Characterization of beads in cell samples

Abstract

An example method performed via a computing device for providing support to an imaging flow-cytometry instrument includes receiving flow-cytometry event data and images representing a plurality of events detected with the imaging flow-cytometry instrument and corresponding to a sample including cells and beads. The method also includes selecting a group of events by applying a set of gating operations to the plurality of events. The group of events includes bead events. The method also includes processing a subset of the images corresponding to the selected group of events with a machine learning model to determine a total number of beads in the subset of the images.

Description

Docket No.: TP389019WO1CHARACTERIZATION OF BEADS IN CELL SAMPLESTECHNICAL FIELD

[0001] Various examples relate generally to biopharmaceutical product release tests and, more specifically but not exclusively, to flow-cytometry-based quality control assays.BACKGROUND

[0002] Flow cytometry is a powerful and flexible technique that can be used to rapidly analyze multiple parameters of individual cells within heterogeneous cell populations. Flow cytometers are utilized in a variety of applications, such as immuno-phenotyping, ploidy analysis, cell counting, fluorescence-activated cell sorting (FACS), fluorescence expression analysis, and more. In some implementations, a flow cytometer performs the corresponding analysis of the sample by passing thousands of cells per second through a laser beam and capturing the fluorescence and scattered light that emerges from each cell. As the cells pass through, fluorescence data are collected and analyzed by flow-cytometry software to report cellular characteristics, such as, for example, size, complexity, phenotype, cell health (e.g., viability, proliferation, and apoptotic states), etc.SUMMARY

[0003] Disclosed herein are, among other things, various examples, aspects, features, and embodiments of image-based analytical methods directed at characterizing residual beads in biopharmaceutical products using an imaging flow cytometer. In one example, the imaging flow cytometer is used to capture brightfield images of cells passing therethrough and record corresponding event data. The captured images are processed for object and pixel attributes, which are further analyzed using a selected set of image-analysis parameters. A sequence of specific combinations of parameter options defines a gating strategy for separation of three main image classes: cell only, bead only, and cell-bead. The latter two classes are used to obtain an accurate residual bead count in the sample, e.g., to provide technical support to product release tests.

[0004] In some examples, a computer vision machine-learning model is used to process the captured images corresponding to a group of events selected via one or more gating operations. The model is trained to recognize in the images at least three classes of objects including healthy cells,Docket No.: TP389019WO1 unhealthy cells, and beads, respectively. The residual bead count in the sample is determined by automatically counting the objects recognized by the model as belonging to the third class of objects. In addition, predictions of the model can be postprocessed to generate a report characterizing the healthy and unhealthy cell populations in the sample based on the pertinent characteristics of the objects that are recognized by the model as belonging to the first and second classes of objects.

[0005] One example provides an apparatus comprising: an imaging flow-cytometry instrument; and a computing device configured to: receive flow-cytometry event data and images representing a plurality of events detected with the imaging flow-cytometry instrument and corresponding to a sample including cells and beads; select a first group of events by applying a set of event-data-based gating operations to the plurality of events; select a second group of events and a third group of events from the first group of events by applying a set of image-based gating operations to the first group of events, the second group including cell events, the third group including bead events, the second and third groups having no events in common; and present a subset of the images corresponding to the third group of events for counting a total number of beads in said subset of the images.

[0006] Another example provides an apparatus comprising: an imaging flow-cytometry instrument; and a computing device configured to: receive flow-cytometry event data and images representing a plurality of events detected with the imaging flow-cytometry instrument and corresponding to a sample including cells and beads; select a group of events by applying a set of gating operations to the plurality of events, the group of events including bead events; and process a subset of the images corresponding to the selected group of events with a machine learning model to determine a total number of beads in the subset of the images.

[0007] Yet another example provides a method performed via a computing device for providing support to an imaging flow-cytometry instrument, the method comprising: receiving flow-cytometry event data and images representing a plurality of events detected with the imaging flow-cytometry instrument and corresponding to a sample including cells and beads; selecting a first group of events by applying a set of event-data-based gating operations to the plurality of events; selecting a second group of events and a third group of events from the first group of events by applying a set of image-based gating operations to the first group of events, the second group including cell events, the third group including bead events, the second and third groups having no events in common; andDocket No.: TP389019WO1 presenting a subset of the images corresponding to the third group of events for counting a total number of beads in said subset of the images.

[0008] Yet another example provides a method performed via a computing device for providing support to an imaging flow-cytometry instrument, the method comprising: receiving flow-cytometry event data and images representing a plurality of events detected with the imaging flow-cytometry instrument and corresponding to a sample including cells and beads; selecting a group of events by applying a set of gating operations to the plurality of events, the group of events including bead events; and processing a subset of the images corresponding to the selected group of events with a machine learning model to determine a total number of beads in the subset of the images.

[0009] Yet another example provides a non-transitory computer-readable medium storing instructions that, when executed by a computing device, cause the computing device to perform operations comprising any one of the above methods.BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The foregoing aspects and many of the attendant advantages of the present disclosure will become more readily appreciated as the same become better understood by reference to the following detailed description, when taken in conjunction with the accompanying drawings.

[0011] FIG. 1 is a block diagram illustrating an imaging flow-cytometry system according to some examples.

[0012] FIG. 2 shows a set of fifteen images captured using the imaging flow-cytometry system of FIG. 1 according to some examples.

[0013] FIG. 3 is a flowchart illustrating an analytical method that can be implemented using the imaging flow-cytometry system of FIG. 1 according to some examples.

[0014] FIG. 4 graphically illustrates an event-data-based gating operation used in the analytical method of FIG. 3 according to one example.

[0015] FIG. 5 is a block diagram illustrating image-processing parameters that can be used to configure image-based gating operations used in the analytical method of FIG. 3 according to some examples.Docket No.: TP389019WO1

[0016] FIGS. 6A-6D are verification plots graphically illustrating the effectiveness of the imagebased gating operations used in the analytical method of FIG. 3 for separating cell and bead / cell- bead populations according to one example.

[0017] FIGS. 7A-7B graphically illustrate the determination of the limit of detection and lower limit of quantification for the analytical method of FIG. 3 according to one example.

[0018] FIG. 8 is a flowchart illustrating a sequence of image-based gating operations that can be used in the analytical method of FIG. 3 according to some examples.

[0019] FIG. 9 graphically illustrates an image-based gating operation used in the sequence of FIG. 8 according to one example.

[0020] FIG. 10 graphically illustrates another image-based gating operation used in the sequence of FIG. 8 according to one example.

[0021] FIG. 11 graphically illustrates yet another image-based gating operation used in the sequence of FIG. 8 according to one example.

[0022] FIG. 12 pictorially illustrates a statistics-display screen used in the sequence of FIG. 8 according to one example.

[0023] FIG. 13 is a block diagram illustrating a computing device used in or network-connected to the imaging flow-cytometry system of FIG. 1 according to some examples.

[0024] FIG. 14 is a flowchart illustrating an analytical method implemented using the imaging flow-cytometry system of FIG. 1 according to additional examples.

[0025] FIG. 15 shows several images captured using the imaging flow-cytometry system of FIG. 1 during a run of a sample including healthy cells, unhealthy cells, and beads according to some examples.

[0026] FIG. 16 is a bar chart illustrating the performance of a machine-learning (ML) model employed in the analytical method of FIG. 14 according to some examples.

[0027] FIG. 17 illustrates a report summary generated using the analytical method of FIG. 14 according to one example.Docket No.: TP389019WO1

[0028] FIG. 18 is a block diagram illustrating a workflow used to implement the analytical method of FIG. 14 according to some examples.DETAILED DESCRIPTION

[0029] Superparamagnetic particles, also sometimes referred to as magnetic beads, are versatile tools that can be used for efficient and effective isolation of biomolecules and / or cells. In a representative example, magnetic beads are small (e.g., < 10 pm in size) spheres made of a composite including a polymer, such as polystyrene, and an iron oxide, such as magnetite (FeaO^. The latter material gives the magnetic beads their superparamagnetic properties. Superparamagnetic beads are different from more-common ferromagnetic beads in that they exhibit magnetic behavior only in the presence of an external magnetic field. This property is dependent on the small size of the superparamagnetic particles in the beads and enables the beads to be separated in suspension, along with anything they are bound to. Since superparamagnetic particles do not attract each other outside of a magnetic field, concerns about unwanted clumping are greatly alleviated.

[0030] Many types of magnetic beads are commercially available on the market. For example, different surface coatings and chemistries give each type of beads its own binding properties. Such customization can beneficially be used, e.g., to enable magnetic separation, isolation, and purification of nucleic acids, proteins, and other biomolecules in a straightforward, effective, and scalable way. The ease-of-use makes the magnetic beads automation friendly and well suited for a range of applications, including sample preparation for next generation sequencing (NGS) and polymerase chain reactions, protein purification, molecular and immunodiagnostics, magnetic activated cell sorting (MACS), and many others.

[0031] Magnetic separation uses a magnetic field to separate small paramagnetic particles from a suspension. In molecular biology, magnetic beads provide a straightforward and reliable way to isolate and purify various types of biomolecules including, but not limited to, genomic DNA, plasmids, mitochondrial DNA, RNA, and proteins. In cell biology, magnetic beads provide a straightforward and reliable way to isolate and purify specific cells, based on the antigen of interest for positive selection or lack of antigens for negative selection. One advantage to using magnetic beads is that the target cells and / or biomolecules can be isolated directly from a crude sample with minimal amount of processing. Herein, the term “biomolecules” refers to chemical compoundsDocket No.: TP389019WO1 found in and / or produced by living organisms. Example types of biomolecules include, but are not limited to, carbohydrates, proteins, nucleic acids, and lipids.

[0032] In one example, sample purification includes the operations of binding, washing, and debeading. For the binding operation, magnetic beads are added to the sample to bind to the target biomolecules or cells. After the binding, an external magnetic field is applied, which attracts the magnetic beads to the corresponding outer edge of the container, thereby immobilizing the magnetic beads together with the bound biomolecules / cells. During the washing operation, the magnetic beads are kept immobilized while the remainder of the sample is washed away. During the debeading operation, an elution buffer is added to release the bound biomolecules / cells from the magnetic beads as a purified sample, ready for downstream processing, analysis, and quantification. However, in a typical implementation, the de-beading operation is less than 100% effective. As a result, a small fraction of beads may remain in the purified sample as a contaminant.

[0033] The Food and Drug Administration (FDA) imposes limits on the levels of contaminants in biological and biopharmaceutical products. For example, with respect to residual beads in cell therapy products, dosing limits may be around 100 beads per 3xl06cells. Accordingly, product release tests typically need to include an accurate residual bead count assay (RBCA) to demonstrate or confirm compliance with the pertinent regulations and similar quality controls (e g., imposed by government agencies or internally).

[0034] In one example, an RBCA is implemented as part of the Chimeric Antigen Receptor (CAR) T-cell workflow. T cells are important cells in the immune system and are responsible for fighting infections and cancer and, as such, are extensively used in many cancer and other disease studies and treatments. One of the steps in the CAR T-cell workflow is the isolation of T cells from the apheresis material. Magnetic beads, such as human CD3 / CD28 T-cell activation beads, may be used for T-cell isolation, and such beads simultaneously activate T cells and can be magnetically sorted.

[0035] For example, the human CD3 / CD28 T-cell activation beads allow for convenient in vitro activation of human T cells without the need for antigen-presenting cells. Human T cells are stimulated with anti-CD3 and anti-CD28 antibodies coated on magnetic beads. In one example, the coated magnetic beads are added directly to isolated T cells, Leukapheresis, or to PBMCs at a 1 : 1 bead-to-cell ratio for robust T-cell activation. In other examples, other suitable bead-to-cell ratiosDocket No.: TP389019WO1 can also be used. Short term activation, e.g., for a few hours to a few days, does not expand the cells, and proteomics expression / detection and nucleic acid assays can be performed. Typically, the cell / bead complexes are not used in flow cytometry in this phase. Long-term activation, e.g., for three days to several weeks, is used to obtain an expansion of the T cells. After de-beading, the presence of residual beads can be characterized using a suitable RBCA, e.g., as described in more detail below.

[0036] At least some of the above-indicated problems in the state of the art can beneficially be addressed using various examples disclosed herein. For example, some examples provide imagebased analytical methods directed at characterizing residual immunoselection beads using an imaging flow cytometer. In one example, an imaging flow cytometer is used to capture brightfield images of cells passing therethrough and record corresponding event data. The captured images are processed for object and pixel attributes, which are further analyzed using a selected set of imageanalysis parameters. A sequence of specific combinations of parameter options defines a gating strategy for separation of three main image classes: Cell Only, Bead Only, and Cell-Bead. The latter two classes can be used to obtain an accurate residual bead count in the sample, e.g., to provide technical support to product release tests.

[0037] FIG. 1 is a block diagram illustrating an imaging flow-cytometry (1FC) system 100 according to some examples. The IFC system 100 includes an excitation light source 110, a highspeed camera 120, a fluidics subsystem 130, a set of optical filters 140, a plurality of optical detectors 150, and an electronic subsystem 160. In one example, the IFC system 100 is implemented using the commercially available Invitrogen™ Attune™ CytPix Flow Cytometer. In other examples, other suitable imaging flow cytometers can also be used.

[0038] The fluidics subsystem 130 is configured to inject a sample 132 into a central core 134 of a funnel (or nozzle) 136. In one example, the sample 132 contains a suspension of cells and / or particles in a carrier fluid. The diameter of the corresponding sample injection hose is typically significantly larger than the size of the cells and / or particles in the sample 132. As a result, the cells and / or particles of the sample 132 are initially distributed substantially randomly in the corresponding volume of the central core 134.

[0039] Peripheral portions of the funnel 136 receive a flow of a sheath fluid 133 from a sheathfluid container via a sheath fluid injection hose. Due to the narrowing of the funnel 136, theDocket No.: TP389019WO1 velocity of the sheath fluid 133 increases along the downstream direction in the funnel 136. The flow of the sample 132 is focused into the center of the funnel 136 due to the Bernoulli effect associated with the velocity change such that the cells and / or particles of the sample 132 approach the exit aperture of the funnel 136 substantially in single file. Under optimal flow conditions (e.g., with laminar flow), there is substantially no mixing in of the sheath fluid 133 into the central portion of an output stream 138 that carries the sample 132.

[0040] The output stream 138 ejected from the funnel 136 passes through one or more laser beams 112 generated by the excitation light source 110. Each of the laser beams 112 has a respective wavelength, which can be in the spectral range from ultraviolet to infrared (depending on the specific embodiment and configuration of the IFC system 100 and on the fluorescent markers used with the sample 132). Light scattering and fluorescence emission generated in the output stream 138 in response to the one or more laser beams 112 are filtered by the set of optical filters 140 and detected using the plurality of optical detectors 150. For illustration purposes and without any implied limitations, five optical detectors 150 labeled FSC, SSC, FL1, FL2, and FL3 are shown in FIG. 1. In other embodiments, a different (from five) number of optical detectors 150 can similarly be used. Electrical output signals generated by the optical detectors 150 are amplified using an amplifier 162 and converted into digital form in an analog-to-digital converter (ADC) 164. Resulting digital signals 166 are then directed to a computing device 168 for processing.

[0041] In some examples, one of the optical detectors 150 is configured to detect light scattered in the forward direction, such as, for example, at an angle that is smaller than approximately 20° with respect to the excitation laser beam axis. An output of this optical detector is typically referred to as the forward scatter channel (FSC). The FSC signals can be used, for example, to estimate the size of the light scattering cells or particles in the stream 138. Another one of the optical detectors 150 is typically configured to detect light scattered at approximately 90° with respect to the excitation laser beam axis. The corresponding output is typically referred to as the side scatter channel (SSC). The SSC signals can be used to provide information about the relative complexity (for example, granularity and internal structure) of the cells and / or particles of the sample 132. Based on a combination of FSC and SSC data, the cells / particles can be differentiated, such as, for example, by cell types and sizes.

[0042] Additional one or more of the optical detectors 150 are configured for fluorescence measurements. In an example configuration, the fluorescent light is collected at approximately 90°Docket No.: TP389019WO1 with respect to the excitation laser beam axis and is directed to the corresponding one of the optical detectors 150 through a corresponding subset of beam splitters and optical filters of the set 140. The combination of optical filters in the set 140 is such that each of the corresponding optical detectors 150 receives light that is spectrally within a different respective spectral (wavelength) band.

[0043] When a cell or particle passes through the optical interrogation spot, a pertinent set of the optical detectors 150 will generate electrical pulses. These pulses in the corresponding set of the output signals manifest the passage of a cell or particle through the laser beam 112. As explained above, different ones of the optical detectors 150 will report on the scattered (FSC and SSC) light and on the fluorescence light. As the cell or particle enters the optical interrogation spot, the output of the corresponding optical detector will begin to rise, reaching a peak when the cell or particle is located approximately in the center of the excitation laser beam 112. At this point, the cell or particle is brightly and fully illuminated (the laser beam 112 typically has the highest intensity in the center portion thereof) and will produce a strongest optical signal. As the cell or particle moves away from the center of the laser beam 112, the optical signal will drop back to the baseline. This generation of an optical pulse defines an “event.” The computing device 168 operates to put a time stamp on each event and record one or more pertinent characteristics of the corresponding pulse, such as, for example, the amplitude, width, an integrated area of the pulse, or a combination thereof.

[0044] A measurement from each optical detector 150 provides a respective event parameter. Each event parameter can be displayed according to its amplitude, area, and width values on histograms and dot plots using appropriate data-processing software. Tens or even hundreds of thousands of cells per second can be examined in this manner based on their event parameters. A combination of event parameters provides an event “signature,” which can be used for performing various sorting and / or gating operations, e.g., as explained in more detail below.

[0045] The high-speed camera 120 operates to capture a respective brightfield image 122 for each flow-cytometry event detected as explained above. A camera readout signal 124 is used to transfer the digital files representing the captured images to the computing device 168. Several examples of the captured images 122 are shown in and described in more detail below in reference to FIG. 2. In one example, the camera 120 is capable of capturing up to 6,000 images / s (depending on the image size and event rate). In some embodiments, the objective of the camera 120 is configured to provide 20x magnification. The pixelated detector of the camera 120 is configured to provide a spatial resolution of approximately 0.3 pm / pixel. In other embodiments, other suitableDocket No.: TP389019WO1 cameras characterized by other values of the operating parameters and characteristics may also be used.

[0046] FIG. 2 shows a set 200 of fifteen images 122 captured using the IFC system 100 during one example run of the sample 132 including cells and beads. More specifically, the set 200 corresponds to the sample 132 including T cells spiked with magnetic beads with the cell-to-bead ratio of 10: 1. The set 200 includes five subsets of images, which are labeled 210-250, respectively, with each of the subsets consisting of three respective images 122. The images of the subset 210 are images of single cells. The images of the subset 220 are images of multiple cells, e.g., in the form of cell clumps. Each of the images in the subset 230 includes at least one cell and at least one magnetic bead. Note that the top image in the subset 230 has a single magnetic bead and multiple cells. The middle image in the subset 230 has a single cell and multiple magnetic beads. The bottom image in the subset 230 has a single cell and a single magnetic bead. The images of the subset 240 are images of magnetic beads. Note that the top image in the subset 240 has multiple magnetic beads. The images of the subset 250 are images of cellular debris, bubbles, and / or some samplepreparation-process scrap.

[0047] FIG. 3 is a flowchart illustrating an analytical method 300 implemented using the IFC system 100 according to some examples. The method 300 includes the computing device 168 running software configured to analyze the flow-cytometry event data and the corresponding set of images 122 obtained by passing the sample 132 through the IFC system 100 as indicated above.The method 300 is described below with continued reference to FIGS. 1-3 and with further reference to FIGS. 4-7.

[0048] A block 302 of the method 300 includes the computing device 168 receiving flowcytometry event data and the corresponding set of images 122 of the sample 132 processed with the IFC system 100. The flow-cytometry event data are received via the digital signals 166. The corresponding set of images 122 is received via the camera readout signal 124. The received information is stored in a memory of the computing device 168 accessible thereby during postacquisition processing.

[0049] A block 304 of the method 300 includes the computing device 168 applying a set of one or more event-data-based gating operations to the information received in the block 302. In various examples, the one or more event-data-based gating operations of the block 304 are configured toDocket No.: TP389019WO1 perform a general clean-up of the data, such as excluding dead-cell events or events involving multiple cells, as well as isolating the target cell population, e.g., using the characteristic size, granularity, expression of various cell markers, etc.

[0050] Herein, the term “gating” refers to a process of selecting a subset of events from a larger set of events detected during a flow cytometry experiment for further analysis and / or data presentation. In some examples, a sequence of two or more different gating operations can be applied to narrow the event selection in a stepwise manner. The input to such sequence may typically include all events detected in the corresponding flow-cytometry experiment. The events inside an applied gate are included in further analysis and / or postprocessing, whereas the events outside the applied gate are excluded.

[0051] FIG. 4 graphically illustrates a configuration of an event-data-based gating operation 400 used in the block 304 of the method 300 according to one example. The gating operation 400 is configured to limit the number of eligible events based on the FSC and SSC signals. In general, the beads and larger, more complex cells and cell-bead complexes will be higher in both parameters. In the example shown, an inclusion gate 402 of the gating operation 400 is relatively broad and causes only a relatively small fraction of events falling into a triangular area 404 located near the origin of the (FSC, SSC) coordinate plane to be excluded. With high likelihood, the excluded events correspond to cellular debris and other sample-preparation-process scrap, e.g., illustrated in the subset 250 of the image set 200 (FIG. 2).

[0052] Referring back to FIG. 3, a block 306 of the method 300 includes the computing device 168 applying a set of one or more image-based gating operations to the set of events included in the gate(s) used in the block 304. In a representative example, the image-based gating operations are directed at identifying three respective groups of images 122 including: (i) cell only images, e.g., illustrated in the subsets 210 and 220 of the set 200; (ii) bead only images, e.g., illustrated in the subset 240 of the set 200; and (iii) cell-bead images, e.g., illustrated in the subset 230 of the set 200. An example sequence 800 of image-based gating operations that can be used in the block 306 is described in more detail below in reference to FIG. 8.

[0053] FIG. 5 is a block diagram illustrating a set 500 of image-processing parameters that can be used to configure the image-based gating operations used in the block 306 of the method 300 according to some examples. The set 500 includes a total of twenty-six image-processingDocket No.: TP389019WO1 parameters sorted into five categories, which are labeled 501-505, respectively. The category 501 includes system feature parameters. The category 502 includes object feature parameters. The category 503 includes pixel feature parameters. The category 504 includes shape feature parameters. The category 505 includes intensity feature parameters. In other examples, other suitable sets of image-processing parameters can also be used.

[0054] A configuration of an example image-based gating operation used in the block 306 of the method 300 is described in more detail below in reference to FIG. 9. In one example, the imagebased gating operation illustrated in FIG. 9 is performed right after the event-data-based gating operation 400 illustrated in FIG. 4.

[0055] FIGS. 6A-6D are verification plots graphically illustrating the effectiveness of the set of image-based gating operations used in the block 306 of the method 300 for separating cell and bead / cell-bead populations according to one example. The corresponding sample 132 includes T cells spiked with magnetic beads with the cell-to-bead ratio of 10: 1.

[0056] FIG. 6A graphically shows a scatter plot 610 mapping the gated events onto the coordinate plane defined by the Normalized Intensity SD parameter and the Minimum Intensity parameter, both of which belong to the intensity features category 505 of the parameter set 500 shown in FIG. 5. In the scatter plot 610, the above-mentioned three populations fall into separate rectangular boxes, which are labeled 602, 604, and 606, respectively. More specifically, the cell only events fall into the box 602; the bead events fall into the box 604; and the cell-bead events fall into the box 606.

[0057] FIGS. 6B-6D graphically show histograms 612, 614, and 616 representing the statistics of the Average Normalized Intensity parameter corresponding to the boxes 602, 604, and 606, respectively. In the brightfield images, the cells are typically represented by pixels having relatively high intensity values (e.g., see the subsets 210, 220 in FIG. 2). Accordingly, the histogram 612 (FIG. 6B) has a peak located at about 95% mark of the Average Normalized Intensity axis. In contrast, the beads are typically represented by pixels having relatively low intensity values (e.g., see the subsets 230, 240 in FIG. 2). Accordingly, the histogram 616 (FIG. 6D) has a peak located at about 20% mark of the Average Normalized Intensity axis. The histogram 614 representing the cell-bead images has a peak located at about 75% mark of the Average Normalized Intensity axis, which is an intermediate location between the peaks of the histograms 612 and 616.Docket No.: TP389019WO1

[0058] Referring back to FIG. 3, a block 308 of the method 300 includes the computing device 168 presenting the groups of the images 122 corresponding to the different groups of events identified in the block 306 for residual bead counting. Cumulatively, the gating operations of the blocks 304 and 306 significantly reduce the number of images 122 that need to be processed by the operator or downstream image processing software to obtain the residual bead count characterizing the sample 132. In a representative example, the gating operations of the blocks 304 and 306 can reduce the number of images to be analyzed from about 300,000 to about 1,000, which represents a reduction of more than two orders of magnitude.

[0059] In different embodiments of the block 308, different counting methods can be applied to the presented groups of images 122. For example, a first counting method that can be used in the block 308 includes the computing device 168 sequentially displaying on a display device the bead and cell-bead images exemplified by the subsets 230 and 240 (see FIG. 2) for manual bead counting by the user. A second counting method that can be used in the block 308 includes the image processing software running on the computing device 168 performing automated bead counting in the presented groups of the images 122. A third counting method that can be used in the block 308 includes feeding the presented groups of the images 122 to a computer vision model adapted for bead counting. In some examples, the computer vision model is a machine-learning (ML) computer vision model trained for the task of bead counting. In one nonlimiting example, such ML computer vision model is based on a convolutional neural network (CNN) architecture. To train the model, a subset of collected images is selected, and each of the images is annotated with ground-truth labels for healthy cells, unhealthy cells, and beads. In other embodiments, other suitable bead counting methods, such as a hybrid manual and algorithm-based counting method, can also be used.

[0060] In some examples of the above-mentioned second counting method, the automated bead counting algorithm is not configurable to count more than one bead event per image. As a result, when there are two or more beads in an image, the algorithm will only count these multiple beads as one bead, thereby undercounting the actual number of beads. To address this issue, the user may manually count the beads in the multiple-bead gate, thereby causing a hybrid (algorithm + manual) method to be applied for bead counting.

[0061] In some examples, operations of the block 308 also include the computing device 168 determining the bead-to-cell ratio characterizing the sample 132. This determination can be made, for example, using the bead count obtained with a selected one of the above-described countingDocket No.: TP389019WO1 methods and further using the total number of cells in the gated events. Operations of the block 308 may further include the computing device 168 reporting the determined bead-to-cell ratio to the user, e.g., by displaying the results on a display screen and / or generating and saving an electronic report file in a memory device accessible through the computing device 168.

[0062] FIGS. 7A-7B graphically illustrate the determination of the limit of detection (LOD) and lower limit of quantification (LLOQ) for the method 300. The LOD represents the lowest concentration level that can be determined to be statistically different from a blank at a 99% confidence level. In other words, it is the lowest quantity of a substance that can be distinguished from the absence of that substance (a blank value) within a stated confidence limit, generally 1%. The LOD value is matrix-, method-, and analyte-specific. In the example shown in FIG. 7A, the LOD range for magnetic bead detection on the IFC system 100 is >12 beads per milliliter (beads / mL).

[0063] The LLOQ is the lowest standard curve point that can still be used for quantification, meeting both linearity and accuracy criteria. The LLOQ indicates the value above which quantitative results may be obtained with a specified degree of confidence. In other words, the LLOQ represents the lowest concentration of an analyte that can be accurately measured. The limits of quantitation are also matrix-, method-, and analyte-specific. In the example shown in FIG. 7B, the LLOQ for the residual bead assay is > 20 beads per 1,000,000 cells, which meets the sensitivities preferred for cell therapy product release testing. Linearity and accuracy were achieved with R2 value of 0.9800 or better and recovery to a theoretical concentration within 70-130%.

[0064] FIG. 8 is a flowchart illustrating a sequence 800 of image-based gating operations that can be used in the block 306 of the method 300 according to some examples. The sequence 800 is described below with continued reference to FIGS. 3, 5, and 8 and with further reference to FIGS. 9- 12.

[0065] A block 802 of the sequence 800 includes the computing device 168 generating a scatter plot representing the gated data received from the block 304 of the method 300 on the coordinate plane defined by the Average Normalized Intensity parameter and the Circularity (%) parameter. The Average Normalized Intensity parameter belongs to the intensity features category 505 of the parameter set 500 (FIG. 5). The Circularity (%) parameter belongs to the shape features categoryDocket No.: TP389019WO1504 of the parameter set 500 (FIG. 5). Operations of the block 802 further include the computing device 168 gating the generated scatter plot, e.g., as illustrated in FIG. 9.

[0066] The Average Normalized Intensity (AverageNormlntensity) parameter is a measure of the object surface complexity. In terms of this measure, the cells will be more complex (e.g., -100 threshold), and the beads will be less complex (e.g., <40 threshold). The Circularity (%) (CircularityPercent) parameter is a measure of how round an object is. In terms of this measure, the score of 100 represents a perfectly round object. In some examples, a Circularity (%) parameter value smaller than 80 may indicate dead or dying cells.

[0067] FIG. 9 graphically illustrates a configuration of an image-based gating operation 900 used in the block 802 of the sequence 800 according to one example. The gating operation 900 is configured to limit the number of eligible events based on the values of the Average Normalized Intensity and Circularity (%) parameters. In the example shown, an inclusion gate 902 of the gating operation 900 is represented by a rectangular box encompassing the range of values of the Average Normalized Intensity parameter from approximately 5 to 100. The inclusion gate 902 causes the events falling into a rectangular area 904 located next to the Circularity (%) coordinate axis to be excluded.

[0068] In some examples, the events falling into the rectangular area 904 tend to hyperinflate the total number of pertinent events and, as such, are gated out. In the example shown, the area 904 accounts for approximately 38% of the total event count. However, inspection of the corresponding images 122 reveals that only thirteen images of the 10,000 total captured images are in this region and, hence, the percentage should be 0.013% and not 38% of the total events. Additionally, visual inspection of the thirteen representative images reveals that these images are either images of bubbles or background noise and do not represent any actual cell or bead events.

[0069] Referring back to FIG. 8, a block 804 of the sequence 800 includes the computing device 168 generating a scatter plot representing the gated events received from the block 802 on the coordinate plane defined by the Minimum Intensity (Minlntensity) parameter and the Normalized Intensity SD (StandardDeviationNormlntensity) parameter. Both of these parameters belong to the intensity features category 505 of the parameter set 500 shown in FIG. 5. The generated scatter plot is then used to separate cell-bead and bead events from cell-only events, e.g., as further illustrated in FIG. 10.Docket No.: TP389019WO1

[0070] FIG. 10 graphically illustrates a configuration of an image-based gating operation 1000 used in the block 804 of the sequence 800 according to one example. As previously mentioned, in the brightfield images 122, the beads will have a relatively low average intensity. Consequently, the use of the Minimum Intensity parameter enables separation of bead and cell events. In the example shown, a rectangular gate 1002 is used to identify cell-only events, and a rectangular gate 1004 is used to identify cell-bead and bead events. The Minlntensity range for the gate 1004 is from 1 to 31. The Minlntensity range for the gate 1002 is >32. In general, the position of a border 1003 between the gates 1002 and 1004 is matrix-, method-, and analyte-specific and, as such, is set based on calibration, e.g., involving inspection of the corresponding subsets of the images 122.

[0071] Referring back to FIG. 8, a block 806 of the sequence 800 includes the computing device 168 generating a histogram displaying the events received from the gate 1004 based on the values of the Major Axis (Maj orDiameterMi crons) parameter, which belongs to the shape features category 504 of the parameter set 500 (FIG. 5) and represents a major size of pertinent objects in the images. This histogram enables separation of single bead events and multiple bead events for proper counting of the beads in the images 122 having multiple beads. As previously mentioned, some automated counting algorithms employed in the block 308 of the method 300 will erroneously count these multiple beads as one bead. Accordingly, the gating operation of the block 806 enables the output of such counting algorithms to be appropriately adjusted to substantially eliminate the corresponding undercounting.

[0072] FIG. 11 graphically illustrates a configuration of an image-based gating operation 1100 used in the block 806 of the sequence 800 according to one example. The image-based gating operation 1100 uses the above-mentioned histogram to differentiate between single bead events and multiple bead events. A first gate 1102 of the image-based gating operation 1100 is used to identify single bead events. A second gate 1104 of the image-based gating operation 1100 is used to identify multiple bead events. In general, the position of a boundary 1103 between the gates 1102 and 1104 is analyte-specific and, as such, is set based on calibration, e.g., involving inspection of the corresponding subsets of the images 122. In the example shown, the boundary 1103 is placed at the 10 pm mark on the MajorDiameterMicrons coordinate axis.

[0073] Referring back to FIG. 8, a block 808 of the sequence 800 includes the computing device 168 displaying the statistics corresponding to the gating operations of the blocks 802, 804, and 806.Docket No.: TP389019WO1In some examples, the displayed statistics can be used by the user to verify that all events for both cell and bead counts are captured. For example, the user can inspect bead counts from both the scatter plot of FIG. 10 and the histogram of FIG. 11 to look for inconsistencies that may manifest errors in the algorithmic counting of beads in multiple bead events and / or bead-cell events. When any inconsistencies are detected by the user, a responsive action directed at manually correcting the bead counts based on visual inspection of the corresponding images 122 can be taken.

[0074] FIG. 12 illustrates a statistics-display screen 1200 produced in the block 808 of the sequence 800 according to one example. The statistics-display screen 1200 includes a first panel 1202 showing the statistics of all cell events and all bead events. The statistics-display screen 1200 includes a second panel 1204 providing further details for the bead events by separately showing the statistics of single bead events and multiple bead or bead-cell events. In the example shown, a responsive action by the user may include visual inspection of all twenty images representing the bead events in the first panel 1202.

[0075] FIG. 13 is a block diagram illustrating a computing device 1300 one or more instances of which can be used in or in conjunction with the IFC system 100 according to some examples. In some examples, the computing device 1300 can be used to implement the computing device 168 (FIG. 1). In some examples, the computing device 1300 is programmed to implement at least some parts of the analytical method 800 (FIG. 8). In some examples, the computing device 1300 is further programmed to run an ML model, e.g., as described in more detail below in reference to FIGS. 14-18.

[0076] The computing device 1300 of FIG. 13 is illustrated as having a number of components, but any one or more of these components may be omitted or duplicated, as suitable for the application and setting. In some embodiments, some or all of the components included in the computing device 1300 may be attached to one or more motherboards and enclosed in a housing. In some embodiments, some of those components may be fabricated onto a single system-on-a-chip (SoC) (e.g., the SoC may include one or more electronic processing devices 1302 and one or more storage devices 1304). Additionally, in various embodiments, the computing device 1300 may not include one or more of the components illustrated in FIG. 13, but may include interface circuitry for coupling to the one or more components using any suitable interface (e.g., a Universal Serial Bus (USB) interface, a High-Definition Multimedia Interface (HDMI) interface, a Controller Area Network (CAN) interface, a Serial Peripheral Interface (SPI) interface, an Ethernet interface, aDocket No.: TP389019WO1 wireless interface, or any other appropriate interface). For example, the computing device 1300 may not include a display device 1310, but may include display device interface circuitry (e.g., a connector and driver circuitry) to which an external display device 1310 may be coupled.

[0077] The computing device 1300 includes a processing device 1302 (e.g., one or more processing devices). As used herein, the terms “electronic processor device” and “processing device” interchangeably refer to any device or portion of a device that processes electronic data from registers and / or memory to transform that electronic data into other electronic data that may be stored in registers and / or memory. In various embodiments, the processing device 1302 may include one or more digital signal processors (DSPs), application-specific integrated circuits (ASICs), central processing units (CPUs), graphics processing units (GPUs), server processors, or any other suitable processing devices.

[0078] The computing device 1300 also includes a storage device 1304 (e.g., one or more storage devices). In various embodiments, the storage device 1304 may include one or more memory devices, such as random-access memory (RAM) devices (e.g., static RAM (SRAM) devices, magnetic RAM (MRAM) devices, dynamic RAM (DRAM) devices, resistive RAM (RRAM) devices, or conductive-bridging RAM (CBRAM) devices), hard drive-based memory devices, solid-state memory devices, networked drives, cloud drives, or any combination of memory devices. In some embodiments, the storage device 1304 may include memory that shares a die with the processing device 1302. In such an embodiment, the memory may be used as cache memory and include embedded dynamic random-access memory (eDRAM) or spin transfer torque magnetic random-access memory (STT-MRAM), for example. In some embodiments, the storage device 1304 may include non-transitory computer readable media having instructions thereon that, when executed by one or more processing devices (e.g., the processing device 1302), cause the computing device 1300 to perform any appropriate ones of the methods disclosed herein below or portions of such methods.

[0079] The computing device 1300 further includes an interface device 1306 (e.g., one or more interface devices 1306). In various embodiments, the interface device 1306 may include one or more communication chips, connectors, and / or other hardware and software to govern communications between the computing device 1300 and other computing devices. For example, the interface device 1306 may include circuitry for managing wireless communications for the transfer of data to and from the computing device 1300. The term “wireless” and its derivativesDocket No.: TP389019WO1 may be used to describe circuits, devices, systems, methods, techniques, communications channels, etc., that may communicate data via modulated electromagnetic radiation through a nonsolid medium. The term does not imply that the associated devices do not contain any wires, although in some embodiments they might not. Circuitry included in the interface device 1306 for managing wireless communications may implement any of a number of wireless standards or protocols, including but not limited to Institute for Electrical and Electronic Engineers (IEEE) standards including Wi-Fi (IEEE 802.11 family), IEEE 802.16 standards, Long-Term Evolution (LTE) project along with any amendments, updates, and / or revisions (e.g., advanced LTE project, ultramobile broadband (UMB) project (also referred to as “3GPP2”), etc.). In some embodiments, circuitry included in the interface device 1306 for managing wireless communications may operate in accordance with a Global System for Mobile Communication (GSM), General Packet Radio Service (GPRS), Universal Mobile Telecommunications System (UMTS), High Speed Packet Access (HSPA), Evolved HSPA (E-HSPA), or LTE network. In some embodiments, circuitry included in the interface device 1306 for managing wireless communications may operate in accordance with Enhanced Data for GSM Evolution (EDGE), GSM EDGE Radio Access Network (GERAN), Universal Terrestrial Radio Access Network (UTRAN), or Evolved UTRAN (E-UTRAN). In some embodiments, circuitry included in the interface device 1306 for managing wireless communications may operate in accordance with Code Division Multiple Access (CDMA), Time Division Multiple Access (TDMA), Digital Enhanced Cordless Telecommunications (DECT), Evolution-Data Optimized (EV-DO), and derivatives thereof, as well as any other wireless protocols that are designated as 3G, 4G, 5G, and beyond. In some embodiments, the interface device 1306 may include one or more antennas (e.g., one or more antenna arrays) configured to receive and / or transmit wireless signals.

[0080] In some embodiments, the interface device 1306 may include circuitry for managing wired communications, such as electrical, optical, or any other suitable communication protocols. For example, the interface device 1306 may include circuitry to support communications in accordance with Ethernet technologies. In some embodiments, the interface device 1306 may support both wireless and wired communication, and / or may support multiple wired communication protocols and / or multiple wireless communication protocols. For example, a first set of circuitry of the interface device 1306 may be dedicated to shorter-range wireless communications such as Wi-Fi or Bluetooth, and a second set of circuitry of the interface device 1306 may be dedicated to longer- range wireless communications such as global positioning system (GPS), EDGE, GPRS, CDMA,Docket No.: TP389019WO1WiMAX, LTE, EV-DO, or others. In some other embodiments, a first set of circuitry of the interface device 1306 may be dedicated to wireless communications, and a second set of circuitry of the interface device 1306 may be dedicated to wired communications.

[0081] The computing device 1300 also includes battery / power circuitry 1308. In various embodiments, the battery / power circuitry 1308 may include one or more energy storage devices (e.g., batteries or capacitors) and / or circuitry for coupling components of the computing device 1300 to an energy source separate from the computing device 1300 (e.g., to AC line power).

[0082] The computing device 1300 also includes a display device 1310 (e.g., one or multiple individual display devices). In various embodiments, the display device 1310 may include any visual indicators, such as a heads-up display, a computer monitor, a projector, a touchscreen display, a liquid crystal display (LCD), a light-emitting diode display, or a flat panel display.

[0083] The computing device 1300 also includes additional input / output (I / O) devices 1312. In various embodiments, the I / O devices 1312 may include one or more data / signal transfer interfaces, audio I / O devices (e.g., microphones or microphone arrays, speakers, headsets, earbuds, alarms, etc.), audio codecs, video codecs, printers, sensors (e.g., thermocouples or other temperature sensors, humidity sensors, pressure sensors, vibration sensors, etc.), image capture devices (e.g., one or more cameras), human interface devices (e.g., keyboards, cursor control devices, such as a mouse, a stylus, a trackball, or a touchpad), etc.

[0084] Depending on the specific embodiment, various components of the interface devices 1306 and / or I / O devices 1312 can be configured to output suitable control signals, receive suitable control / telemetry signals, and receive and transmit data streams. In some examples, the interface devices 1306 and / or I / O devices 1312 include one or more analog-to-digital converters (ADCs) for transforming received analog signals into a digital form suitable for operations performed by the processing device 1302 and / or the storage device 1304. In some additional examples, the interface devices 1306 and / or I / O devices 1312 include one or more digital-to-analog converters (DACs) for transforming digital signals provided by the processing device 1302 and / or the storage device 1304 into an analog form suitable for being transmitted through a communication channel.

[0085] FIG. 14 is a flowchart illustrating an analytical method 1400 implemented using the IFC system 100 according to additional examples. The method 1400 can be viewed as a modifiedDocket No.: TP389019WO1 version of the above-described method 300. One modification implemented in the method 1400 includes replacing the processing block 308 by processing blocks 1402 and 1404. Another optional modification implemented in the method 1400 may include removing the processing block 306. In the embodiments of the method 1400 in which the block 306 is present, the number of gating operations performed in the block 306 of the method 1400 may be smaller than the number of gating operations performed in the block 306 of the method 300. For the description of operations performed in the blocks 302, 304, and 306, the reader is referred to the description of the method 300 provided above in reference to FIGS. 3-12.

[0086] Operations of the block 1402 include feeding the images 122 corresponding to the selected gated groups of events to a previously trained machine-learning (ML) model. In some examples, the ML model operates to identify in the images 122 the following classes of objects: (A) healthy cells, (B) unhealthy cells, and (C) beads. In some examples, the ML model can identify: (i) healthy activated cells, healthy unactivated cells, and clumps of healthy cells for class (A); (ii) unhealthy cells, clumps of unhealthy cells, and cell debris for class (B); and (iii) single beads, beads attached to cells, and clumps of beads for class (C). In some additional examples, the ML model may be configured to identify in the images 122 only the following two classes of objects: (I) cells, including both healthy and unhealthy cells, and (II) beads.

[0087] FIG. 15 shows several images 122 captured using the IFC system 100 during a run of the sample 132 including healthy cells, unhealthy cells, and beads according to some examples. More specifically, an image 1502 captures three healthy cells each of which is classified by the ML model as a class (A) object. An image 1504 captures a clump of unhealthy cells that are classified by the ML model as class (B) objects. An image 1506 captures a bead attached to a healthy cell, which are classified by the ML model as class (C) and class (A) objects, respectively. An image 1508 captures a pair of beads attached to an unhealthy cell, which are classified by the ML model as class (C) and class (B) objects, respectively.

[0088] FIG. 16 is a bar chart 1600 illustrating performance of the ML model employed in the block 1402 of the method 1400 according to some examples. More specifically, the bar chart 1600 illustrates the precision and recall of the ML model with respect to the classes (A), (B), and (C). Herein, in model evaluation, “precision” refers to the proportion of positive predictions that are actually correct, whereas “recall” measures the proportion of actual positive cases that the model correctly identifies. In some examples, the model precision is calculated as: (True Positives) / (TrueDocket No.: TP389019WO1Positives + False Positives); and the model recall is calculated as: (True Positives) / (True Positives + False Negatives). The model evaluation results presented in FIG. 16 indicate high values (close to 100%) of both the model precision and the model recall for each of the three classes. The indicated high precision means that the ML model beneficially makes very few false positive predictions, ensuring that when the ML model predicts something as positive, the prediction is likely to be correct. The indicated high recall indicates that the ML model is good at identifying most of the actual positive cases while minimizing false negatives.

[0089] Referring back to FIG. 14, the block 1404 of the method 1400 includes running one or more postprocessing scripts configured to analyze the ML model’s predictions, tabulate results, and summarize the image data. Operations of the block 1404 further include generating a report containing the results generated with the post-processing scripts and typically a shorter report summary presenting the selected results in a user-friendly form. In various examples, the generated report / summary includes an estimate of the bead-to-cell ratio in the sample and may additionally incorporate information on certain cell features and cell-bead dynamics (e.g., cell health, bead orientation with respect to cells, cell size for activation, cell clustering, etc.) that can provide useful biological information to the user that goes beyond the residual bead detection. Such information can beneficially be used by the user, e.g., to assess bioprocessing benchmarks and adjust the corresponding bioprocessing workflow (when needed).

[0090] FIG. 17 illustrates a report summary 1700 generated in the block 1404 of the method 1400 according to one example. In the example shown, the report summary 1700 includes a sample statistics section 1710 and a graph 1720 detailing the healthy cell-size characteristics of the sample. The graph 1720 includes a bar plot 1722 constructed using the measured diameters of the class (A) objects identified by the ML model in the block 1402 of the method 1400. The graph 1720 also includes a kernel density estimation (KDE) plot 1724 corresponding to the bar plot 1722. In various examples, the format of the report summary generated in the block 1404 may be selectable based on the specific application of the method 1400. As such, various alternative examples of the report summary 1700 may include additional or alternative sections, additional or alternative graphs, and other specific information deemed appropriate by the user.

[0091] FIG. 18 is a block diagram illustrating a workflow 1800 used to implement the method 1400 according to some examples. In the example shown, operations of the block 306 in the method 1400 include three different image-gating operations, which are labeled in FIG. 18 using theDocket No.: TP389019WO1 reference numerals 1808, 1810, and 1816, respectively. The block 1402 of the method 1400 is configured to use two different ML models, which are labeled in FIG. 18 using the reference numerals 1820 and 1830, respectively. The block 1404 of the method 1400 is configured to run two different postprocessing scripts, which are labeled in FIG. 18 using the reference numerals 1822 and 1832, respectively. The postprocessing script 1822 is compatible with the ML model 1820 and generates a report 1824 that characterizes the population that has been gated-in by the image-gating operation 1810. Similarly, the postprocessing script 1832 is compatible with the ML model 1830 and generates a report 1834 that characterizes the population that has been gated-in by the imagegating operation 1816. In some examples, the workflow 1800 can be controlled to selectively activate only one of the ML models 1820, 1830, thereby causing the block 1404 to generate only the corresponding one of the reports 1824, 1834.

[0092] In the example shown, the gating operation 1808 is configured to select a subpopulation 1809 from a population 1807 previously selected by the gating operations of the block 304. The gating operation 1810 is configured to select subpopulations 1812, 1814 from the subpopulation 1809. The images 122 representing the subpopulation 1812 are then fed into the ML model 1820 for object classification and analysis. The gating operation 1816 is configured to select a subpopulation 1818 from the subpopulation 1814. The images 122 representing the subpopulation 1818 are then fed into the ML model 1830 for object classification and analysis.

[0093] In some examples, the subpopulation 1809 may be selected via the gating operation 1808 as being of specific interest for further analysis. The gating operation 1810 may be configured to exclude the events corresponding to bubbles and cell debris. The gating operation 1816 may be configured to exclude the cell-only events that fall into the box 602 and to gate in the bead and cellbead events that fall into the boxes 604 and 606 (also see FIG. 6A).

[0094] In some examples, depending on the throughput of the assay and the computational power available for the postprocessing in the block 1404, either the configuration employing the ML model 1820 or the configuration employing the ML model 1830 can be selected. The first (ML model 1820) selection allows for the analytic tool to be applied earlier in the gating scheme. One benefit of the first selection is that this particular configuration allows for a higher number of images from the sample to be analyzed with the ML model, which typically leads to a more robust report on cell features in addition to the residual bead detection data. The first option also supports a higher degree of automation in the data analysis, which may save time for the user by reducing the need forDocket No.: TP389019WO1 sifting through the images manually on the instrument before feeding them into the ML algorithm. The tradeoff of this option is that more images need to be processed, which may increase the system cost due to the need for a more powerful computer to analyze the data. The second (ML model 1830) selection can be used at the end of a multistep manual gating process, which drives down the number of images that need to be analyzed and tends to provide faster turnaround times for residual bead detection. The tradeoff of the second option is that also it tends to limit the amount of additional cellular feature data that the algorithm can provide to the user.

[0095] In some examples, the ML models 1820, 1830 are computer vision (CV) models that are based on a convolutional neural network (CNN) architecture. To train the model, a subset of collected images is selected, and each of the images is annotated with ground-truth labels for healthy cells, unhealthy cells, and beads. The resulting dataset is then split into training and validation sets to facilitate the model development and hyperparameter tuning. Additionally, a separate test set, consisting of images not included in the training or validation sets, is used to evaluate the model’s performance. Model iterations are trained by tuning the hyperparameters, including the model architecture size, data transformation, and data augmentation. Additional images can be continuously added to the dataset to iteratively improve the model. Once developed, the ML models 1820, 1830 can be deployed to a suitable computing infrastructure, e.g., accessible through the computing device 1300.

[0096] In some examples, each of the ML models 1820, 1830 is constructed based on the RepPoints model architecture described in Ze Yang, Shaohui Liu, Han Hu, et al., “RepPoints: Point Set Representation for Object Detection,” arXiv: 1904.11490v2, 2019, which is incorporated herein by reference in its entirety. Briefly, the RepPoints model architecture comprises a backbone network, a Feature Pyramid Network (FPN), and a RepPoints detector. The RepPoints detector has two branches, for classification and regression, respectively. The classification branch is responsible for predicting the class of each detected object. The regression branch (also referred to as the localization branch) is responsible for refining the positions of the RepPoints and predicting the bounding box coordinates for the detected objects.

[0097] In some examples, the model can be trained using focal loss for classification and smooth LI loss for regression. During the training, the following attributes are controllable: (i) hyperparameters, (ii) transforms, and (iii) augmentations. Example hyperparameters include the number of training epochs and the number of trainable parameters in the model. For example, theDocket No.: TP389019WO1 model platform offers two options, RepPoints-20M and RepPoints-37M. Additional adjustments may be performed by selecting the backbone size (e.g., ResNet-50 versus ResNet-101) and / or adjusting the layers in the FPN. Example transforms include the resize, resize with padding, and scale transforms. The resize transform is used to resize images to a consistent shape, larger or smaller. The resize with padding transform scales the images to the height and width the user enters, while maintaining the original aspect ratio. The crop transform applies the same crop to all images. Example augmentations optionally assign a probability that an augmentation would occur on the training image and may include horizontal flip, vertical flip, random rotate, random brightness, random contrast, hue saturation value, blur, motion blur, Gaussian blur, and random augment.

[0098] Additional embodiments of the ML models 1820, 1830 may benefit from the ML model architectures and / or features described in the following publications: (1) Joseph Redmon, Santosh Divvala, Ross Girshick, and Ali Farhadi, “You Only Look Once: Unified, Real-Time Object Detection,” arXiv: 1506.02640v5, 2016; (2) Ze Yang, Yinghao Xu, Han Xue, et al., “Dense RepPoints: Representing Visual Objects with Dense Point Sets,” arXiv: 1912.11473v3, 2020; (3) Yihong Chen, Zheng Zhang, Yue Cao, et al., “RepPoints v2: Verification Meets Regression for Object Detection,” arXiv: 2007.08508vl, 2020; (4) Wentong Li, Yijie Chen, Kaixuan Hu, et al., “Oriented RepPoints for Aerial Object Detection,” arXiv: 2105.1111 lv4, 2022; and (5) Nicolas Carion, Francisco Massa, Gabriel Synnaeve, et al., “End-to-End Object Detection with Transformers,” arXiv: 2005.12872v3, 2020, all of which are incorporated herein by reference in their entirety.

[0099] According to one example disclosed above, e.g., in the summary section and / or in reference to any one or any combination of some or all of FIGS. 1-18, provided is a method performed via a computing device for providing support to an imaging flow-cytometry instrument, the method comprising: receiving flow-cytometry event data and images representing a plurality of events detected with the imaging flow-cytometry instrument and corresponding to a sample including cells and beads; selecting a first group of events by applying a set of event-data-based gating operations to the plurality of events; selecting a second group of events and a third group of events from the first group of events by applying a set of image-based gating operations to the first group of events, the second group including cell events, the third group including bead events, the second and third groups having no events in common; and presenting a subset of the imagesDocket No.: TP389019WO1 corresponding to the third group of events for counting a total number of beads in said subset of the images.

[0100] In some examples of the above method, the method further comprises computing an estimate of a bead-to-cell ratio in the sample based on the total number of beads and further based on a number of events in the second group.

[0101] In some examples of any of the above methods, the third group of events includes cellbead events and bead-only events.

[0102] In some examples of any of the above methods, at least one of the cell-bead events includes two or more beads.

[0103] In some examples of any of the above methods, at least one of the bead-only events includes two or more beads.

[0104] In some examples of any of the above methods, the set of event-data-based gating operations is configured to isolate a target cell population.

[0105] In some examples of any of the above methods, the set of event-data-based gating operations includes a gating operation based on FSC and SSC signal data from the flow-cytometry event data.

[0106] In some examples of any of the above methods, the set of image-based gating operations is configured using a set of image-processing parameters selectable from a larger plurality of imageprocessing parameters; wherein the larger plurality of image-processing parameters includes more than 10 (or more than 20, or more than 25) parameters; and wherein the set of image-processing parameters has fewer than 10 (or fewer than 6) parameters.

[0107] In some examples of any of the above methods, the plurality image-processing parameters is sorted into a plurality of categories including a category of system feature parameters, a category of object feature parameters, a category of pixel feature parameters, a category of shape feature parameters, and a category of intensity feature parameters.Docket No.: TP389019WO1

[0108] In some examples of any of the above methods, the set of image-processing parameters consists of parameters selected from the category of shape feature parameters and the category of intensity feature parameters.

[0109] In some examples of any of the above methods, the set of image-processing parameters consists of: an average normalized intensity parameter; a circularity parameter; a minimum intensity parameter; a normalized intensity standard deviation parameter; and a major size parameter.

[0110] In some examples of any of the above methods, the set of image-based gating operations includes a gating operation configured to partition the third group of events into a first subgroup including single-bead events and a second subgroup including multiple-bead events.

[0111] In some examples of any of the above methods, the method further comprises: computing statistics of one or more selected gating operations from the set of image-based gating operations; and displaying the computed statistics on a display device.

[0112] In some examples of any of the above methods, the method further comprises allowing adjustments to the total number of beads based on the subset of the images.

[0113] In some examples of any of the above methods, the counting includes manual counting of the beads in the subset of the images.

[0114] In some examples of any of the above methods, the counting includes applying an automated bead counting algorithm to the subset of the images.

[0115] In some examples of any of the above methods, the counting includes feeding the subset of the images to a computer vision model.

[0116] In some examples of any of the above methods, the method is characterized by an LOD value of approximately 10 beads / mL.

[0117] In some examples of any of the above methods, the method is characterized by an LLOQ value of approximately 20 beads per 1,000,000 cells.

[0118] According to another example disclosed above, e.g., in the summary section and / or in reference to any one or any combination of some or all of FIGS. 1-18, provided is a method performed via a computing device for providing support to an imaging flow-cytometry instrument,Docket No.: TP389019WO1 the method comprising: receiving flow-cytometry event data and images representing a plurality of events detected with the imaging flow-cytometry instrument and corresponding to a sample including cells and beads; selecting a group of events by applying a set of gating operations to the plurality of events, the group of events including bead events; and processing a subset of the images corresponding to the selected group of events with a machine learning (ML) model to determine a total number of beads in the subset of the images.

[0119] In some examples of the above method, the method further comprises computing an estimate of a bead-to-cell ratio in the sample based on the determined total number of beads and further based on a number of events in the group.

[0120] In some examples of any of the above methods, the set of gating operations includes or consists of an event-data-based gating operation. In some examples, the set of gating operations consists of a single event-data-based gating operation.

[0121] In some examples of any of the above methods, the set of gating operations further includes an image-based gating operation.

[0122] In some examples of any of the above methods, the image-based gating operation is configured using a set of image-processing parameters selectable from a larger plurality of imageprocessing parameters; wherein the larger plurality of image-processing parameters includes more than 10 parameters; and wherein the set of image-processing parameters has fewer than 10 parameters.

[0123] In some examples of any of the above methods, the group of events includes cell-bead events and bead-only events.

[0124] In some examples of any of the above methods, at least one of the cell-bead events includes two or more beads.

[0125] In some examples of any of the above methods, at least one of the bead-only events includes two or more beads.

[0126] In some examples of any of the above methods, the group of events further includes cell- only events.Docket No.: TP389019WO1

[0127] In some examples of any of the above methods, the ML model is selected from the group consisting of: a computer vision machine-learning model; a convolutional neural network architecture; and a transformer encoder-decoder architecture.

[0128] In some examples of any of the above methods, the ML model is trained to recognize in the images: healthy cells as belonging to a first class of objects; unhealthy cells as belonging to a second class of objects; and beads as belonging to a third class of objects.

[0129] In some examples of any of the above methods, the total number of beads is determined by counting in the subset of the images the objects recognized by the ML model as belonging to the third class of objects.

[0130] In some examples of any of the above methods, the method further comprises generating a report characterizing the healthy and unhealthy cells in the sample based on characteristics in the subset of the images of objects recognized by the ML model as belonging to the first and second classes of objects.

[0131] In some examples of any of the above methods, the report includes one or more of: an estimated size distribution of the healthy cells; an estimated mean healthy cell diameter; an estimate of cells viability; an estimated portion of cells in clumps; and an estimated portion of beads attached to cells.

[0132] In some examples of any of the above methods, the method further comprises displaying the report on a display device.

[0133] In some examples of any of the above methods, the method further comprises selecting the ML model from a plurality of trained ML models, wherein each of the trained ML models is configured to process an output of a different respective set of gating operations.

[0134] In some examples of any of the above methods, the plurality of trained ML models includes: a first model configured to process the output of a first number of gating operations; and a second model configured to process the output of a different second number of gating operations.

[0135] In some examples of any of the above methods, the subset of the images includes at least 10,000 images or at least 100,000 images.Docket No.: TP389019WO1

[0136] A non-transitory computer-readable medium storing instructions that, when executed by the computing device, cause the computing device to perform operations comprising any one of the above methods.

[0137] According to yet another example disclosed above, e.g., in the summary section and / or in reference to any one or any combination of some or all of FIGS. 1-18, provided is an apparatus, comprising: an imaging flow-cytometry instrument; and a computing device configured to: receive flow-cytometry event data and images representing a plurality of events detected with the imaging flow-cytometry instrument and corresponding to a sample including cells and beads; select a first group of events by applying a set of event-data-based gating operations to the plurality of events; select a second group of events and a third group of events from the first group of events by applying a set of image-based gating operations to the first group of events, the second group including cell events, the third group including bead events, the second and third groups having no events in common; and present a subset of the images corresponding to the third group of events for counting a total number of beads in said subset of the images.

[0138] According to yet another example disclosed above, e.g., in the summary section and / or in reference to any one or any combination of some or all of FIGS. 1-18, provided is an apparatus comprising: an imaging flow-cytometry instrument; and a computing device configured to: receive flow-cytometry event data and images representing a plurality of events detected with the imaging flow-cytometry instrument and corresponding to a sample including cells and beads; select a group of events by applying a set of gating operations to the plurality of events, the group of events including bead events; and process a subset of the images corresponding to the selected group of events with a machine learning model to determine a total number of beads in the subset of the images.

[0139] It is to be understood that the above description is intended to be illustrative and not restrictive. Many implementations and applications other than the examples provided would be apparent upon reading the above description. The scope should be determined, not with reference to the above description, but should instead be determined with reference to the appended claims, along with the full scope of equivalents to which such claims are entitled. It is anticipated and intended that future developments will occur in the technologies discussed herein, and that the disclosed systems and methods will be incorporated into such future examples. In sum, it should be understood that the application is capable of modification and variation.Docket No.: TP389019WO1

[0140] All terms used in the claims are intended to be given their broadest reasonable constructions and their ordinary meanings as understood by those knowledgeable in the technologies described herein unless an explicit indication to the contrary is made herein. In particular, use of the singular articles such as “a,” “the,” “said,” etc. should be read to recite one or more of the indicated elements unless a claim recites an explicit limitation to the contrary.

[0141] The Abstract is provided to allow the reader to quickly ascertain the nature of the technical disclosure. It is submitted with the understanding that it will not be used to interpret or limit the scope or meaning of the claims. In addition, in the foregoing Detailed Description, it can be seen that various features are grouped together in various examples for the purpose of streamlining the disclosure. This method of disclosure is not to be interpreted as reflecting an intention that the claimed subject matter incorporate more features than are expressly recited in each claim. Rather, as the following claims reflect, inventive subject matter lies in fewer than all features of a single disclosed example. Thus, the following claims are hereby incorporated into the Detailed Description, with each claim standing on its own as a separately claimed subject matter.

[0142] Unless explicitly stated otherwise, each numerical value and range should be interpreted as being approximate as if the word “about” or “approximately” preceded the value or range.

[0143] Although the elements in the following method claims, if any, are recited in a particular sequence with corresponding labeling, unless the claim recitations otherwise imply a particular sequence for implementing some or all of those elements, those elements are not necessarily intended to be limited to being implemented in that particular sequence.

[0144] Unless otherwise specified herein, the use of the ordinal adjectives “first,” “second,” “third,” etc., to refer to an object of a plurality of like objects merely indicates that different instances of such like objects are being referred to, and is not intended to imply that the like objects so referred-to have to be in a corresponding order or sequence, either temporally, spatially, in ranking, or in any other manner.

[0145] Unless otherwise specified herein, in addition to its plain meaning, the conjunction “if’ may also or alternatively be construed to mean “when” or “upon” or “in response to determining” or “in response to detecting,” which construal may depend on the corresponding specific context. For example, the phrase “if it is determined” or “if [a stated condition] is detected” may be construed toDocket No.: TP389019WO1 mean “upon determining” or “in response to determining” or “upon detecting [the stated condition or event]” or “in response to detecting [the stated condition or event].”

[0146] Also, for purposes of this description, the terms “couple,” “coupling,” “coupled,” “connect,” “connecting,” or “connected” refer to any manner known in the art or later developed in which energy is allowed to be transferred between two or more elements, and the interposition of one or more additional elements is contemplated, although not required. Conversely, the terms “directly coupled,” “directly connected,” etc., imply the absence of such additional elements.

[0147] The functions of the various elements shown in the figures, including any functional blocks labeled as “processors” and / or “controllers,” may be provided through the use of dedicated hardware as well as hardware capable of executing software in association with appropriate software. When provided by a processor, the functions may be provided by a single dedicated processor, by a single shared processor, or by a plurality of individual processors, some of which may be shared. Moreover, explicit use of the term “processor” or “controller” should not be construed to refer exclusively to hardware capable of executing software, and may implicitly include, without limitation, digital signal processor (DSP) hardware, network processor, application specific integrated circuit (ASIC), field programmable gate array (FPGA), read only memory (ROM) for storing software, random access memory (RAM), and nonvolatile storage. Other hardware, conventional and / or custom, may also be included. Similarly, any switches shown in the figures are conceptual only. Their function may be carried out through the operation of program logic, through dedicated logic, through the interaction of program control and dedicated logic, or even manually, the particular technique being selectable by the implementer as more specifically understood from the context.

[0148] As used in this application, the terms “circuit,” “circuitry” may refer to one or more or all of the following: (a) hardware-only circuit implementations (such as implementations in only analog and / or digital circuitry); (b) combinations of hardware circuits and software, such as (as applicable): (i) a combination of analog and / or digital hardware circuit(s) with software / firmware and (ii) any portions of hardware processor(s) with software (including digital signal processor(s)), software, and memory(ies) that work together to cause an apparatus, such as a mobile phone or server, to perform various functions); and (c) hardware circuit(s) and or processor(s), such as a microprocessor(s) or a portion of a microprocessor(s), that requires software (e.g., firmware) for operation, but the software may not be present when it is not needed for operation.” This definitionDocket No.: TP389019WO1 of circuitry applies to all uses of this term in this application, including in any claims. As a further example, as used in this application, the term circuitry also covers an implementation of merely a hardware circuit or processor (or multiple processors) or portion of a hardware circuit or processor and its (or their) accompanying software and / or firmware. The term circuitry also covers, for example and if applicable to the particular claim element, a baseband integrated circuit or processor integrated circuit for a mobile device or a similar integrated circuit in server, a cellular network device, or other computing or network device.

[0149] It should be appreciated by those of ordinary skill in the art that any block diagrams herein represent conceptual views of illustrative circuitry embodying the principles of the disclosure. Similarly, it will be appreciated that any flow charts, flow diagrams, state transition diagrams, pseudo code, and the like represent various processes which may be substantially represented in computer readable medium and so executed by a computer or processor, whether or not such computer or processor is explicitly shown.

[0150] Any numerical range recited herein includes all values from the lower value to the upper value. For example, if a range is stated as 1% to 50%, it is intended that the narrower ranges thereof, such as 2% to 40%, 10% to 30%, 1% to 3%, etc., are expressly enumerated by said statement. These specific examples represent only a limited subset of what is intended to be covered, and all possible combinations of numerical values between and including the lowest value and the highest value of the enumerated range are to be considered to be expressly stated in this application. Concentration ranges, pH ranges, and other ranges of specific parameters are intended to be interpreted in a manner similar to the “%” example.

[0151] The modifier “about” or “approximately” used in connection with a quantity is inclusive of the stated value and has the meaning dictated by the context (for example, it includes at least the degree of error associated with the measurement of the particular quantity). The modifier “about” or “approximately” should also be considered as disclosing the range defined by the absolute values of the two endpoints. For example, the expression “from about 2 to about 4” also discloses the range “from 2 to 4.” The term “about” may refer to plus or minus 10% of the indicated number. For example, “about 10%” may indicate a range of 9% to 11%, and “about 1” may mean from 0.9-1.1. Other meanings of “about” may be apparent from the context, such as rounding off, so that, for example, “about 1” may also mean from 0.5 to 1.4.Docket No.: TP389019WO1

[0152] For purposes of this disclosure, the chemical elements are identified in accordance with the Periodic Table of the Elements, CAS version, Handbook of Chemistry and Physics, 75th Ed., inside cover, and specific functional groups are generally defined as described therein. Additionally, the present disclosure relies on general principles of organic chemistry, inorganic chemistry, and material science, as accepted in the pertinent arts. For example, specific functional moieties and reactivity in accordance with some of such principles are described in Organic Chemistry, Thomas Sorrell, University Science Books, Sausalito, 1999; Smith and March, March's Advanced Organic Chemistry, 5th Edition, John Wiley & Sons, Inc., New York, 2001; Larock, Comprehensive Organic Transformations, VCH Publishers, Inc., New York, 1989; Carruthers, Some Modem Methods of Organic Synthesis, 3rd Edition, Cambridge University Press, Cambridge, 1987, the entire contents of each of which are incorporated herein by reference.

[0153] “BRIEF SUMMARY OF SOME SPECIFIC EMBODIMENTS” in this specification is intended to introduce some example embodiments, with additional embodiments being described in “DETAILED DESCRIPTION” and / or in reference to one or more drawings. “BRIEF SUMMARY OF SOME SPECIFIC EMBODIMENTS” is not intended to identify essential elements or features of the claimed subject matter, nor is it intended to limit the scope of the claimed subject matter.

Claims

Docket No.: TP389019WO1CLAIMSWhat is claimed is:

1. A method performed via a computing device for providing support to an imaging flowcytometry instrument, the method comprising: receiving flow-cytometry event data and images representing a plurality of events detected with the imaging flow-cytometry instrument and corresponding to a sample including cells and beads; selecting a group of events by applying a set of gating operations to the plurality of events, the group of events including bead events; and processing a subset of the images corresponding to the selected group of events with a machine learning (ML) model to determine a total number of beads in the subset of the images.

2. The method of claim 1, further comprising computing an estimate of a bead-to-cell ratio in the sample based on the determined total number of beads and further based on a number of events in the group.

3. The method of claim 1, wherein the set of gating operations includes or consists of an event- data-based gating operation.

4. The method of claim 3, wherein the set of gating operations further includes an image-based gating operation.

5. The method of claim 4, wherein the image-based gating operation is configured using a set of image-processing parameters selectable from a larger plurality of image-processing parameters; wherein the larger plurality of image-processing parameters includes more than 10 parameters; and wherein the set of image-processing parameters has fewer than 10 parameters.

6. The method of claim 1, wherein the group of events includes cell-bead events and bead-only events.Docket No.: TP389019WO17. The method of claim 6, wherein at least one of the cell-bead events includes two or more beads.

8. The method of claim 6, wherein at least one of the bead-only events includes two or more beads.

9. The method of claim 6, wherein the group of events further includes cell-only events.

10. The method of claim 1, wherein the ML model is selected from the group consisting of: a computer vision machine-learning model; a convolutional neural network architecture; and a transformer encoder-decoder architecture.

11. The method of claim 1, wherein the ML model is trained to recognize in the images: healthy cells as belonging to a first class of objects; unhealthy cells as belonging to a second class of objects; and beads as belonging to a third class of objects.

12. The method of claim 11, wherein the total number of beads is determined by counting in the subset of the images the objects recognized by the ML model as belonging to the third class of objects.

13. The method of claim 11, further comprising generating a report characterizing the healthy and unhealthy cells in the sample based on characteristics in the subset of the images of objects recognized by the ML model as belonging to the first and second classes of objects.

14. The method of claim 13, wherein the report includes one or more of: an estimated size distribution of the healthy cells; an estimated mean healthy cell diameter; an estimate of cells viability; an estimated portion of cells in clumps; andDocket No.: TP389019WO1 an estimated portion of beads attached to cells.

15. The method of claim 13, further comprising displaying the report on a display device.

16. The method of claim 1, further comprising selecting the ML model from a plurality of trained ML models, wherein each of the trained ML models is configured to process an output of a different respective set of gating operations.

17. The method of claim 16, wherein the plurality of trained ML models includes: a first model configured to process the output of a first number of gating operations; and a second model configured to process the output of a different second number of gating operations.

18. The method of claim 1, wherein the subset of the images includes at least 10,000 images or at least 100,000 images.

19. A non-transitory computer-readable medium storing instructions that, when executed by the computing device, cause the computing device to perform operations comprising the method of any one of claims 1-18.

20. An apparatus, comprising: an imaging flow-cytometry instrument; and a computing device configured to: receive flow-cytometry event data and images representing a plurality of events detected with the imaging flow-cytometry instrument and corresponding to a sample including cells and beads; select a group of events by applying a set of gating operations to the plurality of events, the group of events including bead events; and process a subset of the images corresponding to the selected group of events with a machine learning model to determine a total number of beads in the subset of the images.

Images