics that are or would be determined to not be BECs using known discrimination methods, such as cytology.

[0083] In some embodiments, the preselected 2D BEC sensitivity can be at least 85%, at least 90%, at least 95%, or within the range of 85% to 100%, 85% to 99%, 85% to 95%, 85% to 90%, 90% to 100%, 90% to 99%, 90% to 95%, 95% to 100%, or 95% to 99%.

[0084] In some embodiments, the preselected 2D BEC specificity can be at least 60%, at least 65%, at least 70%, or within the range of 60% to 100%, 60% to 90%, 60% to 80%, 60% to 70%, 65% to 100%, 65% to 90%, 65% to 80%, 65% to 70%, 70% to 100%, 70% to 90%, or 70% to 80%.

[0085] In the context of AI-based 2D cell classification using both 2D cell classifiers, an abnormal cell referral rate can also be calculated. The abnormal cell referral rate is the proportion of cells that are identified as having abnormal characteristics or as having BEC-like characteristics and are therefore 3D imaged, and are or would be determined to be abnormal using known identification methods, such as cytology. The abnormal cell referral rate can be higher than the 2D abnormal cell sensitivity, because cells that are abnormal but not identified as having abnormal characteristics can still be identified as having BEC-like characteristics and, for that reason, can be subjected to 3D imaging.

[0086] In some embodiments, the preselected abnormal cell referral rate is at least 70%, at least 80%, at least 85%, at least 90%, at least 95%, at least 98%, at least 99%, or between 70% and 99.9%, between 80% and 99.9%, between 85% and 99.9%, between 90% and 99.9%, between 95% and 99.9%, between 98% and 99.9%, between 99% and 99.9%, ...%, between 80% and 99.9%, between 85% and 99.9%, between 90% and 99.9%, between 95% and 99.9%, between 98% and 99.9%, between 99% and 99.9%, between 70% and 99%, between 80% and 99.9%, between 85% and 99.9%, between 90% and 99.9%, between 95% and 99.9%, between 98% and 99.9%, between 99% and 99.9%, between 70% and 99%, between 80% and 99.9%, between 95% and 99.9%, between 98% and 99.9%, between 99% and 99.9%, between 95% and 99.9%, between 98% and 99.9%, between 99% and 99.9%, between 95% and 99.9%, between 98% and 99.9%, between 99 The range may be within the following ranges: 9%, 85% to 99%, 90% to 99%, 95% to 99%, 98% to 99%, 70% to 98%, 80% to 98%, 85% to 98%, 90% to 98%, 95% to 98%, 70% to 95%, 80% to 95%, 85% to 95%, 90% to 95%, 70% to 90%, 80% to 90%, 85% to 90%, 70% to 85%, 80% to 85%, or 70% to 80%.

[0087] In the context of AI-based 2D cell classification using the abnormal cell 2D classifier alone or in combination with the BEC2D classifier, an abnormal cell true positive referral rate may also be calculated. The abnormal cell true positive referral rate is the proportion of cells identified by the abnormal cell 2D classifier as having abnormal characteristics that are or would be determined to be abnormal using known identification methods, such as cytology. The abnormal cell true positive referral rate may not actually be as meaningful as the abnormal cell referral rate, since accurate results may be obtained regardless of why the abnormal cells were 3D imaged. In some embodiments, the abnormal cell true positive referral rate may be at least 60%, at least 70%, at least 80%, at least 90%, or within the ranges of 60% to 99.9%, 70% to 99.9%, 80% to 99.9%, 90% to 99.9%, 60% to 99%, 70% to 99%, 80% to 99%, 90% to 99%, 60% to 90%, 70% to 90%, 80% to 90%, 60% to 80%, 70% to 80%, or 60% to 70%.

[0088] In the context of AI-based 2D cell classification using both 2D cell classifiers, the BEC referral rate can also be calculated and considered when determining accuracy. The BEC referral rate is the percentage of cells that are identified as having abnormal or BEC-like characteristics and are therefore 3D imaged, and are or would be determined to be BECs using known identification methods, such as cytology. Because BECs that are not identified as having BEC-like characteristics can still be identified as having abnormal characteristics and therefore undergo 3D imaging, the BEC referral rate can be higher than the BEC 2D classifier sensitivity. A given number of BECs are typically imaged to ensure the accuracy of this type of lung cancer assay, and therefore the number of BECs actually imaged in 3D can affect how long it takes to perform a lung cancer assay using a given patient sample, or whether the sample is deemed suitable for accurate results.

[0089] In some embodiments, the preselected BEC referral rate is at least 70%, at least 75%, at least 80%, at least 85%, at least 90%, at least 91%, at least 95%, or between 70% and 99.9%, 75% and 99.9%, 80% and 99.9%, 85% and 99.9%, 90% and 99.9%, 91% and 99.9%, 95% and 99.9%, 70% and 99%, 75% and 99%, 80% and 99%, 85% and 99%, or The range may be in the ranges of % to 99%, 90% to 99%, 91% to 99%, 95% to 99%, 70% to 95%, 75% to 95%, 80% to 95%, 85% to 95%, 90% to 95%, 70% to 91%, 75% to 91%, 80% to 91%, 85% to 91%, 90% to 91%, 70% to 90%, 75% to 90%, 80% to 90%, 85% to 90%, 70% to 85%, 75% to 85%, or 80% to 85%.

[0090] In the context of AI-based 2D cell classification using the BEC2D classifier alone or in combination with the abnormal cell 2D classifier, a BEC true positive referral rate may also be calculated. The BEC true positive referral rate is the proportion of cells identified by the BEC2D classifier as having BEC-like characteristics that are or would be determined to be BECs using known identification methods, such as cytology. The BEC true positive referral rate may not actually be as meaningful as the BEC referral rate, since accurate results may be obtained regardless of why the BECs were 3D imaged. In some embodiments, the BEC true positive referral rate can be at least 50%, at least 60%, at least 70%, at least 80%, at least 90%, or in the range of 50% to 99.9%, 60% to 99.9%, 70% to 99.9%, 80% to 99.9%, 90% to 99.9%, 50% to 99%, 60% to 99%, 70% to 99%, 80% to 99%, 90% to 99%, 50% to 90%, 60% to 90%, 70% to 90%, 80% to 90%, 50% to 80%, 60% to 80%, 70% to 80%, 50% to 70%, 60% to 70%, or 50% to 60%.

[0091] The AI-based cell classification method 300 may also have a preselected 2D rejection rate, which is the percentage of the total number of enumerated analyzed cells that are identified as not having abnormal or BEC-like characteristics and therefore are not imaged in 3D. In some embodiments, the 2D rejection rate can be 70% or less, 60% or less, 55% or less, 50% or less, or within the ranges of 0% to 70%, 1% to 70%, 10% to 70%, 25% to 70%, 40% to 70%, 50% to 70%, 55% to 70%, 60% to 70%, 0% to 60%, 1% to 60%, 10% to 60%, 25% to 60%, 40% to 60%, 50% to 60%, 55% to 60%, 0% to 55%, 1% to 55%, 10% to 55%, 25% to 55%, 40% to 55%, 50% to 55%, 0% to 50%, 1% to 50%, 10% to 50%, 25% to 50%, or 40% to 50%.

[0092] Accuracy values can also be calculated for 3D classifiers. Specifically, an AI-based cell classification method can have a 3D abnormal cell sensitivity, which is the percentage of cells that are identified by the 3D classifier as abnormal (or positive, such as for lung cancer) that are or would be determined to be abnormal using a known identification method, such as cytology (e.g., review by a pathologist). When calculating the 3D abnormal cell sensitivity, the number of false-negative cells that are not referenced for 3D imaging by the 2D classifier must be subtracted from the known total number of abnormal cells, since the 3D classifier has no opportunity to classify such cells as abnormal. A preselected minimum 3D abnormal cell sensitivity helps prevent false-negative test results. Similar calculations can be performed for other accuracy parameters of the 3D cell classifier.

[0093] The lung cancer detection method 200 also has a pre-selected lung cancer accuracy parameter for the lung cancer test as a whole. Lung cancer sensitivity is the proportion of patients who actually have lung cancer who are identified by the lung cancer detection method as having lung cancer (a positive test result). The AI-based cell classification method 300 also has a pre-selected assay accuracy parameter, which in some embodiments may be the same as the lung cancer accuracy parameter.

[0094] Generally, the sensitivity of any lung cancer detection method using 3D imaging of the type used in method 300 depends on the number of BECs that are 3D imaged. The number of BECs that are 3D imaged from a patient sample may be determined from the total number of BECs enumerated and may be thresholded so that a preselected assay precision, such as a preselected assay sensitivity, is statistically likely to have been achieved. The method 300 may continue to analyze the cells 10 until a preselected threshold is reached, which reflects the total number of enumerated BECs and may be at least 250, at least 400, at least 500, at least 600, at least 700, at least 800, at least 900, at least 1000, at least 1100, at least 1200, at least 1300, at least 1400, at least 1500, at least 2000, at least 2500, or within the range of 250 to 2500, 250 to 2000, 400 to 1900, 600 to 1800, 800 to 1700, 1000 to 1600, 1200 to 1600, or 1300 to 1500.

[0095] Method 300 does not use abnormal cell 2D classifiers and BEC2D classifiers, and instead can reach a preselected sensitivity more quickly than similar methods that generate 3D images of all identified or putative cells, because time is not spent capturing 3D images of cells that are unlikely to indicate lung cancer.

[0096] Assay abnormal cell sensitivity is the proportion of cells that are or would be identified as abnormal using known discrimination methods that are identified as abnormal by the AI-based cell classification method 300. Assay abnormal cell sensitivity can be approximated by multiplying the 2D abnormal cell sensitivity by the 3D abnormal cell sensitivity. Assay abnormal cell specificity is the proportion of cells that are or would be identified as normal using known discrimination methods that are identified as normal by the AI-based cell classification method 300. Assay abnormal cell sensitivity can be approximated by multiplying the 2D abnormal cell specificity by the 3D abnormal cell specificity.

[0097] Meyer, MG et al. (2015), CELL-CT(R) 3-dimensional cell imaging technology platform enables the detection of lung cancer using the noninvasive LuCED sputum test, Cancer Cytopathology, 123:512~523(doi.org / 10.1002 / cncy.21576); Wilbur, DC et al. (2015), Automated 3-dimensional morphologic analysis of sputum specimens for lung cancer detection: Performance characteristics support use in lung cancer screening, Cancer Cytopathology, 123:548-556 (doi.org / 10.1002 / cncy.21565); U.S. Patent No. 6519355, U.S. Patent No. 6522775, U.S. Patent No. 6591003, U.S. Patent No. 6636623, U.S. Patent No. 6697508, U.S. Patent No. 7197355, U.S. Patent No. 7494809, U.S. Patent No. 7569789, U.S. Patent No. 7738945, U.S. Patent No. 7811825, U.S. Patent No. 7835561, U.S. Patent No. 7867778, U.S. Patent No. 7787112, U.S. Patent No. 7907765, U.S. Patent No. Nos. 7,933,010, 8,090,183, 8,155,420, 8,947,510, 9,594,072, 10,753,857, 11,069,054, and U.S. Patent Application Publication No. 20200018704 are each incorporated by reference herein in their entirety and particularly as they relate to the components of the optical tomography system and lung cancer detection method and system described herein, basic operations including staining and enrichment of potential cells, image formation including formation of pseudoprojection images, and 3D classifiers. Training Methods

[0098] The present disclosure further includes a method 500, as set forth in the flowchart of Figure 5, for training a 2D cell classifier that may be used in method 200, method 300, or method 310. The 2D classifier training is governed by a binary ground truth: for an abnormal cell 2D classifier, cells either have abnormal features or do not have abnormal features, while for a BEC 2D classifier, cells either have BEC-like features or do not have BEC-like features.

[0099] Elements of optical tomography system 100 are referenced in this description of training method 500 by way of example. Similar components of different optical tomography systems may also be used in connection with training method 500. However, the trained values generated using training method 500 as implemented in the abnormal cell 2D classifier and BEC2D classifier may be most useful when used with the same optical tomography system used for training method 500, or with an optical tomography system that has identical or very similar components and performance specifications that are unlikely to produce any difference in results outside of a preselected accuracy.

[0100] In step 510, optical tomography system 100 is used to generate a set of representative 2D images of abnormal cells of a plurality of cells identified as abnormal by known identification methods, such as cytopathology, a set of representative 2D images of BECs of a plurality of cells identified as BECs by known identification methods, and a set of representative 2D images of normal / non-BECs of a plurality of lung sample cells identified as normal and non-BECs by known identification methods, for example, according to steps 400 and 410 of method 310. In some embodiments, for training purposes, a cell is identified as abnormal by known identification methods if it is confirmed by a cytopathologist to be in one of the following diagnostic categories: 1) atypia, dysplasia, precancerous cell, malignant cell, or 2) pleomorphic dyskeratosis, type II pneumocyte. In some embodiments, for training purposes, a cell is identified as a BEC by known identification methods if it is independently confirmed as a BEC by two cytopathologists. In some embodiments, a cell is identified as normal if it is randomly obtained from a patient sample and identified as such by a cytopathologist because it does not meet the abnormal cell criteria, and a cell is identified as a non-BEC if any cytopathologist identifies the cell as not being a BEC.

[0101] Cells for training purposes may, in some embodiments, be obtained from the same type of patient sample (e.g., sputum) or prepared in the same manner (e.g., placed in the same optical medium or stained with the same dye) as those used in connection with the AI-based cell classification methods and systems operated using the trained values.

[0102] Although method 500 illustrates simultaneous training of the abnormal cell 2D classifier and the BEC2D classifier for simplicity, the 2D cell classifiers may be trained separately, optionally using separate sets of training cells, as described in the Examples.

[0103] In step 520, the processor calculates a plurality of cell feature measurements for each of the plurality of cells using the representative 2D images of the cells generated in step 510.

[0104] In step 530, the processor performs a regression between the cell feature measurements and the known identifiers of the cells ((i) abnormal, ii) BEC, or iii) normal and non-BEC, or in methods for training the 2D cell classifiers separately, either iii) normal or non-BEC, the other side of other known identifiers used) to assign an abnormal 2D cell classification score and a BEC2D cell classification score for each cell.

[0105] In step 540, the processor assigns a test identifier for each cell based on the abnormal 2D cell classification score and the BEC2D classification score as: i) having abnormal features, ii) having BEC-like features, or iii) not having abnormal or BEC-like features.

[0106] In step 550, each test identifier is compared to each known identifier for each cell to calculate the accuracy of both 2D cell classifiers.

[0107] In step 560, the accuracy of the 2D cell classifier may be compared to a preselected accuracy requirement, and if the accuracy requirement is met, the 2D cell classifier is designated as trained, or if the accuracy requirement is not met, in step 570, the parameters of the regression for the 2D cell classifier that does not meet the accuracy requirement are adjusted, and steps 530 to 560 are repeated for at least the adjusted 2D cell classifier.

[0108] In some embodiments, the processor tunes the 2D cell classifier algorithm using adaptively boosted logistic regression, random forests, decision trees, or any combination thereof. Adaptively boosted logistic regression uses logistic regression in an iterative loop to improve overall classification merit.

[0109] In some embodiments, the preselected 2D abnormal cell sensitivity value may be at least 90%, at least 95%, or within the ranges of 90% to 100%, 90% to 99.9%, 90% to 99%, 90% to 98%, 95% to 100%, 95% to 99.9%, 95% to 99%, or 95% to 98%, or any other 2D abnormal cell sensitivity value listed herein for the abnormal cell 2D classifier.

[0110] In some embodiments, the pre-selected 2D abnormal cell specificity value may be at least 65%, at least 70%, at least 75%, or in the range of 65% to 100%, 65% to 99%, 65% to 90%, 65% to 80%, 65% to 75%, 70% to 100%, 70% to 99%, 70% to 90%, 70% to 80%, 70% to 75%, 75% to 100%, 75% to 90%, or 75% to 80%, or any other 2D abnormal cell specificity value listed herein for the abnormal cell 2D classifier.

[0111] In some embodiments, the preselected 2D BEC sensitivity value can be at least 85%, at least 90%, at least 95%, or in the ranges of 85% to 100%, 85% to 99.9%, 85% to 99%, 95% to 90%, 90% to 100%, 90% to 99.9%, 90% to 95%, 95% to 100%, 95% to 99.9%, or 95% to 99%, or any other 2D BEC sensitivity value listed herein for the BEC2D classifier.

[0112] In some embodiments, the pre-selected 2D BEC singular value can be at least 60%, at least 65%, at least 70%, or in the ranges of 60% to 100%, 60% to 99%, 60% to 90%, 60% to 80%, 60% to 75%, 60% to 70%, 60% to 65%, 65% to 100%, 65% to 99%, 65% to 90%, 65% to 80%, 65% to 75%, 65% to 70%, 70% to 100%, 70% to 99%, 70% to 90%, 70% to 80%, or 70% to 75%, or any other 2D BEC singular value listed herein for the BEC2D classifier.

[0113] Throughout the specification, claims, and drawings, the following terms take the meaning expressly associated therewith, unless the context clearly dictates otherwise: The term "herein" refers to the specification, claims, and drawings associated with this application. The phrases "in one embodiment," "in another embodiment," "in various embodiments," "in some embodiments," "in other embodiments," and other variations thereof, refer to one or more features, structures, functions, limitations, or characteristics of the present disclosure and are not limited to the same or different embodiments, unless the context clearly dictates otherwise.

[0114] As used herein, any concentration range, percentage range, ratio range, or integer range should be understood to include any value or subrange within the recited range unless otherwise indicated. It should also be noted that the term "or" is generally used in an inclusive sense (i.e., to mean either one, both, or any combination of the alternatives) unless the context dictates otherwise. Also, as used herein and in the appended claims, the singular forms "a," "an," and "the" include plural referents unless the context dictates otherwise. The terms "comprise" and "have," as well as variations thereof, are used synonymously and should be construed as open-ended. The term "combinations thereof," as used herein, refers to all possible combinations of the items listed preceding the term. For example, "A, B, C, or combinations thereof" is intended to refer to any one of A, B, C, AB, AC, BC, or ABC. Similarly, the term "combinations thereof," as used herein, refers to all possible combinations of the items listed preceding the term. For example, "A, B, C, and combinations thereof" is intended to refer to all of A, B, C, AB, AC, BC, and ABC. [Example]

[0115] [Example 1] Training a 2D cell classifier The lung cancer detection method was performed using the CELL-CT® optical tomography system and was designed to detect lung cancer in the pre-invasive stage, when treatment is easier and more likely to be successful.

[0116] The following method was used:

[0117] 1. Patient samples for testing were fixed, stained, and enriched for BEC.

[0118] 2. The patient sample was then suspended in an oil-based optical medium. The cells in the optical medium were then inserted into a glass microcapillary tube with an inner diameter of approximately 60 μm. Pressure was applied to the medium to move the cells into the optical path of a high-magnification microscope within the optical tomography system (see Figure 1).

[0119] 3. Once the cell was in the optical path of the high-magnification microscope, the cell search mode was initiated. In cell search mode, the CELL-CT® continuously swept its focal plane through the lumen of the microcapillary tube in 1 μm steps to identify the linear, radial, and angular positions of the cell. Detected dark objects above a certain size were identified for potential 3D capture. The 2D image set was filtered to capture a central image of the cell. This 2D image was then evaluated using a 2D cell classifier to determine whether the cell possessed abnormal or BEC-like features, as described below.

[0120] 4. Each cell was also analyzed in projection image acquisition mode, during which the tube was rotated, allowing the optical tomography system to generate 500 high-resolution images of the same diagonal cross-section of the capillary tube taken through a 360° tube rotation. These images were pseudoprojection images, which are simulations of projection images generated by integrating light from the objective lens as the focal plane was swept through the nucleus of any cell included in the image. Thus, the pseudoprojection image represented the complete nuclear content within a single image obtained from a single viewpoint.

[0121] 5. Pseudoprojection images were processed to correct for residual noise and motion artifacts.

[0122] 6. The corrected pseudoprojection images were processed using filtered backprojection techniques to generate 3D tomographic representations of the cells, also known as 3D images. 3D tomographic representations were generated for a significant number of cells, including multiple cells identified within the optical medium volume in the optical path of a high-magnification microscope before the optical medium was advanced into different volumes by pressure. An example of such a 3D image for both squamous intermediate cells and dysplastic cells is shown in Figure 6.

[0123] 7. The 3D images were evaluated by a cytopathologist to determine whether the cells were abnormal, BECs, or normal, not BECs.

[0124] The abnormal 2D cell classification score indicated the likelihood that the object under investigation had abnormal characteristics. The abnormal 2D cell classification scores for several imaged cells were used to prepare a composite abnormal 2D cell classification score, which determined whether the patient was likely to have lung cancer.

[0125] The BEC2D cell classification score indicated the likelihood that the object under investigation had BEC-like characteristics. The BEC classification scores for several imaged cells were used to generate a composite BEC2D cell classification score, which helped determine how many BECs were analyzed.

[0126] The abnormal cell 2D classifier and BEC2D classifier for use in the lung cancer detection method were trained and tested by a cytopathologist using conventional known cell identification methods, using images of cells obtained as described above, to assign known identifiers to each cell that was also given a test identifier in the training method.

[0127] The development of the abnormal cell 2D classifier followed a supervised learning process governed by binary known identifiers (abnormal / normal, or with abnormal features / without abnormal features). Cells are classified according to the diagnostic categories they fall into by the cytopathologist: - Atypia, dysplasia, precancerous cells, malignant cells - Polymorphic dyskeratosis, type II pneumocytes If a specimen is confirmed to be in one of these categories, it is designated with the well-known identifier "anomalous" or "having anomalous characteristics." A cell was designated as having no abnormal features or as “normal” if it was obtained randomly and designated as normal by a cytopathologist (in the absence of the abnormal criteria specified above).

[0128] Development of the BEC2D classifier followed a supervised learning process governed by a binary known identifier (BEC / non-BEC, or with BEC-like features / without BEC-like features). When two cytotechnologists independently agreed that a cell was a BEC, the cell was designated with the known identifier "BEC" or "with BEC-like features."

[0129] Cells classified as non-BEC by any cytopathologist were designated with the known identifiers "non-BEC" or "does not have BEC-like characteristics."

[0130] For training purposes, only the designation "does not have abnormal features" was used with the abnormal cell 2D classifier, and only the designation "does not have BEC-like features" was used with the BEC 2D classifier, but even during training, cells may be designated as having abnormal features or BEC-like features.

[0131] Presence of abnormal features was designated as 1 for both the known and test identifiers. Absence of abnormal features was designated as 0 for both the known and test identifiers. A training set of 16,046 cells was imaged, and known identifiers were provided by a cytopathologist and test identifiers were provided by the abnormal cell 2D classifier. To provide test identifiers, cells and nuclei were segmented from the block of voxels containing the cells, and features were automatically measured for each cell by an optical tomography system and then used to assign a test identifier. The test identifier for each cell was then compared to the known identifier to determine whether the test identifier was correct. This aggregate data set formed the training set for creating the abnormal cell 2D classifier. Adaptively boosted logistic regression (AdaBoost) was used to refine the abnormal cell 2D classifier. AdaBoost uses logistic regression in an iterative loop to improve overall classification merit. Cross-validation was used to limit the complexity of the model and thereby ensure its generalization.

[0132] The trained abnormal cell 2D classifier was used to classify cells that were not part of the training set. The results showing the known identifiers and corresponding test identifiers for the trained abnormal cell 2D classifier are shown in Table 1 and Figure 7.

[0133] The 2D abnormal cell sensitivity was 95% and the 2D abnormal cell specificity was 74.7%. For an abnormal cell 2D classifier to be considered trained, the sensitivity was required to be greater than 90% and the specificity was required to be greater than 65%, and therefore the trained abnormal cell 2D classifier exceeded a preselected accuracy value.

[0134] [Table 1]

[0135] Figure 7 shows the cumulative probability density curve for the abnormal cell 2D classifier trained using the results from Table 1. For cells with known identifiers for abnormal features, 93.3% (1-0.067) received the correct test identifier (red line) and were then imaged in 3D. 96.2% of cells with known identifiers for BEC-like features (green line) were then identified by the BEC2D classifier, and 87.5% of cells with known identifiers without BEC-like features (light blue line) were then identified by the BEC2D classifier. Thus, ultimately, 12.5% of the heterogeneous cells (cells without abnormal or BEC-like features) were imaged in 3D.

[0136] Possession of BEC-like features was designated as 1 for both the known and test identifiers. Absence of BEC-like features was designated as 0 for both the known and test identifiers. A training set of 12,572 cells was imaged to provide known identifiers by cytopathologists and test identifiers by the BEC2D classifier. To provide test identifiers, cells and nuclei were segmented from the block of voxels containing the cells, and features were automatically measured by an optical tomography system for each cell and then used to assign a test identifier. The test identifier for each cell was then compared to the known identifier to determine whether the test identifier was correct. This aggregate data set formed the training set for creating the BEC2D classifier. Adaptively boosted logistic regression (AdaBoost) was used to refine the abnormal cell 2D classifier. AdaBoost uses logistic regression in an iterative loop to improve the overall classification merit. Cross-validation was used to limit the complexity of the model and thereby ensure its generalization.

[0137] The trained BEC2D classifier was used to classify cells that were not part of the training set. The results showing the known identifiers and corresponding test identifiers for the trained BEC cell classifier are shown in Table 2 and Figure 8.

[0138] The 2D BEC sensitivity was 88% and the 2D BEC specificity was 66%. For a BEC2D classifier to be considered trained, sensitivity was required to be greater than 85% and specificity was required to be greater than 60%, and therefore the trained BEC2D classifier exceeded a preselected accuracy value.

[0139] [Table 2]

[0140] Figure 8 shows the cumulative probability density curve for the BEC2D classifier trained using the results from Table 2. For cells with known identifiers that have BEC-like features, 86.6% (1-0.134) received the correct test identifier (green line) and were then imaged in 3D. For cells with known identifiers that do not have BEC-like features, 81% received the correct test identifier (light blue line).

[0141] The cumulative results of Figures 7 and 8 are as follows: 93.3% of cells with abnormal features (based on known identifiers) were imaged in 3D. 5.8% (1-0.962) + 0.866 * 0.962 * 100% = 89% of cells with BEC-like features (based on known identifiers) were imaged in 3D. 0.875*0.81*100%=71% of cells with no abnormal or BEC-like features were imaged in 2D only.

[0142] [Example 2] Evaluation of 2D cell classifiers To ensure robustness, the trained 2D cell classifier underwent three independent tests after development. In each test, the lung cancer detection method was run using the trained 2D classifier and the trained 3D classifier, but as in the training of Example 1, 3D images were acquired for all objects for which 2D images were acquired so that accuracy could be evaluated.

[0143] Algorithm Verification Research

[0144] The trained 2D cell classifier was embedded into executable code that also implemented the rest of the standard lung cancer test and tested using cells from five adenocarcinoma and two squamous cell carcinoma cell lines. Various measures of accuracy were evaluated. The abnormal cell referral rate was 98.7%. The BEC2D classifier demonstrated a BEC2D sensitivity of 90% and a BEC2D specificity of 44%. Pilot Clinical Accuracy Study (CAS)

[0145] A pilot CAS was conducted using the same set of specimens used to train the 2D cell classifier to evaluate the referral rate of the 2D cell classifier. 34 specimens, represented by 267,659 cells, were examined. The abnormal cell referral rate was calculated to be 95%. The BEC referral rate was 91%. The abnormal cell referral rate and BEC referral rate indicated that training the 2D algorithm generalized well to a population of cells that were not part of the training process. Case-level accuracy (whether a patient was positive or negative for cells indicative of lung cancer) was unchanged by the use of the 2D cell classifier and the exclusion of some cells from the 3D imaging. CAS-A

[0146] A CAS-A study was conducted to further evaluate the performance of the 2D cell classifier. 38 specimens represented by 757,627 cells were examined as in the pilot CAS. The abnormal cell referral rate was calculated to be 90% by pooling the data with those from the pilot CAS, as there were too few abnormal cells in the CAS-A. The BEC referral rate was calculated to be 87%.

[0147] The 2D rejection rate was not specified, but it was estimated to be 53% during training of the 2D cell classifier. The measured 2D rejection rate in the CAS-A study was 60%, with a median 2D rejection rate by case of 52%. Summary of clinical results

[0148] Overall, the abnormal cell referral rate and BEC referral rate met specifications, while the 2D rejection rate exceeded the values measured in training, indicating that the training of the 2D cell classifier generalized well to populations of cells that were not part of the training process.

[0149] [Example 3] Test duration A lung cancer detection method according to the present disclosure was performed that uses a 2D cell classifier to exclude some cells from 3D imaging, and when the optical tomography image acquisition portion of the AI-based cell classification method was compared to the optical tomography image acquisition portion of a conventional cell classification method in which all cells were imaged in 3D, a 52% reduction in cells to 3D images per patient was observed, along with a 45% reduction in method duration, while maintaining preselected accuracy.

[0150] The various embodiments described above may be combined to provide further embodiments. All U.S. patents, U.S. patent application publications, U.S. patent applications, foreign patents, foreign patent applications, and non-patent publications mentioned herein and / or listed in the Application Data Sheet, including U.S. Patent Application No. 63 / 394,550, filed August 2, 2022, are incorporated herein by reference in their entirety. Aspects of the embodiments may be modified, as necessary, to employ concepts from various patents, applications, and publications to provide still further embodiments.

[0151] These and other changes can be made to the embodiments in light of the above detailed description. Generally, in the appended claims, the terms used should not be construed to limit the claims to the specific embodiments disclosed in the specification and claims, but should be construed to include all possible embodiments, along with the full range of equivalents to which such claims are entitled. Accordingly, the claims are not limited by this disclosure.

Claims

1. a) an optical tomography system generating representative 2D images of cells from a patient sample containing a plurality of cells; b) evaluating the representative 2D images using an abnormal cell 2D classifier to determine whether the cells have abnormal characteristics; c) evaluating the representative 2D images using a BEC2D classifier to determine whether the cells have BEC-like characteristics; and d) if the cell is determined to have abnormal or BEC-like characteristics, the optical tomography system generates a 3D image of the cell. An artificial intelligence (AI)-based cell classification method, comprising:

2. e) repeating a) through d) for a subset of cells within said plurality of cells to generate patient sample data reflecting a total number of enumerated analyzed cells having abnormal characteristics, a total number of enumerated analyzed cells having BEC-like characteristics, or a total number of enumerated analyzed cells. The method of claim 1 further comprising:

3. f) comparing the total number of enumerated analyzed cells having BEC-like characteristics to a threshold to determine whether the total number of enumerated analyzed cells having BEC-like characteristics is equal to or greater than the threshold and therefore sufficient for an accurate assay, or whether the total number of enumerated analyzed cells having BEC-like characteristics is less than the threshold and therefore insufficient for an accurate assay. The method of claim 2 further comprising:

4. 4. The method of claim 3, further comprising repeating steps a) through f) until the total number of enumerated analyzed cells having BEC-like characteristics is sufficient to detect the presence of cells indicative of lung cancer in the patient sample with a preselected accuracy.

5. 5. The method of claim 4, wherein the threshold is preselected by the number of enumerated cells having BEC-like characteristics that are statistically capable of detecting the presence of cells indicative of lung cancer in the patient sample with the preselected accuracy.

6. 6. The method of claim 1, further comprising, prior to a), enriching the patient sample for BECs or for cells likely to be indicative of cancer.

7. 6. The method of claim 1, wherein generating the 2D image of the cell by the optical tomography system comprises: the optical tomography system sweeping a focal plane of the optical tomography system in 1 μm steps across a single cell to generate single-plane 2D images of the single cell stepwise; compiling a set of single-plane 2D images of the single cell; and filtering the set of single-plane 2D images of the single cell to generate the representative 2D image of the cell.

8. 6. The method of claim 1, wherein evaluating the representative 2D images with an abnormal cell 2D classifier and evaluating the representative 2D images with a BEC 2D classifier both comprise determining values ​​for a plurality of cell feature measurements.

9. 1. A method for training an AI-based cell classification system, comprising: a) operating an optical tomography system to generate a representative 2D image of a cell; b) assigning a known identifier to each of a plurality of said cells using said representative 2D images of such cells and known cell identification methods, said known identifier being i) abnormal or having abnormal characteristics, ii) BEC or having BEC-like characteristics, or iii) normal or non-BEC or not having abnormal or BEC-like characteristics; c) calculating a plurality of cell feature measurements for each of said plurality of cells from said representative 2D images of such cells; d) performing a regression between the cell feature measurements and the known identifier of each cell to assign an abnormal 2D cell classification score and a BEC2D cell classification score for each cell; e) assigning a test identifier for each cell that: i) has abnormal characteristics, ii) has BEC-like characteristics, or iii) does not have abnormal or BEC-like characteristics based on the abnormal 2D cell classification score and the BEC 2D classification score of the cell; f) comparing the respective test identifiers with the known identifiers of each cell to calculate the accuracy of the abnormal cell 2D classifier and the BEC 2D classifier; A method comprising:

10. 10. The method of claim 9, further comprising comparing the accuracies of both of the 2D cell classifiers to a preselected accuracy requirement, and designating the 2D cell classifier as trained if the preselected accuracy requirement is met by the 2D cell classifier.

11. 11. The method of claim 9 or 10, further comprising comparing the accuracies of both of the 2D cell classifiers to a preselected accuracy requirement, and if the preselected accuracy requirement is not met by a 2D cell classifier, adjusting parameters of the regression for the affected 2D cell classifier, and repeating steps d) to f).

12. 11. The method of claim 9 or 10, wherein the regression comprises adaptively boosted logistic regression, a random forest, a decision tree, or any combination thereof.

13. 11. The method of claim 9 or 10, further comprising generating 3D images of a plurality of said cells and using said 3D images in said known cell identification method.

14. 11. The method of claim 9 or 10, wherein the cellular feature measurements comprise object shape features, cell shape features, cytoplasmic features, cell nuclear features, chromatin distribution, nuclear size features, nuclear texture features, other morphometric elements, or any combination thereof.

15. 1. A cell sorting system comprising: an optical tomography system, generating a representative 2D image of cells from a patient sample; generating a 3D image of the cell if the cell has abnormal or BEC-like characteristics; an optical tomography system operable to: a processor, comparing the representative 2D image of the cell with an abnormal cell 2D classifier to determine whether the cell has abnormal characteristics; comparing the representative 2D image of the cell with a BEC 2D classifier to determine whether the cell has BEC-like characteristics; a processor and A cell classification system comprising:

Images