System and methods for artificial intelligence inference of cell phenotype using an integrated biosensor

Abstract

A biosensor comprises a complementary metal-oxide semiconductor (CMOS) image sensor (CIS), wherein the CIS comprises a plurality of light sensitive diodes disposed over a silicon substrate, and one or more metal layers for interconnect disposed therein above, and wherein the silicon substrate is back-lapped to a predetermined thickness. It further comprises a microwell layer disposed over the back-lapped silicon layer having etched therein an array of microwells, each microwell having a bottom which is the back-lapped silicon substrate and is further adapted to capture therein one or more biological cells. The biosensor can be used to train an artificial intelligence (AI) engine to generate an AI model for inference or prediction of a phenotype of a biological cell captured by the biosensor. The training is performed on images from both a fluorescence microscope and the biosensor.

Description

SYSTEM AND METHODS FOR ARTIFICIAL INTELLIGENCE INFERENCE OF CELL PHENOTYPE USING AN INTEGRATED BIOSENSORCROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims the benefit of US Provisional Application No. 63 / 725,207 filed November 26, 2024, the contents of which are hereby incorporated by reference.TECHNICAL FIELD

[0002] The present disclosure relates generally to the training of an artificial intelligence (Al) model and use of the generated model for inference of cell phenotype and more specifically a biosensor for such inference purposes in conjunction with the Al model and generation thereof.BACKGROUND

[0003] Accurate, high-throughput cell analysis is critical in biotechnology, healthcare, and pharmaceutical industries, driving innovations in diagnostics, drug development, and personalized medicine. However, traditional methods like flow cytometry and microscopy often fall short in scalability, data integrity, and predictive capabilities, limiting their utility in real-world applications where rapid, actionable insights are needed.

[0004] The area of complementary metal-oxide semiconductor (CMOS) image sensors (IS) has developed rapidly over the past two decades and such backlit devices are commonly manufactured. The process of preparing such an CMOS IS (CIS) involves manufacturing diodes on a silicon substrate, and thereafter depositing above them several layers of metal, using processes well-known in the semiconductor industry. Thereafter, the wafer on which the device is manufactured is essentially flipped, and a back lapping process takes place. The substrate is reduced in thickness to about 2-3 pm which allows light to penetrate and reach the light sensitive diodes. In certain cases, an optical filter may be added on that back-lapped bottom side, and / or lenses may be further deposited. The result is a backlit CIS.

[0005] Characterization of live cell phenotypes is currently done mostly by flow cytometers and fluorescence microscopes by observing cell morphological features combined with quantification of fluorescent reporters and dyes which indicate cell cycle stage, maturationand activation status, differentiation etc. Flow cytometry cannot observe cells over time, while fluorescence microscopy is expensive, necessitates proficiency, and allows measurement of cells in low throughput.

[0006] Artificial intelligence (Al) models have emerged as powerful tools for phenotype inference in biological cells, revolutionizing the way researchers analyze and interpret cellular characteristics. Phenotypes, defined as observable traits or behaviors resulting from gene expression and environmental interactions, are critical to understanding cellular functions and diseases. Al models, particularly those employing machine learning (ML) and deep learning (DL) techniques, have significantly enhanced the speed and accuracy of phenotype analysis.

[0007] One prominent application of Al in this field involves image-based analysis. High- content screening (HCS) platforms generate large-scale microscopic images of cells, which are then analyzed by Al algorithms to extract phenotype-related features. Convolutional neural networks (CNNs), a type of DL model, are frequently used to classify cellular morphologies, detect subtle variations in shape or structure, and identify unique phenotypic markers. For example, Al models can distinguish between healthy and diseased cell phenotypes or predict responses to drug treatments based on changes in cell appearance.

[0008] Additionally, Al models integrate multimodal data sources to infer phenotypes more comprehensively. By combining genomic, transcriptomic, and proteomic datasets, machine learning algorithms can correlate gene expression profiles with phenotypic traits. This holistic approach is particularly beneficial for understanding complex cellular systems, such as cancer biology or stem cell differentiation.

[0009] Al-driven phenotype inference has applications in drug discovery, personalized medicine, and diagnostics. For instance, researchers use these models to predict how individual cells will respond to therapeutic interventions, enabling targeted and effective treatments. However, challenges remain, including the need for robust training datasets, the potential for bias in algorithms, the complexity of interpreting Al-generated insights, and the need for better sets of images, captured from microwell arrays, that can provide improved detection. It would therefore be advantageous to provide a solution thatcombines the advantages of a CIS with the capabilities of Al to overcome the challenges noted above.SUMMARY

[0010] A summary of several example embodiments of the disclosure follows. This summary is provided for the convenience of the reader to provide a basic understanding of such embodiments and does not wholly define the breadth of the disclosure. This summary is not an extensive overview of all contemplated embodiments and is intended to neither identify key or critical elements of all embodiments nor to delineate the scope of any or all aspects. Its sole purpose is to present some concepts of one or more embodiments in a simplified form as a prelude to the more detailed description that is presented later. For convenience, the term “certain embodiments” may be used herein to refer to a single embodiment or multiple embodiments of the disclosure.

[0011] Some example embodiments disclosed herein include a biosensor comprising: a complementary metal-oxide semiconductor (CMOS) image sensor (CIS), wherein the CIS comprises a plurality of light sensitive diodes disposed over a silicon substrate, and one or more metal layers for interconnect disposed therein above, and wherein the silicon substrate is back-lapped to a predetermined thickness; and, a microwell layer disposed over the back-lapped silicon layer having etched therein an array of microwells, each microwell having a bottom which is the back-lapped silicon substrate and is further adapted to capture therein one or more biological cells.

[0012] Some example embodiments disclosed herein also include a system for training an artificial intelligence model comprising: a fluorescence microscope, wherein the fluorescence microscope further comprises a fluorescence light source and a microscope camera; a biosensor a complementary metal-oxide semiconductor (CMOS) image sensor (CIS), wherein the CIS comprises a plurality of light sensitive diodes disposed over a silicon substrate, and one or more metal layers for interconnect disposed therein above, and wherein the silicon substrate is back-lapped to a predetermined thickness, and, a microwell layer disposed over the back-lapped silicon layer having etched therein an array of microwells, each microwell having a bottom which is the back-lapped silicon substrate and is further adapted to capture therein one or more biological cells; and, an artificialintelligence (Al) engine for image processing, communicatively connected to the microscope camera and to the biosensor for receipt of images of cells captured in one or more microwells of the biosensor; wherein the Al engine trains a model based on a plurality of images received from the microscope camera and the biosensor, such that upon completion of a training session at least a phenotype of cells captured in one or more images received from the biosensor may be identified independent of the fluorescence microscope.

[0013] Some example embodiments disclosed herein also include a computerized method for training an artificial intelligence (Al) model to identify a biological cell phenotype based on an image from a biosensor, the method comprises: collecting from a fluorescence microscope and the biosensor a plurality images of captured cells in microwells of the biosensor; training an Al engine by at least multiplexing the plurality of images; generating the Al model that upon receipt of an image from the biosensor it may identify at least a cell phenotype of a cell captured in a microwell of the biosensor; wherein the biosensor comprises: a complementary metal-oxide semiconductor (CMOS) image sensor (CIS), wherein the CIS comprises a plurality of light sensitive diodes disposed over a silicon substrate, and one or more metal layers for interconnect disposed therein above, and wherein the silicon substrate is back-lapped to a predetermined thickness; and, a microwell layer disposed over the back-lapped silicon layer having etched therein an array of microwells, each microwell having a bottom which is the back-lapped silicon substrate and is further adapted to capture therein one or more biological cells.

[0014] Some example embodiments disclosed herein also include a computerized method for identification of a biological cell’s phenotype based on an image from a biosensor, the method comprises: capturing an image of a plurality of biological cells captured by a biosensor; identifying at least a phenotype of a cell of the plurality of biological cells using an artificial intelligence (Al) model; and, reporting the identified phenotype; wherein the biosensor comprises: a complementary metal-oxide semiconductor (CMOS) image sensor (CIS), wherein the CIS comprises a plurality of light sensitive diodes disposed over a silicon substrate, and one or more metal layers for interconnect disposed therein above, and wherein the silicon substrate is back-lapped to a predetermined thickness; and, a microwell layer disposed over the back-lapped silicon layer having etched thereinan array of microwells, each microwell having a bottom which is the back-lapped silicon substrate and is further adapted to capture therein one or more biological cells; and, wherein the Al model was generated by training an Al engine by at least multiplexing images of biological cells captured by the biosensor and a fluorescence microscope.BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The subject matter disclosed herein is particularly pointed out and distinctly claimed in the claims at the conclusion of the specification. The foregoing and other objects, features, and advantages of the disclosed embodiments will be apparent from the following detailed description taken in conjunction with the accompanying drawings.

[0016] Figure 1 A is a cross-section of an integrated biosensor according to an embodiment.

[0017] Figure 1 B is a top view of an integrated biosensor according to an embodiment.

[0018] Figure 1 C is a cross-section of an integrated biosensor showing a single well according to an embodiment.

[0019] Figure 2 is a cross-section of an integrated biosensor using an excitation filter according to an embodiment.

[0020] Figure 3 is a system using a biosensor to generate a model using artificial intelligence training according to an embodiment.

[0021] Figure 4 is a flowchart of generating an artificial intelligence model using a biosensor and a fluorescence microscope for inference or prediction of a cell phenotype according to an embodiment.

[0022] Figure 5 is a flowchart of using an artificial intelligence generated model for inference or prediction of a cell phenotype according to an embodiment.

[0023] Figure 6 is a flowchart of a process for manufacturing a biosensor according to an embodiment.DETAILED DESCRIPTION

[0024] It is important to note that the embodiments disclosed herein are only examples of the many advantageous uses of the innovative teachings herein. In general, statements made in the specification of the present application do not necessarily limit any of the various claimed embodiments. Moreover, some statements may apply to some inventivefeatures but not to others. In general, unless otherwise indicated, singular elements may be in plural and vice versa with no loss of generality. In the drawings, like numerals refer to like parts through several views.

[0025] A biosensor comprises a complementary metal-oxide semiconductor (CMOS) image sensor (CIS), wherein the CIS comprises a plurality of light sensitive diodes disposed over a silicon substrate, and one or more metal layers for interconnect disposed therein above, and wherein the silicon substrate is back-lapped to a predetermined thickness. The biosensor further comprises a microwell layer disposed over the back-lapped silicon layer having etched therein an array of microwells, each microwell having a bottom which is the back-lapped silicon substrate and is further adapted to capture therein one or more biological cells. The biosensor may be used to train an artificial intelligence (Al) engine to generate an Al model for inference or prediction of a phenotype of a biological cell captured by the biosensor. The training is performed on images from both a fluorescence microscope and the biosensor.

[0026] Reference is now made to Figs. 1A, 1 B, and 1 C. Fig. 1A depicts an example crosssection 100A of an integrated biosensor 100 according to an embodiment. Fig. 1 B depicts an example top view 100B of an integrated biosensor 100 according to an embodiment. Fig. 1 C depicts an example cross-section 100C of an integrated biosensor 100 showing a single well (also referred to herein as microwell) according to an embodiment. In an embodiment, the integrated biosensor 100 comprises a complementary metal-oxide semiconductor (CMOS) structure 102 comprising an image sensor 120 and a microwell array 101 , the microwell array 101 comprising a plurality of wells designed to capture therein biological cells (also referred to herein as cell or cells as the case may be). Each well 160 comprises vertical sidewalls 140 that form the well 160, where one or more cells may be captured. Moreover, an emission filter 150 layer is at a bottom of the well 160.

[0027] Accordingly, in an example embodiment of biosensor 100, an intrinsic silicon or opaque polymer microwell array 101 , comprising a plurality of wells 160, is deposited on the backside of a CMOS image sensor (CIS) 120 having a plurality of individual pixels 122. Each well 160 is defined by the sidewalls of structure 140. In an embodiment, a plurality of individual pixels 122 are organized such that they are located under each well 160 and not under the structures 140 that defines the sidewalls. In an exampleembodiment, the sidewall structures 140 have a wall height of, for example, 20pm (range may vary between, for example 15-25pm), of the well array. In another embodiment, the pixels (or diodes) 122 are simply placed in a predetermined pattern within the image sensor 120 including under the structures 140. By having a plurality of pixels 122 under each well, a higher resolution image is provided. In an embodiment, the width of the well 160 is at least four times the width of a pixel 122, thereby ensuring at least 16 pixels under each well 160. The CIS 120 may be fabricated using standard state-of-the-art fabrication process of a backside illuminated CIS. The biosensor may further comprise an electronic circuitry 110 that provides interconnect circuitry as well as a processing circuitry for the CIS 120.

[0028] In an embodiment, the CIS 120 and the silicon / opaque polymer well array 101 will be separated by a thin, for example, 2pm, silicon layer 130 that remains after the silicon wafer thinning process (also known as back lapping) at the end of the CIS 120 fabrication process. As an example, the thin silicon layer 130 may vary between 2-3pm. In an embodiment, the pixel size of the CIS 120 will be about 1 pm or less. White light illuminated from the top, i.e., from the direction of microwell array 101 , would be scattered from the cells dispersed in the microwells 160 and reach the sensors 122. Image acquisition is based on the standard operation of the CIS 120. Electric wiring connected to the bottom of each pixel (the standard metal interconnection of the backside illuminated CIS) allows collecting the data from each pixel and integrating it into an image with or without any additional processing. The diffracted light may create a hologram, or shadow, on the CMOS pixels 122 which will contain information about the three-dimensional (3D) structure of the cell. This Near Field imaging feature offers an advantage compared to regular two-dimensional (2D) microscopy which provides information mainly on the focus plane.

[0029] Reference is now made to Figure 2 depicting an example cross-section 200 of an integrated biosensor using an excitation filter 220 according to an embodiment. An emission filter 150 may be placed at the bottom of each well 160. Such configuration allows for adaptation of the biosensor 100 for fluorescence measurements by adding an excitation filter 220 between a light source 210 and cells 240 trapped within the plurality of wells 160. The emission filter 150, positioned between the cells 240 and the pixels 122,is excited by the emission wavelength 250 of the cells 240. Such design may also be used for quasi-brightfield imaging by illuminating the apparatus with a center wavelength of a bandpass filter.

[0030] In an embodiment, the pixels 122 of different colors may be positioned under the same well. Such configuration is made possible by ensuring that there is a plurality of pixels 122 under each well 160 as further discussed hereinabove. As a result, several emission wavelengths may be measured simultaneously by different color pixels 122 under the same well 160. In an embodiment, the emission filter layer 150 may be replaced or contain a polarization layer that enables fluorescence polarization measurements.

[0031] Reference is now made to Fig. 3 that shows an example system 300 using a biosensor 100A to generate a model using artificial intelligence training according to an embodiment. The training system includes a fluorescence microscope (only relevant parts of which are shown for ease of description) that images the biosensor 110A with simultaneous holographic detection from pixels 122. Light in a specific wavelength arrives from a light source 310 in the fluorescence microscope, excites the sample, and continues to the pixels 122 after being scattered from cells 240. This causes fluorescence emission from the cells 240 on the one hand and acts as a brightfield imaging for the CIS 102. The emitted fluorescence passes through a dichroic mirror 320 on its way to the microscope’s camera 330. The images collected from the CIS 102 (brightfield images) and the microscope camera 330 (fluorescence images) are collected simultaneously.

[0032] In concert with the image capture done by the fluorescence microscope using the microscope camera 330, the biosensor captures brightfield images using the light sensitive diodes (of the pixels) 122. The images captured by the microscope camera 330 are transferred via link 350 to an Al engine 340 for image processing. The images captured by the biosensor light sensitive diodes 122 are transferred via link 360 to the Al engine 340 for image processing. The Al engine is set to generate an Al model that may infer or predict phenotypes of cells based on the images, preferably simultaneously, collected by the microscope camera 340 and the biosensor 100. After training, the phenotypes of the cells may be inferred or predicted by applying the trained model to the images captured by the biosensor 110A.

[0033] The simultaneous image collection, which may be timestamped, is used for training an Al model. The generated model enables predicting fluorescence data (which is correlated to specific phenotypes) from images captured from CIS 102 alone. In an embodiment, a time period is permitted between a timestamp of images used by the model from different image capture sources. By using the CIS 102 at steady-state operation, use of the more complex, time consuming, and costly fluorescence microscope is eliminated. In an embodiment, a retraining is performed periodically to ensure accuracy of the prediction.

[0034] It should therefore be understood that the use of Al, and machine learning in particular, to generate a model, combined with the ability to accurately collect multi-layered data of live cells in high throughput, offers an opportunity to infer the phenotype of a cell based on its morphology, i.e., its brightfield images. Training of the Al generated model is performed by multiplexing the brightfield images and the fluorescence microscope images of the biosensor 110A, with the data simultaneously captured by the CIS 102 and the fluorescence microscope using the same illumination from the microscope. This ensures that the two sets of images are captured under the same illumination conditions. The result of training is an Al model that is deployed onto an autonomous biosensor 110A, i.e., one which operates independently from a fluorescence microscope. The biosensor 110A, after case-specific sample preparation, is then used to image cells 240 under white illumination. Based on the data collected by the CIS 102, the Al model is then used to infer and make predictions on the current or future phenotype of the cells 240 that were imaged.

[0035] Reference is now made to Fig. 4 which is an example flowchart 400 of generating an artificial intelligence-based model using a biosensor and a fluorescence microscope for inference or prediction of a cell phenotype according to an embodiment.

[0036] In S410, a system comprising a biosensor, for example biosensor 110A, and a fluorescence microscope, is used to simultaneously generate images of cells, for example cells 240, that are captured in microwells 160 of biosensors 110A. In an embodiment, the images from both sources are timestamped to ensure correlation between images captured from both sources. Furthermore, in an embodiment, the process of imagecapturing is performed in real-time, continuously or periodically. In yet another embodiment, a range of time between timestamped images is permitted.

[0037] In S420, an Al engine, using for example a machine learning algorithm, is trained using the images received from both sources.

[0038] In S430, a model is generated based on the images collected from both sources. The model allows for the prediction or inference of phenotypes of cells that are captured, for example at the biosensor 100A, without the use of a fluorescence microscope.

[0039] In S440, the generated model is stored, for example in memory (not shown) for use at the inference or prediction phase as further discussed in Fig. 5.

[0040] Reference is now made to Fig. 5 that depicts an example flowchart 500 of using an artificial intelligence generated model for inference or prediction of a cell phenotype according to an embodiment.

[0041] In S510, a brightfield image is received, for example from a biosensor 110A. The biosensor 110A does not make use of a fluorescence microscope and the source of light for imaging may simply be a white light source.

[0042] In S520, an Al generated model for inference or prediction of cell phenotypes generated according to principles disclosed herein is used to detect phenotypes of one or more cells that appear in the received image.

[0043] In S530, based on detection of possible phenotypes the inferred, typically present, or predicted, typically in the future, phenotypes of each of the one or more cells is determined, using the model.

[0044] In S540, the inferred or predicted phenotypes of the one or more cells are provided.

[0045] One of ordinary skill in the art would appreciate that the use of a laser light source is also possible without departing from the scope of the disclosed embodiments.

[0046] Reference is now made to Fig. 6 showing an example flowchart 600 of a process for manufacturing of a biosensor according to an embodiment. The process uses initially a standard manufacturing process of a CIS and then continues with the elements that are unique to the biosensors, for example biosensor 100 described in greater detail herein.

[0047] In S610, diodes are manufactured on a silicon substrate, and thereafter several layers of metal are deposited above the diode, using processes well-known in the semiconductor industry, and therefore not repeated herein. In an embodiment, electronic circuits,including CMOS transistors, may be further integrated without departing from the scope of the disclosed embodiments. Such circuits may include circuits for gathering electronic signals from the diodes and further perform analog and / or digital manipulations thereof.

[0048] In S620, the wafer on which the devices (CIS) are manufactured is essentially flipped, and a back lapping process is performed. The substrate is reduced in thickness to about 2-3 pm which allows light to penetrate and reach the light sensitive diodes. The back- lapping is performed to allow light to reach the diodes without damaging the structural integrity of the device. A typical back-lapping process will reduce the thickness of the silicon layer 130 (for example, see Fig. 1 C) to a thickness of 2-3pm depending, for example, on considerations of structural integrity and transparency.

[0049] In S630, a well layer, having a thickness of around 20 pm, is deposited over the back- lapped back side of the wafer. An opaque polymer, of a type compatible with both attaching to the silicon on one hand and being compatible with accepting cells in wells bored therein, is used.

[0050] In S640, a lithography process is used for the purpose of defining the microwells of the array of the biosensor. In an embodiment, an alignment is performed to ensure that each microwell is defined over a respective array of diodes.

[0051] In S650, the well layer is etched to open the microwells where defined. In an embodiment, the back-lapped silicon surface is used as a stop barrier for the etch.

[0052] Additional steps may be added to this flow. For example, an optical filter layer may be deposited at the bottom of one or more microwells of the plurality of microwells. In another embodiment, a lens layer may be deposited at the bottom of one or more microwells of the plurality of microwells.

[0053] It should be appreciated that the solution described herein provides a number of advantages over known solutions. Firstly, the use of fluorescence microscopy on a regular basis is costly as it requires expensive equipment and complicated setups in order to capture many images of the biological cells captured in microwells of a carrier chip of the known techniques. Secondly, while the initial training of the Al model employs fluorescence microscopy, in the production mode, i.e., the steady-state operation after the Al model has been established, the flow operates apart from the fluorescence microscopy, and the biosensor described herein is simply used for both training andsteady-state operations. Thirdly, the biosensor described herein is a portable device and may be deployed for image capturing of cells in environments where the use of fluorescence microscopy is impractical, difficult, or impossible. Fourthly, the same biosensor is used for both training and steady-state imaging, thereby providing improved accuracy over solutions that may use different capturing devices for training of an Al model and the steady-state operation. Fifthly, the biosensor provides lens free imaging of the cells. One of ordinary skill in the art would readily appreciate that the only difference between training and inference / prediction is that, during training, additional data is collected from the fluorescent images. The brightfield images used during training and the brightfield images taken at the steady-state for inference / prediction are captured using the same mechanisms.

[0054] All examples and conditional language recited herein are intended for pedagogical purposes to aid the reader in understanding the principles of the disclosed embodiment and the concepts contributed by the inventor to furthering the art and are to be construed as being without limitation to such specifically recited examples and conditions. Moreover, all statements herein reciting principles, aspects, and embodiments of the disclosed embodiments, as well as specific examples thereof, are intended to encompass both structural and functional equivalents thereof. Additionally, it is intended that such equivalents include both currently known equivalents as well as equivalents developed in the future, i.e., any elements developed that perform the same function, regardless of structure.

[0055] It should be understood that any reference to an element herein using a designation such as “first,” “second,” and so forth does not generally limit the quantity or order of those elements. Rather, these designations are generally used herein as a convenient method of distinguishing between two or more elements or instances of an element. Thus, a reference to first and second elements does not mean that only two elements may be employed there or that the first element must precede the second element in some manner. Also, unless stated otherwise, a set of elements comprises one or more elements.

[0056] As used herein, the phrase “at least one of” followed by a listing of items means that any of the listed items can be utilized individually, or any combination of two or more of

Claims

the listed items can be utilized. For example, if a system is described as including “at least one of A, B, and C,” the system can include A alone; B alone; C alone; 2A; 2B; 2C; 3A; A and B in combination; B and C in combination; A and C in combination; A, B, and C in combination; 2A and C in combination; A, 3B, and 2C in combination; and the like.CLAIMSWhat is claimed is:

1. A biosensor comprising: a complementary metal-oxide semiconductor (CMOS) image sensor (CIS), wherein the CIS comprises a plurality of light sensitive diodes disposed over a silicon substrate, and at least one metal layer for interconnect disposed therein above, and wherein the silicon substrate is back-lapped to a predetermined thickness; and, a microwell layer disposed over the back-lapped silicon substrate having etched therein an array of microwells, each microwell having a bottom which is the back- lapped silicon substrate and wherein each microcell of the array of microcells is further adapted to capture therein one or more biological cells.

2. The biosensor of claim 1 , wherein the bottom of at least one microcell of the array of microcells has an optical filter layer.

3. The biosensor of claim 1 , wherein the bottom of at least one microcell of the array of microcells has a lens.

4. The biosensor of claim 1 , wherein the light sensitive diodes are disposed in a plurality of groups and wherein each of the microwells is disposed over a corresponding group of the light sensitive diodes.

5. The biosensor of claim 1 , further comprising: an analog electronic circuitry adapted to receive electronic signals from the light sensitive diodes.

6. The biosensor of claim 1 , further comprising: a digital circuitry for at least processing electronic signal of the light sensitive diodes.

7. The biosensor of claim 6, wherein the digital circuitry is communicatively connected to an analog electronic circuitry.

8. A system for training an artificial intelligence model comprising: a fluorescence microscope, wherein the fluorescence microscope further comprises a fluorescence light source and a microscope camera; a biosensor comprising a complementary metal-oxide semiconductor (CMOS) image sensor (CIS), wherein the CIS comprises a plurality of light sensitive diodes disposed over a silicon substrate, and at least one or more metal layers for interconnect disposed therein above, and wherein the silicon substrate is back-lapped to a predetermined thickness, and, a microwell layer disposed over the back-lapped silicon substrate having etched therein an array of microwells, each microwell having a bottom which is the back-lapped silicon substrate and is wherein each microwell of the array of microwells is further adapted to capture therein one or more biological cells; and, an artificial intelligence (Al) engine configured for image processing, communicatively connected to the microscope camera and to the biosensor for receipt of images of cells captured in one or more microwells of the biosensor; wherein the Al engine trains a model based on a plurality of images received from the microscope camera and the biosensor, such that upon completion of a training session at least a phenotype of cells captured in one or more images received from the biosensor may be identified independent of the fluorescence microscope.

9. The system of claim 8, wherein the training of the model comprises at least multiplexing of images collected from the fluorescence microscope and the biosensor.

10. The system of claim 9, wherein the multiplexed images are captured simultaneously by the fluorescence microscope and the biosensor.11 . The system of claim 9, wherein the multiplexed images are checked for having essentially a timestamp which is within a predetermined time period.

12. A computerized method for training an artificial intelligence (Al) model to identify a biological cell phenotype based on an image from a biosensor, the method comprises: collecting, from a fluorescence microscope and the biosensor, a plurality of images of captured cells in microwells of the biosensor; training an Al engine by at least multiplexing the plurality of images; generating the Al model that upon receipt of an image from the biosensor it may identify at least a cell phenotype of a cell captured in a microwell of the biosensor; wherein the biosensor comprises: a complementary metal-oxide semiconductor (CMOS) image sensor (CIS), wherein the CIS comprises a plurality of light sensitive diodes disposed over a silicon substrate, and one or more metal layers for interconnect disposed therein above, and wherein the silicon substrate is back-lapped to a predetermined thickness; and, a microwell layer disposed over the back-lapped silicon substrate having etched therein an array of microwells, each microwell having a bottom which is the back-lapped silicon substrate and wherein each microwell of the array of microwells is further adapted to capture therein one or more biological cells.

13. The computerized method of claim 12, wherein the identification of a phenotype comprises at least one of: inference and prediction.

14. The computerized method of claim 12, wherein the collection of images is performed simultaneously on the fluorescence microscope and the biosensor.

15. The computerized method of claim 12, wherein the collection of images further comprises a timestamp for each image that is captured.

16. The method of claim 15, wherein multiplexing of images collected is within a predetermined time range between timestamps of images collected from the fluorescence microscope and images collected from the biosensor.

17. A computerized method for identification of a biological cell’s phenotype based on an image from a biosensor, the method comprises: capturing an image of a plurality of biological cells captured by a biosensor; identifying at least a phenotype of a cell of the plurality of biological cells using an artificial intelligence (Al) model; and, reporting the identified at least a phenotype; wherein the biosensor comprises: a complementary metal-oxide semiconductor (CMOS) image sensor (CIS), wherein the CIS comprises a plurality of light sensitive diodes disposed over a silicon substrate, and one or more metal layers for interconnect disposed therein above, and wherein the silicon substrate is back-lapped to a predetermined thickness; and, a microwell layer disposed over the back-lapped silicon substrate having etched therein an array of microwells, each microwell having a bottom which is the back-lapped silicon substrate and wherein each microwell of the array of microwells is further adapted to capture therein one or more biological cells; and, wherein the Al model was generated by training an Al engine by at least multiplexing images of biological cells captured by the biosensor and a fluorescence microscope.

18. The computerized method of claim 16, wherein the identification of a phenotype comprises at least one of: inference and prediction.

Images