Method and system for determining an indication of blood coagulability

Abstract

A computer-implemented methods and systems are described for determining a coagulability of a sample of blood. A method comprises obtaining a visual representation of a test membrane to which a blood sample has been applied and has diffused along the test membrane via capillary action. The method further comprises applying a trained machine learning model to the visual representation of the test membrane, the trained machine learning model trained on a plurality of training visual representations of training membranes to determine an indication of coagulability of a training blood sample applied to a respective training membrane. The method further comprises generating, by the trained machine learning model, an indication of coagulability of the blood sample, and issuing a notification comprising the indication of coagulability of the blood sample.

Description

Method and system for determining an indication of blood coagulabilityCross-Reference to Related Applications

[0001] The present application claims priority from Australian Provisional Patent Application No 2024902177 filed on 15 July 2024, the contents of which are incorporated herein by reference in their entirety.Technical Field

[0002] Aspects of the disclosure relate generally to systems and methods for image processing using machine learning techniques and, more specifically, to the determination of an indication of blood coagulability based on an image of blood diffusion through a porous membrane.Background

[0003] The demand for prompt, precise, and cost-efficient coagulability testing is particularly pronounced within the demanding environments of intensive care units and cardiovascular surgery settings, where the judicious administration of anticoagulants is critical. Navigating the fine line between reducing thrombotic events and minimizing bleeding risks is a complex and vital aspect of patient care. Traditional coagulation tests, such as the activated clotting time (ACT), the most commonly used point-of-care (POC) coagulation test, prothrombin time (PT), and activated partial thromboplastin time (APTT), which assess the extrinsic and intrinsic coagulation pathways, are often cumbersome and costly, and their use may be limited in urgent care situations due to time and resource-constrained settings.

[0004] Although activated clotting time, prothrombin time, and activated partial thromboplastin time provide essential data, they fall short in offering immediate guidance for precise anticoagulant dosing due to their reliance on analyses of platelet-poor plasma samples and the need for blood processing.

[0005] The optimization of antithrombotic medication is a critical component in preventing severe complications such as stroke, pulmonary embolism, and myocardial infarction. However, without rapid diagnostic methods to evaluate complex thrombotic mechanisms, significant bleeding incidents often result from excessive anti coagulation. The financial impactof coagulation status monitoring further challenges healthcare systems and patients, particularly in areas with economic disadvantages.

[0006] Coagulation testing devices, which measure the international normalized ratio (INR), offer accurate coagulation metrics but can be prohibitively expensive due to complex methods of detection, often electrochemical, and sometimes necessitate repeated testing due to complex user instructions and differences in handling.

[0007] It is desired to address or ameliorate one or more shortcomings or disadvantages associated with the prior art, or to at least provide a useful alternative hereto.

[0008] Any discussion of documents, acts, materials, devices, articles or the like which has been included in the present specification is solely for the purpose of providing a context for the present invention. It is not to be taken as an admission that any or all of these matters form part of the prior art base or were common general knowledge in the field relevant to the present invention as it existed before the priority date of each claim of this application.Summary

[0009] In accordance with an aspect of the present disclosure, there is provided a computer- implemented method for determining a coagulability of a sample of blood. The method comprises obtaining a visual representation of a test membrane to which a blood sample has been applied and has diffused along the test membrane via capillary action. The method further comprises applying a trained machine learning model to the visual representation of the test membrane, the trained machine learning model trained on a plurality of training visual representations of training membranes to determine an indication of coagulability of a training blood sample applied to a respective training membrane. The method further comprises generating, by the trained machine learning model, an indication of coagulability of the blood sample, and issuing a notification comprising the indication of coagulability of the blood sample.

[0010] In some embodiments, the trained machine learning comprises a deep learning convolutional neural network classifier, configured to extract one or more features from the visual representation of the test membrane.

[0011] In some embodiments, the method further comprises extracting one or more features from the visual representation of the test membrane, and providing the one or more features to the trained machine learning model.

[0012] In some embodiments, the the one or more features comprise one or more of: the red blood cell diffusion front; a red blood cell diffusion distance; a red blood cell diffusion area; a red blood cell diffusion rate; the plasma diffusion front; a plasma diffusion distance; a plasma diffusion area; a plasma diffusion rate; an area ratio of the plasma diffusion area to the red blood cell diffusion area; an distance ratio of the plasma diffusion distance to the red blood cell diffusion distance; a colorimetric parameter indicative of a change in a blood component composition of the blood sample; a colorimetric parameter indicative of a change in a pathology of the blood sample; and a colorimetric indicator of coagulability of the blood sample.

[0013] In some embodiments, the extracting one or more features from the visual representation of the test membrane comprises at least one of: calculating the red blood cell diffusion area based on the red blood cell diffusion front; calculating the red blood cell diffusion rate based on a first position of the red blood cell diffusion front at a first time, and a second position of the red blood cell diffusion front at a second time; calculating the plasma diffusion area based on the plasma diffusion front; calculating the plasma diffusion rate based on a first position of the plasma diffusion front at a first time, and a second position of the plasma diffusion front at a second time; calculating a ratio of the plasma diffusion rate to the red blood cell diffusion rate; calculating the red blood cell diffusion distance based on the red blood cell diffusion front; calculating the plasma diffusion distance based on the plasma diffusion front; and calculating the distance ratio of the plasma diffusion distance to the red blood cell diffusion distance.

[0014] In some embodiments, the the trained machine learning model is configured to, perform one or more of: in response to the area ratio exceeding an area threshold, determine that the indication of coagulability comprises high coagulability; in response to the area ratio not exceeding the area threshold, determine that the indication of coagulability comprises normal coagulability; and in response to the distance ratio exceeding a distance threshold, determine that the indication of coagulability comprises high coagulability.

[0015] In some embodiments, extracting the colorimetric parameter comprises determining a color intensity of a colorimetric region within the visual representation. In some embodiments,the colorimetric region comprises a region of the test membrane to which analytic dye has been applied, and through which the blood sample has diffused. In some embodiments, the analytic dye in the dyed region is configured to effect a color change in the dyed region in response to the presence of one or more of: free haemoglobin; platelet biomarkers; and fibrin biomarkers, in the blood sample.

[0016] In some embodiments, the trained machine learning model is configured to, in response to the colorimetric parameter indicating color intensity exceeding a color intensity threshold, determine the indication of coagulability comprises high coagulability.

[0017] In some embodiments, the method further comprises mapping the indication of coagulability to a medication recommendation, and and outputting the medication recommendation to a user.

[0018] In some embodiments, the method further comprises preprocessing the visual representation, wherein preprocessing the visual representation comprises one or more of: normalizing the visual representation for color intensity; normalizing the visual representation for brightness; normalizing the visual representation for contrast; and performing region of interest alignment.

[0019] In some embodiments, the trained machine learning model comprises a classifier, configured to classify the visual representation of the test membrane into one of a plurality of blood coagulability classifications. In some embodiments, the classifier has been trained, using weak supervision, the test visual representations comprise a plurality of videos, and labels are provided at the video level, and not for individual frames of a video of the plurality of videos.

[0020] In some embodiments, the trained machine learning model comprises a cross entropy loss function. In some embodiments, the trained machine learning model comprises a center loss algorithm, the center loss algorithm configured to apply a distance query to features extracted from the image of the test membrane to determine a second indication of coagulability. In some embodiments, the trained machine learning model is configured to adjust the indication of coagulability based on the second indication of coagulability.

[0021] In accordance with an aspect of the present disclosure, there is provided a computer- implemented method for determining a coagulability of a sample of blood, the method comprising receiving a visual representation of a test membrane to which a blood sample hasbeen applied and has diffused along the test membrane via capillary action to produce a red blood cell diffusion front and a plasma diffusion front. The method further comprising applying a trained machine learning model to the visual representation of the test membrane, the trained machine learning model trained on a plurality of training visual representations of training membranes to determine an indication of coagulability of a training blood sample applied to a respective training membrane. The method further comprising generating, by the trained machine learning model, an indication of coagulability of the blood sample, and issuing a notification comprising the indication of coagulability of the blood sample.

[0022] In accordance with an aspect of the present disclosure, there is provided a computer- implemented method for determining a coagulability of a sample of blood. The method comprises: receiving an image of a test membrane to which a blood sample has been applied and has diffused along the test membrane via capillary action; determining a red blood cell diffusion front; determining a plasma diffusion front; and applying a trained machine learning model to the image of the test membrane, the red blood cell diffusion front and the plasma diffusion front, the trained machine learning model configured to determine an indication of the coagulability of the blood sample.

[0023] In some embodiments, the indication of the coagulability of the blood sample comprises a classification of the coagulability of the blood sample.

[0024] In some embodiments, applying a trained machine learning model to the image of the test membrane comprises:

[0025] determining an area ratio of the plasma diffusion area to the red blood cell diffusion area.

[0026] In some embodiments, the trained machine learning model comprises a random forest classifier, configured to classify the image of the test membrane into one of a plurality of blood coagulability classifications.

[0027] In some embodiments, the trained machine learning model was trained on a dataset comprising images of recalcified citrated whole blood samples to simulate different levels of coagulability in patients.

[0028] In some embodiments, the test membrane comprises a nitrocellulose strip. In some embodiments, the test membrane is a component of a lateral flow assay.

[0029] According to another aspect of the present disclosure, there is provided a non- transitory storage medium storing machine-readable storage medium storing instructions which, when executed by one or more processors, individually or in combination, cause the one or more processors to perform a method disclosed herein.

[0030] According to another aspect of the present disclosure, there is provided a system comprising one or more processors, and memory comprising computer executable instructions, which when executed by the one or more processors, individually or in combination, cause the system to perform a method disclosed herein.Brief Description of Drawings

[0031] The embodiments of the disclosure will now be described with reference to the accompanying drawings, in which:Figure l is a block diagram of system for determining an indication of the coagulability of a blood sample, according to an embodiment;Figure 2 illustrates the components of a lateral flow assay cassette, in accordance with an embodiment;Figure 3 illustrates a membrane to which a blood sample has diffused, in accordance with an embodiment;Figure 4 illustrates another membrane to which a blood sample has diffused, in accordance with an embodiment;Figure 5 illustrates a graph of experimental data indicating the diffusion distances of the RBC component and plasma component over time as the blood components of a first blood sample diffuse through a membrane, in accordance with an embodiment;Figure 6 illustrates a graph of experimental data indicating the diffusion distances of the RBC component and plasma component over time as the blood components of a second blood sample diffuse through a membrane, in accordance with an embodiment;Figure 7 illustrates a process of applying a blood sample to a membrane of a lateral flow assay test cassette, in accordance with an embodiment;Figure 8 illustrates a plurality of test cassettes to which a blood sample has been applied and allowed to diffuse, in accordance with an embodiment;Figure 9 illustrates a process flow diagram for a computer-implemented process 900 to determine an indication of the coagulability of a blood sample, in accordance with an embodiment;Figure 10 illustrates submodules of the evaluation module, in accordance with an embodiment;Figure 11 illustrates a color test image, a greyscale representation of the test image and a binary representation of the test image, in accordance with an embodiment;Figure 12 illustrates a random forest classifier comprising a plurality of decision trees, in accordance with an embodiment;Figure 13 illustrates a table representing a plurality of different recalcification concentrations applied to citrated whole blood samples, and the corresponding coagulation classification associated with that recalcification concentrations, in accordance with an embodiment;Figure 14A illustrates a binned pixel intensity histograms for recalcification treatment group Band 1, in accordance with an embodiment;Figure 14B illustrates a binned pixel intensity histograms for recalcification treatment group Band 2, in accordance with an embodiment;Figure 15 A illustrates a binned pixel intensity histograms for recalcification treatment group Band 3, in accordance with an embodiment;Figure 15B illustrates a binned pixel intensity histograms for recalcification treatment group Band 4, in accordance with an embodiment;Figure 16 illustrates a first ROC curve graph, in accordance with an embodiment;Figure 17 illustrates a second ROC curve graph, in accordance with an embodiment;Figure 18 illustrates a precision recall graph, in accordance with an embodiment;Figure 19 illustrates a confusion matrix, in accordance with an embodiment;Figure 20 illustrates the results of an analysis of wicked diffusion distance comparing increasing concentrations of CaCL with 20 U of enoxaparin, as a bar chart represented as mean ± SD, in accordance with an embodiment;Figure 21 illustrates the results of an analysis of wicked diffusion distance comparing increasing doses of enoxaparin, as a bar chart represented as mean ± SD, in accordance with an embodiment;Figure 22 illustrates the client device 110, in accordance with an embodiment;Figure 23 illustrates a binned pixel intensity histograms for recalcification treatment group Band 5, in accordance with an embodiment; andFigure 24 illustrates a test membrane to which a blood sample has been applied and diffused along the test membrane, the test membrane comprises a strip dyed with analytic dye, in accordance with an embodiment.Description of Embodiments

[0032] Lateral flow assays (LFAs), with their simple design and use of affordable materials such as cellulose or nitrocellulose (NC), present an affordable approach to accessible diagnostics and means for point-of-care monitoring of blood coagulability. However, the results of a LFA are typically not easily interpretable, especially by untrained users. Accordingly, translating the results of a LFA test into clinical action can be challenging or may be unreliably executed.

[0033] Provided herein is a computer-implemented method and system for evaluating the coagulability of a blood sample via processing a visual representation (e.g., an image or a video) of wicked diffusion of the blood sample through a porous membrane. Some embodiments described herein utilise lateral flow assays comprising a nitrocellulose (NC) membrane.

[0034] Embodiments of the systems and methods described herein integrate machine-learning based techniques to determine the coagulability of a blood sample. Embodiments of the systems and methods described herein may produce an automated classification of coagulability of a blood sample, thus offering insights into individual red blood cell and platelet function and fibrin production of the patient’s blood.

[0035] Embodiments of the systems and methods described herein may provide enhanced insight into the coagulation status and assist with dosing precision of the antithrombotic medications. Embodiments of the systems and methods described herein may elevate the accessibility, affordability, and efficiency of coagulability testing, potentially leading toimproved patient outcomes and supporting at home monitoring and clinical decision-making processes.System architecture

[0036] Figure 1 is a block diagram of system 100 for determining an indication of the coagulability of a blood sample, according to an embodiment. The system 100 of Figure 1 provides means for implementing the method illustrated in the process flow diagram of Figure 9.

[0037] As illustrated, the system 100 may comprise: one or more client device(s) 110; external data storage 122; a server 124; and / or one or more third party server(s) 170 in communication over a network 120.

[0038] Client device 110 may comprise a mobile or handheld computing device (e.g., a computer) such as a smartphone or tablet, a laptop, or a PC, and may, in some embodiments, comprise multiple computing devices. The client device 110 may comprise one or more processor(s) 112, memory (e.g., a storage medium configured to store machine-readable instructions) 114 and / or communications interface 118. Memory 114 may be non-transitory. The processor(s) 112 may comprise one or more microprocessors, central processing units (CPUs), application specific instruction set processors (ASIPs), application specific integrated circuits (ASICs) or other processors capable of reading and executing machine-readable instructions. The processor(s) 112 may be configured to receive stored machine-readable instructions (i.e. program code) from memory 114, which when executed by the one or more processors 112, individually or in combination, cause the client device 110 to function according to the described embodiments. Client device 110 comprises one or more display screens 140, each of the one or more display screens 140 being configured to display a graphical user interface (GUI) 145 in implementing a method, such as that illustrated in Figure 22. A display device may comprise one or more individual display screens. The functionality and content of the GUI 145 is provided by the processor(s) 112, and the memory 114, which may be cooperating with application 180.

[0039] The client device 110 may be operated by a user 102. The user may control the operation of the client device via the user interface 145 and may receive output from the client device via the display screen 140.

[0040] The client device 110 may comprise a camera 155 configured to capture visual representations to be processed by application 180. The camera may be integrated with the client device or may external to the client device. In some embodiments, the camera may transmit visual representations to the client device via network 120.

[0041] The functionality of the system 100 may be defined by an application 180. In some embodiments, the application comprises a back-end and a front-end. In the embodiment illustrated in Figure 1, the back-end of application 180 is configured to execute on the server 124, and the front-end of application 180 is configured to execute on the client device 110.

[0042] Application 180 may comprise an pre-processing module 150. Alternatively, the preprocessing module 150 may be an application separate from application 180. Application 180 may comprise a evaluation module 190. Alternatively, the evaluation module 190 may be an application separate from application 180. Application 180 may be executed, in part or in full, on client device 110. Application 180 may be executed, in part or in full, on server 124. Machine-readable code (e.g. software) defining application 180 may be stored, in part or in full, on client device 110. Machine-readable code (e.g. software) defining application 180 may be stored, in part or in full, on server 124. Application 180 may receive inputs (e.g. visual representations of blood coagulability tests) from data storage 122, or from other sources internal to the server 124, internal to the client 110, or accessible over the network 120. Application 180 may store the output products (including coagulability classifications and treatment recommendations) in data storage 122, in memory 130, memory 114, and / or transmit the output products over network 122.

[0043] The memory 114 may comprise application 180 which comprises computer executable code, which when executed by the one or more processors 112, is configured to allow client device 110 to facilitate the intuitive viewing and navigation of data displayed on a screen 140 of the client device 110. The communications interface 118 facilitates communications with components of the communications interface 118 across the network 120, such as: data storage 122, server 124, and / or third party server(s) 170. The communications interface 118 may comprise a combination of network interface hardware and network interface software suitable for establishing, maintaining and facilitating communication over a relevant communication channel.

[0044] The network 120 may include, for example, at least a portion of one or more networks having one or more nodes that transmit, receive, forward, generate, buffer, store, route, switch, process, or a combination thereof, etc. one or more messages, packets, signals, some combination thereof, or so forth. The network 120 may include, for example, one or more of: a wireless network, a wired network, an internet, an intranet, a public network, a packet-switched network, a circuit-switched network, an ad hoc network, an infrastructure network, a public- switched telephone network (PSTN), a cable network, a cellular network, a satellite network, a fibre-optic network, some combination thereof, or so forth.

[0045] The data storage 122 may form part of or be local to the system 100, or may be remote from and accessible to the system 100, for example, via the communications network 120. The data storage 122 may be configured to store data associated with the system 100. The data storage 122 may be a centralised data storage.

[0046] In some embodiments, the server 124 may comprise one or more processors 126 and memory 130 storing instructions (e.g. program code) which when executed by the processor(s) 126, individually or in combination, causes the system 100 to function according to the described methods. The processor(s) 126 may comprise one or more microprocessors, central processing units (CPUs), application specific instruction set processors (ASIPs), application specific integrated circuits (ASICs) or other processors capable of reading and executing instruction code.

[0047] The memory 130 may comprise one or more volatile or non-volatile memory types. For example, memory 130 may comprise one or more of random access memory (RAM), readonly memory (ROM), electrically erasable programmable read-only memory (EEPROM) or flash memory. Memory 130 is configured to store program code accessible by the processor(s) 126.Lateral flow assays

[0048] A blood sample may be applied to a porous membrane, which wicks the blood sample along the membrane, leaving a diffusion of the blood sample through the membrane. In some embodiments, the membrane is a component of a lateral flow assay (LFA). In some embodiments, the LFA is positioned within a test cassette, which provides an external structure for the LFA. In some embodiments, the membrane may comprise a stand-alone component (e.g., a strip of nitrocellulose). In some embodiments, the membrane may comprise acomponent within a device other than a LFA test cassette. The membrane may be referred to as a test strip.

[0049] Figure 2 illustrates the components of a lateral flow assay (LSA) cassette 200, in accordance with an embodiment. The LDA cassette is illustrated as being deconstructed into an upper component 202 and a lower component 204. The lower component comprises a wicking membrane 206 (hereafter, the membrane). In some embodiments, the membrane is paper-based. In some embodiments, the membrane comprises a nitrocellulose (NC) membrane.

[0050] The membrane 220 may be overlayed by a pad 230, which may comprise a sample pad and / or a conjugate pad. In some embodiments, the membrane 220 is in abutment with the pad 230. As used herein, the line 208 represents the end of the pad 230 and the wicking start line of the membrane. In some embodiments, a label 216 is printed on the paper-based NC membrane 206 indicating the type of lateral flow assay.

[0051] The upper component 202 of the lateral flow assay cassette comprises a lid which, in use, is affixed on top of the lower component 204 by protrusions (not shown) located on the underside of the upper component 202 which engage with recesses 218 located on the lower component 204.

[0052] The upper component of the lateral flow assay cassette comprises a sample recess 210, to which, when the cassette is in use, one or more drops of a blood sample may be deposited. The upper component of the lateral flow assay cassette further comprises a viewing window 212, which, when the cassette is in use, a user can view the diffusion of the blood sample along the membrane 206. The upper component of the lateral flow assay cassette further comprises another viewing window 214, via which a user can view the label 216.Diffusion of blood along LFA

[0053] The capillary action of the blood sample through the membrane led to the separation of the blood into its constituent components: red blood cells (RBCs); and plasma. The plasma comprises water, electrolytes, proteins (coagulation factors etc.) and immunoglobulins . The RBC blood component may comprise blood components other than red blood cells. Similarly, the plasma blood component may comprise blood components other than plasma.

[0054] Figure 3 illustrates a membrane 302 to which a blood sample has diffused, in accordance with an embodiment. The diffusion of the blood sample is illustrated from thewicking start line 304. The diffusion of the red blood cells is illustrated as a hashed portion 306 with the RBC diffusion front illustrated as line 312.

[0055] The plasma component of the blood sample has diffused further through the membrane. The diffusion of the plasma component is illustrated as dotted portion 308, with the plasma diffusion front illustrated as line 316.

[0056] A diffusion front comprises a line indicating where, on the membrane, the blood component has diffused to at a point in time (or has ceased diffusing, in the case of a diffusion performed to completion). The diffusion front may extend from one side of the membrane to the opposite side of the membrane. Alternatively, in some embodiments, the diffusion front may comprise a portion of an entire diffusion front that spans from one side of the membrane to the other side of the membrane. The diffusion front may comprise a point on the membrane, rather than a line. For example, the diffusion front may be determined as point on the left hand side of the membrane at which the diffusion has ceased.

[0057] A diffusion distance, for either the RBC component or the plasma component, may be determined as a distance from a predefined reference point to the respective diffusion front. For example, a diffusion distance, for either the RBC component or the plasma component, may be determined as a distance between the wicking start line 304 and the respective diffusion front. In some embodiments, the diffusion distance 310 of the RBC component of the blood sample is measured from the wicking start line 304 to the peak RBC diffusion front 312. In some embodiments, the diffusion distance 314 of the plasma component of the blood sample is measured from the wicking start line 304 to the peak plasma diffusion front 316.

[0058] In Figure 3, the RBC diffusion front and the plasma diffusion front are illustrated as arced lines, which are symmetrical along the length of the membrane. In practice, however, the diffusion front is typically an irregular line that extends from one side of the membrane to the other side of the membrane.

[0059] Figure 4 illustrates another membrane 402 to which a blood sample has diffused, in accordance with an embodiment. The RBC diffusion front 412 is an irregularly formed line. Similarly, the plasma diffusion front 416 is an irregularly formed line.

[0060] In some embodiments, the application 180 is configured to determine a diffusion distance by determining an average diffusion distance from the wicking start line 404 acrossthe diffusion front. For example, the diffusion distance for the RBC component may be determined as average distance 410, and the diffusion distance for the plasma component may be determined as average distance 414. In some embodiments, the application 180 is configured to determine a diffusion distance by determining a weighted average diffusion distance from the wicking start line 404 across the diffusion front.

[0061] In some embodiments, the application 180 is configured to determine a diffusion distance by determining a peak diffusion distance from the wicking start line 404 across the diffusion front. For example, the diffusion distance for the RBC component may be determined as peak distance 480, and the diffusion distance for the plasma component may be determined as peak distance 460.

[0062] In some embodiments, the application 180 is configured to determine a midline diffusion distance by determining a diffusion distance at the midline of the test membrane (e.g., along line 470).

[0063] In some embodiments, the application 180 is configured to determine an RBC diffusion area and / or a plasma diffusion area. A diffusion area comprises a surface area of the test membrane over which the blood component has diffused. Notably, the plasma diffusion area will be greater than the RBC diffusion area as the plasma diffusion area encompasses and extends beyond the RBC diffusion area.

[0064] A diffusion area may be determined based on the width of the test membrane and any one or more of: peak diffusion distance; average diffusion distance; and midline diffusion distance.Diffusion over time

[0065] Figures 5 and 6 each illustrate a graph (respectively graph 500 and graph 600) of experimental data indicating the diffusion distances (e.g., wicked distances, in millimetres) of the RBC component and plasma component over time as the blood components diffuse through a membrane, in accordance with an embodiment. Graph 500 illustrates the diffusion distances for a blood sample comprising 0 mM CaCh recalcification concentration, to simulate normal blood coagulability. Graph 600 illustrates the diffusion distances for a blood sample comprising 100 mM CaCh recalcification concentration, to simulate severe blood coagulability.

[0066] Graphs 500 and 600 demonstrate that, under both recalcification conditions, the plasma components travel faster and further than the RBC diffusion front. The diffusion distances of the RBC component stabilizes (saturates) after approximately 8 mins and 3.5 mins for 0 mM and 100 mM recalcification, respectively, reaching maximum diffusion rates of 2.11 mm / min and 0.684 mm / min. This stabilisation occurs alongside a decreasing gradient as the membrane becomes saturated.

[0067] Plasma diffusion rates significantly exceed those observed at the RBC diffusion front, peaking at 3.34 mm / min and 1.78 mm / min for 0 mM and 100 mM recalcification, respectively, and saturating simultaneously with the RBC front.

[0068] The distance travelled and the rate of diffusion of both the RBC and plasma diffusion fronts adhere to the relationships governed by Washbum's equation [References 1 , 2] and Darcy’s Law [Reference 3], Furthermore, increasing the CaCh concentration to 100 mM dramatically stimulates the coagulation process, enhancing the dynamic viscosity of both the RBC and plasma components resulting in shorter diffusion distances and slower rates of diffusion as the blood transitions from a viscoelastic liquid to solid gel, resulting in significant changes in the rheological properties. These changes occur as a result of CaCL-stimulated conversion of fibrinogen into fibrin, increasing RBC aggregation, a major determinant of blood viscosity, particularly under low shear conditions and as such, larger effects changes can be observed in the RBC diffusion front. These findings suggest that observing changes in RBC diffusion distance may serve as a more sensitive metric of coagulation and changes in blood viscosity, a metric which has been linked to many cardio vascular diseases and historically been under-utilised due to difficulties in dynamic measurements.Pore size effects

[0069] The stages of dynamic variation in blood viscosity may be better distinguished by increasing the sensitivity of paper-based LFAs by adjusting the porosity and hydrophobicity of the fluid membrane. Smaller pores may result in greater capillary forces and slower flow rates because the liquid must overcome greater resistance to move through the narrow passages. Conversely, larger pores may allow for faster flow rates as there is less resistance to fluid movement, as it influences the effective diffusion coefficient. For paper-based LFAs with very small pore sizes (<1 pm), the blood cell suspensions are not able to diffuse, thus decreasing the sensitivity of the assay. Testing process

[0070] Figure 7 illustrates a process of applying a whole blood sample 702 to a membrane of a lateral flow assay test cassette, in accordance with an embodiment. In particular, Figure 7 illustrates a first test cassette 710, to which a sample of blood 702 from vial 701 is being transferred to membrane 706 via pipette 704. By way of example, the volume of the blood sample applied to the membrane 706 is approximately 10 microlitres.

[0071] In some embodiments, the blood sample 702 in the vial 701 comprises citrated whole blood. In some embodiments, a sample of non-citrated whole blood from a patient may be applied to a LFA test cassette. The blood sample may be applied from a vial or from a pin prick extraction from the patient.

[0072] The diffusion 722 of the blood sample 702 is depicted in a second instance 720 of first cassette 710, depicted at 10 minutes after the blood sample 702 has been applied to membrane 706.

[0073] Figure 7 also illustrates a second cassette 730, to which a second blood sample has been applied and allowed to diffuse 724 for 10 minutes. Notably, for experimental purposes, the blood sample 702 applied to cassette 730 comprises citrated whole blood which has been treated with calcium chloride (CaCL) to simulate a blood coagulability level higher than normal blood coagulability. The recalcified blood applied to the second test cassette 730 has a shorter diffusion distance than the diffusion distance of the blood 702 applied to the first test cassette 720.

[0074] Notably, Figure 7 does not illustratively distinguish between the diffusion of the RBC component of the blood sample and the plasma component of the blood sample. In other words, for simplicity of illustration, Figure 7 illustrates a single diffusion front, rather than a RBC diffusion front and a plasma diffusion front.Classification

[0075] To monitor a patient’s health and to provide appropriate medical intervention, is desirable to determine an evaluation (e.g. in the form of a grading, classification, categorisation, a measure, or a qualitative or quantitative indication) of the coagulability of a blood sample. An evaluation of a blood sample may provide insight whether the patient has a blood coagulation disorder, requiring treatment.

[0076] For example, an evaluation of the coagulability of a blood sample may be determined based on the diffusion front of the blood sample applied to the test membrane, as a blood sample with normal coagulability will diffuse further through a test membrane than a blood sample with a higher coagulability through the same type of test membrane. Similarly, an evaluation of the coagulability of a blood sample may be determined based on other features of the blood diffusion along the test membrane, the color of the diffused blood sample, or a color change effected as the diffusing blood sample reacts with a dye strip of the test membrane.

[0077] Figure 8 illustrates a plurality of test cassettes to which a blood sample has been applied and allowed to diffuse, in accordance with an embodiment. The blood sample applied to test cassette 802 has low coagulability, as indicated by the long diffusion distance. In contrast, the blood sample applied to test cassette 804 has high blood coagulability, as indicated by the short diffusion distance.

[0078] Notably, Figure 8 does not illustratively distinguish between the diffusion of the RBC component of the blood sample and the plasma component of the blood sample. In other words, for simplicity of illustration, Figure 8 illustrates a single diffusion front, rather than a RBC diffusion front and a plasma diffusion front.Evaluation process

[0079] Methods described here employ image-based machine learning to determine an evaluation of the coagulability of a blood sample

[0080] Figure 9 illustrates a process flow diagram for a computer-implemented process 900 to determine an indication of the coagulability of a blood sample, based on a visual representation of a test membrane to which the blood sample has been applied and has diffused through, in accordance with an embodiment. Process 900 may be performed by application 180 on client device 110. In some embodiments, process 900 is performed in part on a client device 110 and in part on a server 124.

[0081] Figure 9 illustrates process steps performed in an illustrative method, and may not recite the complete process or all operations of the method. Although various operations of process 900 are described below and depicted in Figure 9, the operations need not necessarily all be performed, and in some cases may be performed simultaneously or in a different order than the order shown.

[0082] In operation 902, the system 100 is configured to obtain one or more visual representations of a test membrane to which a blood sample has been applied and allowed to diffuse. The one or more visual representations may comprise one or more still images (e.g., photos) depicting at least part of the test membrane. The one or more visual representations may comprise one or more videos depicting at least part of the test membrane.

[0083] The one or more visual representations may be obtained by a camera of the system 100. For example, the one or more visual representations may be obtained by the client device 110 via camera 155. For example, following the step of applying the blood sample to the test membrane, the user takes an image which includes the test membrane on their mobile platform. In other example, following the step of applying the blood sample to the test membrane, the system 100 obtains a video of the test membrane. The visual representation (e.g., image and / or video) may also include a depiction of a test cassette or surrounding objects.

[0084] The application 180 provides the visual representation to the pre-processing module 150 of the application 180. In embodiments, the visual representation may be stored in data storage 120 before being provided to the pre-processing module 150. In some embodiments, the pre-processing module may obtain the image via network 122. In embodiments, in which the pre-processing module is executed on the server 124, the visual representation may be transmitted, by the client device, via network 120, to the pre-processing module 150 on the server 124.

[0085] In some embodiments, the visual representation of a test membrane comprises a digital file. The digital file may be in the format of a Joint Photographic Experts Group (JPEG) format, a Portable Network Graphics (PNG) format, a Tag Image File Format (TIFF), a raw image, or another digital format.Pre-processing

[0086] Optionally, to facilitate the automated determination of an indication of coagulability of a blood sample, in operation 904 the pre-processing module 150 is configured to process the received visual representation (e.g., image or video of a test membrane) to produce a visual representation to be input into the evaluation module 190.

[0087] In some embodiments in which the visual representation comprises an image, the preprocessing module 150 is configured to process a received image by applying image registrationand segmentation to isolate the image of the test membrane (e.g., nitrocellulose membrane) and to standardize the input for consistency across the different test runs.

[0088] In some embodiments, the pre-processing module 150 is configured to apply cropping and / or isolating to the test image to produce a test image that only depicts the test membrane. For example, pre-processing may comprise cropping the test image such that the resulting test image only depicts the portion of the membrane from the wi eking start line (e.g., 304) to above the plasma diffusion point (e.g., 316). In embodiments in which the membrane is positioned in a LFA test cassette, the application 180 may crop the test image to remove depictions of the test cassette.Evaluation module

[0089] Figure 10 illustrates submodules of the evaluation module 190, in accordance with an embodiment. The evaluation module 190 receives a visual representation 1002 of a test membrane from the pre-processing module 150.

[0090] The evaluation module 190 comprises a data augmentation module 1004, configured to augment the received visual representation 1002 to produce an augmented visual representation. In some embodiments, data augmentation module 1004 is configured to apply augmentation techniques such as, but not limited to: rotation; translation; scaling; application of a filter; color jittering; adjusting the orientation of the raw visual representation 1002; adjusting the size of the visual representation 1002; upscaling or downscaling the visual representation 1002; applying a filter to the visual representation 1002, for example to compensate for lighting effects; applying color normalization techniques to the visual representation 1002; or any combination thereof.

[0091] In response to the visual representation 1002 comprising a video, the data augmentation module 1004 may be configured to apply techniques such as, but not limited to: frame rate reduction; frame interpolation; extraction of still images from the video; slicing the video into a plurality of clips with overlapping or non-overlapping temporal windows; or any combination thereof.

[0092] In some embodiments, the data augmentation module 1004 is configured to perform temporal jittering of the visual representation 1002. In some embodiments, the data augmentation module 1004 is configured to adjust the visual representation 1002 by performing frame skipping. The data augmentation module 1004 may adjust the visual representation 1002by spatial cropping. The data augmentation module 1004 may perform biological domainspecific transformations for video pre-processing.

[0093] The data augmentation module 1004 may adjust the visual representation 1002 by performing region-of-interest (ROI) alignment, to align the visual representation in accordance with one or more regions of interest. The region-of-interest may comprise one or more of the plasma diffusion front and the RBC diffusion front. The region-of-interest may comprise a predefined colorimetric region 2404 within the visual representation. The region-of-interest may comprise a dyed region 2440 within the visual representation.

[0094] The data augmentation module 1004 may adjust the visual representation 1002 by performing color intensity normalization to mitigate variations in illumination caused by differing ambient lighting conditions or camera settings.Feature extractor

[0095] The evaluation module 190 comprises a feature extractor 1006. In some embodiments, in operation 906, the feature extractor 1006 is configured to extract one or more features 1008 from the augmented visual representation 1002. The extracted features 1008 may include, but are not limited to: the first-order shape and region features; and grey-level co-occurrence matrix (GLCM) features.

[0096] In some embodiments, the feature extractor 1006 is configured to determine one or more binned pixel intensity histograms of the visual representation 1002. The histograms may spanned pixel intensities from 0 (black) to 255 (white),

[0097] Advantageously, the inventors determined that blood samples with a higher coagulation produced larger ratios of plasma diffusion area to red blood cell diffusion area compared to blood samples with a lower coagulation. Similarly, the inventors determined that blood samples with a higher coagulation produced larger ratios of plasma diffusion distance to red blood cell diffusion distance compared to blood samples with a lower coagulation.

[0098] The feature extractor 1006 is configured to extract the location and / or form of the RBC diffusion front and the location and / or form plasma diffusion front from the visual representation 1002.

[0099] Preferably, the feature extractor 1006 is configured to extract, from the visual representation 1002 on or more of: a RBC diffusion distance (which may comprise one or moreof: a RBC peak diffusion distance; a RBC average diffusion distance; and a RBC midline diffusion distance; or a combination thereof). Preferably, the feature extractor 1006 is configured to extract from the visual representation 1002 a plasma diffusion distance (which may comprise one or more of: a plasma peak diffusion distance; a plasma average diffusion distance; a plasma midline diffusion distance; or a combination thereof).

[0100] Preferably, the feature extractor 1006 is configured to extract, from the visual representation 1002, a RBC diffusion area. Preferably, the feature extractor 1006 is configured to extract, from the visual representation 1002, a plasma diffusion area.

[0101] In some embodiments, the feature extractor 1006 is configured to determine, as a feature, a ratio (e.g. a distance ratio) comprising the plasma diffusion distance and the RBC diffusion distance. In some embodiments, the distance ratio comprises a ratio of the plasma diffusion distance to the RBC diffusion distance. Advantageously, it is noted that the distance ratio of a test image may be correlated with coagulation severity.

[0102] In some embodiments, the evaluation module 190 is configured to determine, as a feature, a ratio (e.g. an area ratio) comprising the plasma diffusion area and the RBC diffusion area. In some embodiments, the area ratio comprises a ratio of the plasma diffusion area to the RBC diffusion area. Advantageously, it is noted that the area ratio of a visual representation 1002 may be correlated with coagulation severity.

[0103] In some embodiments, the feature extractor 1006 is configured to determine, as a feature, a difference (e.g., a distance difference) between the plasma diffusion distance and the RBC diffusion distance. In some embodiments, the distance difference comprises the plasma diffusion distance minus the RBC diffusion distance. Advantageously, it is noted that the distance difference of a visual representation may be correlated with coagulation severity.

[0104] In some embodiments, the evaluation module 190 is configured to determine, as a feature, a difference (e.g. an area difference) between the plasma diffusion area and the RBC diffusion area. In some embodiments, the area difference comprises the plasma diffusion area minus the RBC diffusion area. Advantageously, it is noted that the area difference of a visual representation may be correlated with coagulation severity.

[0105] In some embodiments, the feature extractor 1006 is configured to determine, as a feature of the visual representation 1002, a plasma diffusion rate, wherein a plasma diffusionrate is indicative of a rate at which the plasma diffusion front progresses along the test membrane as the blood sample diffuses along the test membrane via capillary action. In some embodiments, a plasma diffusion rate is indicative of a rate at which the plasma diffusion area increases as the blood sample diffuses along the test membrane via capillary action.

[0106] In some embodiments, the feature extractor 1006 determines the plasma diffusion rate by determining the difference between the position of the plasma diffusion front at a first time, and the position of the plasma diffusion front at a second time. In an example wherein the visual representation 1002 comprises a plurality of images, and an indication of the times at which each of the plurality of images was obtained. Accordingly, the feature extractor 1006 may be configured to determine the plasma diffusion rate by determining the position change of the plasma diffusion front over a period of time. In an example, wherein the visual representation 1002 comprises a video, the feature extractor 1006 may be configured to determine the position change of the plasma diffusion front over a plurality of video frames, with reference to the frame rate of the video.

[0107] In some embodiments, the feature extractor 1006 is configured to determine, as a feature of the visual representation 1002, a RBC diffusion rate. In some embodiments, a RBC diffusion rate is indicative of a rate at which the RBC diffusion front progresses along the test membrane as the blood sample diffuses along the test membrane via capillary action. In some embodiments, a RBC diffusion rate is indicative of a rate at which the RBC diffusion area increases as the blood sample diffuses along the test membrane via capillary action.

[0108] In some embodiments, the feature extractor 1006 determines the RBC diffusion rate by determining the difference between the position of the RBC diffusion front at a first time, and the position of the RBC diffusion front at a second time. In an example wherein the visual representation 1002 comprises a plurality of images, and an indication of the times at which each of the plurality of images was obtained. Accordingly, the feature extractor 1006 may be configured to determine the RBC diffusion rate by determining the position change of the RBC diffusion front over a period of time. In an example, wherein the visual representation 1002 comprises a video, the feature extractor 1006 may be configured to determine the position change of the RBC diffusion front over a plurality of video frames, with reference to the frame rate of the video.

[0109] In some embodiments, the feature extractor 1006 is configured to determine, as a feature of the visual representation 1002, a ratio of a plasma diffusion rate to a RBC diffusion rate.Greyscale and binary

[0110] Figure 11 illustrates a color test image 1110, a greyscale representation 1120 of the test image and a binary 1130 representation of the test image, in accordance with an embodiment. Colour test image 1110 depicts a membrane, to which a blood sample has been applied and allow to diffuse. The test image 1110 depicts the portion of a membrane from the wicking start line to above the plasma diffusion front 1112.

[0111] In operation 906, the feature extractor 1006 converts the color test image 1110 to greyscale test image 1120 and extracts features from the greyscale test image 1120. In some embodiments, the feature extractor 1006 converts the color test image 1110 to a binary test image 1130 and extracts features from the binary test image 1130.

[0112] The RBC diffusion front 1114 and the plasma diffusion front 1112 are discernible (e.g., by the human eye, or via computer aided image processing) from the color test image 1110. Similarly, the RBC diffusion front 1124 and the plasma diffusion front 1122 are discernible from the greyscale test image 1120. In the binary test image 1134, only the RBC diffusion front 1134 is discernible.Coagulability evaluator

[0113] The evaluation module comprises a coagulability evaluator 1010. The coagulability evaluator 1010 comprises a trained machine learning model.

[0114] In operation 908, the evaluation module 190 is configured to apply the extracted features 1008 to the trained machine learning model 1010 to determine an indication of the coagulability of the blood sample depicted in image 1002. The trained machine learning model may comprise one or more of, but not limited to: an artificial neural network; a convolutional neural network; a logistic regression model; a linear regression model; a classifier; a decision tree; a random forest classifier; a K-Nearest Neighbour (KNN) model; a deep learning algorithm; or any combination thereof. The trained machine learning model may be trained by supervised or unsupervised learning techniques.

[0115] In some embodiments, the trained machine learning model 1010 comprises an ensemble learning algorithm. The ensemble learning algorithm may aggregate two or more learners (e.g. regression models, neural networks) in order to produce better predictions.

[0116] In some embodiments, the trained machine learning model comprises a classifier, trained to classify the test image 1002 into one of a plurality of blood coagulability classifications, based on the features 1008 extracted by the feature extractor 1006. In some embodiments, the coagulability evaluator 1010 comprises a decision tree classifier.

[0117] In one embodiment, the coagulability evaluator 1010 comprises a trained random forest classifier 1200 comprising a plurality of decision trees, as illustrated in Figure 12. The random forest classifier is configured to classify the test image into one of a plurality of blood coagulability classifications. In one embodiment, the plurality of blood coagulability classifications comprises: Normal 1202; Mild 1204; Moderate 1206; Significant 1208; and Severe 1210.

[0118] The evaluator 1010 has been trained by applying a plurality of extracted features from a plurality of test images depicting the diffusion of blood samples with varying levels of coagulability. Each of the plurality of test images has been labelled with a blood coagulability classification in accordance with the recalcification concentration applied to the blood sample depicted in that test image.

[0119] The training features encompassed a broad spectrum of variables, including statistical metrics (mean, min-max), entropy, geometric properties (area, major axis length), color features, color intensity features and textural features (grey -level co-occurrence matrix).

[0120] The extracted features 1008 of the test image 1002 are applied to each of the plurality of trained decision trees of the evaluator 1010. The decision trees independently predict the coagulability classification of the blood sample used to produce the test image. The coagulability classification with the most votes from the decision trees is chosen as the predicted classification 1020 (e.g., the coagulation evaluation, or the indication of coagulability).

[0121] In one embodiment, the predicted classification 1020 was refined using a cross entropy loss (e.g., logarithmic loss) algorithm between the prediction 1020 and ground truth 1012.

[0122] In one embodiment, the evaluation module 190 further comprises a center loss algorithm 1070, configured to determine a coagulability classification based on the distancebetween the features of the test image, the class centers for a training set 1080 of test images with various coagulability classification. Class centers from the training set were selected as the support set. The features of the test images serve as the distance query 1014, and the probability of classification is obtained by measuring the distance from the query to each center point 1016 in the support set. Preferably, the evaluation module 190 determines the coagulation classification 1040 (e.g., the indication of coagulability) by balancing the predicted classification 1020 from the classifier 1010 with the predicted classification 1030 from the center loss algorithm 1070.

[0123] In some embodiments, the evaluation module 190 comprises a trained machine learning model which is trained to, in response to the ratio (e.g., area ratio) of plasma diffusion area to RBC diffusion area exceeding an area threshold, determine that the indication of coagulability comprises higher than normal coagulability (e.g., Mild, Moderate, Significant or Severe). In some embodiments, the evaluation module 190 comprises a trained machine learning model which is trained to, in response to the ratio (e.g., distance ratio) of plasma diffusion distance to RBC diffusion distance exceeding a distance threshold, determine that the indication of coagulability comprises higher than normal coagulability (e.g., Mild, Moderate, Significant or Severe).

[0124] In some embodiments, the evaluation module 190 comprises a trained machine learning model which is trained to, in response to the ratio (e.g., area ratio) of plasma diffusion area to RBC diffusion area not exceeding a threshold, determine that the indication of coagulability comprises a normal coagulability. In some embodiments, the evaluation module 190 comprises a trained machine learning model which is trained to, in response to the ratio (e.g., distance ratio) of plasma diffusion distance to RBC diffusion distance not exceeding a threshold, determine that the indication of coagulability comprises normal coagulability.

[0125] In a preferred embodiment, the evaluation module 190 comprises a deep learning convolutional neural network classifier. The deep learning convolutional neural network is configured to extract deep features to determine a indication of coagulability of a blood sample.Training datasets

[0126] In some embodiments, a training dataset is compiled to train the coagulability evaluator 1010 to determine an indication of coagulability of a blood sample. The training dataset comprises a plurality of training visual representations, wherein each training visualrepresentation is a visual representation of a test membrane (e.g., a training membrane) to which a blood sample (e.g., a training blood sample) has been applied and allowed to diffuse. The training visual representations may be labelled with features to facilitate supervised learning of the coagulability evaluator 1010.

[0127] To train an embodiment of the coagulability evaluator 1010, a training set of visual representations were obtained by applying the testing procedure to samples of citrated whole blood to which different recalcification concentrations had been applied to represent different levels of blood coagulability. In particular, the blood samples were treated with various concentrations of calcium chloride (CaCL) solutions, (e.g., 0 mM, 25 mM, 50 mM, 75 mM and 100 mM) to simulate a set of blood samples having various blood coagulabilities. The various concentrations of CaCL solutions cause the blood sample to exhibit different blood coagulabilities, with the higher concentrations of CaCL cause the blood to have a higher blood coagulability.

[0128] Figure 13 illustrates a table 1300 representing a plurality of different recalcification concentrations applied to citrated whole blood samples, and the corresponding coagulation classification associated with that recalcification concentrations, in accordance with an embodiment. In particular, for a sample comprising citrated whole blood treated with a recalcification concentration of 50 mM, the evaluation module is trained to determine a coagulation classification of Moderate. Similarly, for a sample comprising citrated whole blood treated with a recalcification concentration of 75 mM, the evaluation module is trained to determine a coagulation classification of Significant.

[0129] The machine learning model of the evaluation module 190 was trained on features extracted from a training set of visual representations. The features extracted from a training set of visual representations included the diffusion distance of the RBC front through detailed binned pixel intensity histograms for each recalcification treatment group, as illustrated in Figure 14A (Band 1, coagulation severity “Normal”), Figure 14B (Band 2, coagulation severity “Mild), Figure 15A (Band 3, coagulation severity “Moderate”), Figure 15B (Band 4, coagulation severity “Significant”), and Figure 23 (Band 5, coagulation severity “Severe”). These histograms spanned pixel intensities from 0 (black) to 255 (white), capturing the gamut from non-recalcified to 100 mM recalcified blood.

[0130] Notably, there was a discernible shift toward higher pixel intensities with increasing CaCl concentrations, signifying a reduction in the diffusion distance of the RBC front. Specifically, the average pixel intensity escalated from 125.3 in the non-recalcified cohort to 162.8 in the 100 mM CaCh group.

[0131] Using a normalised area, the binned pixel intensities of 20 different repeats of each concentration were plotted on an image histogram. Feature extraction was performed subsequently, first by defining specific functions to obtain some first order statistics, shape and region and GLCM texture features.

[0132] A “for” loop was defined to run through the different sets of the stored images in each of the 5 collections, and through data augmentation exceeding >100 images and storing the extracted features into one variable through ’v-stacking’. A feature data frame was then produced using the “pandas" function containing the extract features from the 5 numerical bands representing coagulation severity with associated labels “Normal”, “Mild”, “Moderate”, “Significant” and “Severe” corresponding to the different recalcification concentrations in ascending order. The produced data frame was then split into training and testing sets based on 0.3 allocation, before a random forest learning model 1010 was trained using the training data and associated labels.

[0133] A separate verification dataset was then used to test the model 1010 using blinded recalcification amounts, and the prediction metrics of precision, recall and Fl -score were displayed in a classification report. Receiver operating characteristics curves, precision-recall curves for the different classes (associated labels) and across different sized test sets (0.2, 0.5, 0.7) were plotted to assess the quality of the classification model, and a confusion matrix outlining the proportions of correct and incorrect predictions was produced.

[0134] In some embodiments, a training dataset is compiled to train the coagulability evaluator 1010 to determine an indication of coagulability of a blood sample to which an analytic dye has been applied to the test membrane. Such a training set comprises training visual representations of test membranes comprising one or more dye strips to which a blood sample has been applied.

[0135] In some embodiments, color jittering techniques may be applied to training visual representations, to expand the training dataset. Color jittering may comprise randomly, or strategically, altering color properties such as brightness, contrast, saturation or hue.Advantageously, training an embodiment of the coagulability evaluator 1010 with a training dataset that has been expanded through the application of color jittering may result in in improved model accuracy and performance of the coagulability evaluator 101. For example, training the coagulability evaluator 1010 with a training dataset that has been expanded through the application of color jittering may result in the coagulability evaluator 1010 becoming less sensitive to variations in lighting conditions and color tones of visual representations.Medication recommendation

[0136] A high blood coagulation level may represent a higher risk of blood clots and may indicate that anticoagulant medication should be administered to the patient. In contrast, a low blood coagulation level may represent a higher risk of excessive internal or external bleeding if a blood vessel is injured. A low blood coagulation level may indicate that medication should be administered to normalise the patient’s blood coagulation.

[0137] In operation 910, the system 100 is configured to issue a notification of the indication of coagulability 1020 to the user. Issuing the notification of the coagulability 1020 may comprise one or more of, but not limited to: displaying the indication of coagulability 1020 on a display screen 140 of the client device 110; storing the indication of coagulability 1020 in a data storage location (e.g., in data storage 122); transmitting the indication of coagulability 1020 via network 120.

[0138] Optionally, in operation 912, the application 190 is configured to determine a treatment recommendation in light of the indication of coagulability 1040 determined by the evaluation module 190. The treatment recommendation may comprise a medication recommendation.

[0139] Figure 22 illustrates the client device 110, in accordance with an embodiment. The client device comprises a smartphone, on which the GUI of the application 180 is displayed. Figure 22 illustrates the GUI displayed by the application 180 in response to the evaluation module 190 issuing a notification of a indication of coagulability 1020, in operation 910. The GUI may be displayed on display screen 140. In this embodiment, the indication of coagulability comprises a classification, which classifies the blood sample as having a High coagulability. The indication of coagulability is displayed as item 2204.

[0140] In response to the indication of coagulability indicating a coagulability other than a normal coagulability, the application 180 may be configured to display recommended actionsto the user (e.g., the patient or caregiver). Recommendation 2206 comprises a recommendation to administer medication with the view to normalising the patient’s blood coagulability.

[0141] In some embodiments, the application 180 determine a medication recommendation based on the indication of coagulability. In some embodiments, the application 180 determine a medication recommendation based on the indication of coagulability in conjunction with one or more of: patient medical history; patient physiology; patient age; patient weight; or any combination thereof. The medication recommendation may comprise: medication type; medication name; dosage, frequency of administration; method of administration; or any combination thereof.

[0142] In response to the indication of coagulability indicating a coagulability other than a normal coagulability, the application 180 may be configured to provide a recommendation 2208 to test another sample of the patient’s blood at a later time. The later time may be after medication has been administered to the patient.

[0143] The application 180 may also provide a user interface item 2202 to facilitate capturing an image of another test membrane. In response to the user selecting user interface item 2202, the application 180 will cause the camera 145 to capture an image, ostensibly of a test membrane to which a sample of the patient’s blood has been applied and allowed to diffuse.Colorimetric analysis

[0144] In some embodiments, a colorimetric analysis of the image of the test membrane may be performed by the system 100, to determine an indication of the coagulability of a blood sample. Colorimetric analysis comprises a technique to determine an indication of the coagulability of a blood sample by measuring the intensity of the color of the blood sample, as diffused along a test membrane.Colorimetric analysis using analytic dye

[0145] In some embodiments, an analytic dye is applied to the test membrane 302 such that the blood sample reacts with the analytic dye to effect a colour change in the blood sample as the blood sample diffuses along the test membrane.

[0146] The analytic dye may be applied to a specific region (e.g., a dyed region) of the test membrane. In some embodiments, the analytic dye in the dyed region reacts with haemoglobin in a blood sample to effect a color change in the dyed region of the test membrane. In bloodsamples exhibiting haemolysis (rupture of red blood cells), causing the presence of free haemoglobin in the sample, the analytic dye will react with the free haemoglobin to effect a color change in the dyed region.

[0147] In some situations, the presence of free haemoglobin in the plasma of a blood sample is indicative of the occurrence of haemolysis in the blood sample. Furthermore, in some situations, haemolysis may be indicative of a hypercoagulable state of the blood sample. Accordingly, determining an indication of haemolysis in a blood sample can provide an indication of coagulability of a blood sample.

[0148] In some embodiments, a greater color intensity of dyed region, through which the blood sample has diffused and reacted with the analytic dye, is indicative of a larger volume of free haemoglobin in the blood sample. Conversely, the lesser color intensity may be indicative of a lower volume (or absence) of free haemoglobin in the blood sample.

[0149] In some embodiments, the analytic dye comprises o-Tolidine (also referred to as orthotolidine). In some embodiments, the analytic dye comprises o-Tolidine in combination with hydrogen peroxide.

[0150] In some embodiments, the analytic dye comprises one or more of: horseradish peroxidase; tetramethylbenzidine; hydrogen peroxide; and chromozym TH. Advantageously, platelet function (PAC-1 signalling) may be determined with colour changes via interactions with horseradish peroxidase and Tetramethylbenzidine) and hydrogen peroxide. Additionally, thrombin can be detected as a proxy for fibrin production in the blood using chromozym TH.Colorimetric parameters

[0151] The evaluation module 190 may be configured to determine one or more colorimetric parameters from a visual representation of the test membrane 2402. In some embodiments, the feature extractor 1006 is configured to extract one or more colorimetric parameters, from a visual representation of the test membrane 2402, as features 1008 to be provided to coagulability evaluator 1010.

[0152] A colorimetric parameter may quantify, or classify, the color of the blood sample, as diffused along a test membrane. A colorimetric parameter may define the color of the diffused blood sample. A colorimetric parameter may define the color intensity of the diffused blood sample. A colorimetric parameter may define the color of a region of the test membrane (e.g.,a colorimetric region) through which the blood sample has diffused. A colorimetric parameter may define the color intensity a region of the test membrane (e.g., a colorimetric region) through which the blood sample has diffused. In some embodiments, a colorimetric parameter is indicative of a change in a blood component composition of the blood sample for assessing thrombotic conditions.

[0153] In some embodiments, the colorimetric parameters comprise Lab coordinates, wherein L represents lightness, a represents the red / green value, and b represents the yellow / blue value. In some embodiments, the colorimetric parameters comprise Red, Green, Blue (RGB) color space parameters. In some embodiments, the colorimetric parameters comprise Hue, Saturation, Value (HSV) color space parameters.

[0154] In some embodiments, the coagulability evaluator 1010 comprises a trained machine learning model configured to classify the coagulability of a blood sample, which has been diffused along a test membrane, in accordance with a plurality of colorimetric classifications. Each of the colorimetric classifications may correspond with an indication of coagulability of the blood sample. Accordingly, the trained machine learning model may be trained to provide an indication of coagulability of the blood sample based, at least, on a colorimetric parameter of the visual representation of the test membrane to which the blood sample has been applied.Colorimetric region

[0155] The system 100 may be configured to determine at least one colorimetric parameter from a colorimetric sample region (e.g., colorimetric region) of the test membrane. In an embodiment in which the test membrane comprises a dyed region, the colorimetric region may comprise at least part of the dyed region. In some embodiments, the system is configured to determine a plurality of colorimetric parameters, from a plurality of colorimetric regions.

[0156] Figure 24 illustrates a test membrane 2402 to which a blood sample has been applied, and diffused along the test membrane, in accordance with an embodiment. In Figure 24, a first colorimetric region comprises a dyed strip 2440 of the test membrane 2402, wherein the dyed strip comprises analytic dye applied to (e.g., embedded in) the test membrane 2402. The dye strip 2440 extends the width of the test membrane 2402, perpendicular to the direction of diffusion of the blood sample.

[0157] Figure 24 further illustrates a second colorimetric region 2450, from which the system 100 may be configured to extract one or more colorimetric parameters. In embodiments, thelocation of the colorimetric region may be defined in terms of one or more of, but not limited to: a distance 2408 from the plasma diffusion front 2416; a distance 2406 from the red blood cell diffusion front 2412; a distance from the start line 2420; a vertical midline 2418 of the test membrane 2402; and a horizontal midline of the test membrane 2402.

[0158] In a preferred embodiment, a colorimetric region comprising the dyed strip 2440 is located within the plasma diffusion area, and outside of the RBC diffusion area, so that the colorimetric effect of the reaction of the analytic dye to the free haemoglobin in the plasma is readily visible with being obscured by the deep color of the RBC diffusion front.

[0159] In some embodiments, the evaluation model 190 may be configured to determine an average colorimetric parameter (e.g., an average color intensity value) across a colorimetric region, by considering different colorimetric parameter values within the colorimetric region.

[0160] In embodiments, the evaluation model 190 may be configured to determine colorimetric parameters from a plurality of colorimetric regions. The evaluation model 190 may be configured to determine representative colorimetric parameters based on the colorimetric parameters determined from a plurality of colorimetric regions by applying a function (e.g., determining an average, or weighted average) to the colorimetric parameters determined from each colorimetric region of the plurality of colorimetric regions.

[0161] In some embodiments, the evaluation module 190 comprises a trained machine learning model which is trained to, in response to the color intensity value exceeding a color intensity threshold, determine that the indication of coagulability comprises higher than normal coagulability (e.g., Mild, Moderate, Significant or Severe).

[0162] In some embodiments, the evaluation module 190 comprises a trained machine learning model which is trained to classify the color intensity of a visual representation in accordance with a color intensity classification of a set of color intensity classifications.Color normalization

[0163] The evaluation model 190 may be configured to perform color normalization techniques to the visual representation 1002, to ameliorate the effects that environmental variations may have on the accuracy of the colorimetric analysis of the visual representation of the test membrane 2402. Environmental variations may comprise variations in ambient lighting, or variations in humidity.

[0164] The evaluation model 190 may be configured to perform color normalization techniques to ameliorate the effects that variations in parameter settings may have on the accuracy of the colorimetric analysis of the visual representation of the test membrane 2402. Variations in parameter settings may result from the use of different camera equipment, changing camera settings, changing lighting settings, variations in image processing techniques or other parameters.

[0165] The evaluation model 190 may be configured to perform color normalization techniques to ameliorate the effects that variations in test membrane materials may have on the accuracy of the colorimetric analysis.

[0166] In some embodiments, the system 100 comprises a light source (not illustrated), configured to emit light onto the test membrane as the system captures an image of the test membrane 2402. The light source may be configured to emit light of in a specific wavelength, or wavelength range.

[0167] In some embodiments, the evaluation model 190 may be configured to adjust the image of the test membrane 2402, to normalise the image with respect to a reference image of a test membrane. The system 100 may apply techniques such as scaling, offsetting, color deconvolutions, color jittering, RGB (red green blue) max normalization, YCbCr normalization, and stain normalization.Experimental data

[0168] In experimental data, the inventors noted a significant impact of recalcification on coagulation dynamics in citrate-anticoagulated whole blood. Notably, non-recalcified blood exhibited a significantly greater diffusion distance (1.304 cm) compared to recalcified groups, which showed progressively shorter diffusion distances at 25 mM (0.997 cm), 50 mM (0.713 cm), 75 mM (0.531 cm), and 100 mM (0.426 cm) of CaCh, displaying an inverse relationship between RBC wicked distance and increasing CaCh concentration. Statistical analysis revealed marked differences among recalcified groups, especially between 25 mM and 50 mM (p < 0.0001), 50 mM and 75 mM (p = 0.0004), and 75 mM and 100 mM (p < 0.0488) concentrations.

[0169] Analysis of experimental results through analysis of variance (ANOVA) revealed significant variations in the mean fluorescent intensities of platelet integrin receptor glycoprotein lib (CD41)-labelled platelets across at least three distinct groups (F(3,69) = 19.23, p < 0.0001) within the LFA framework, indicating distinct platelet activation dynamics. Thisvariation was not pronounced with increasing CaCL concentrations from 0 to 100 mM (p = 0.2437), suggesting that while platelet activation as indicated by CD41 signalling may increase, it does not linearly correlate with CaCh concentration. Notably, the twofold increase in the CD41 signal, coupled with the observed definition in the resulting images, points towards an increased tendency for platelet aggregation or “clustering” at higher CaCl levels.

[0170] In parallel, fibrin formation displayed significant inter-group disparities (F(3,72) = 41.72, p < 0.0001), with the 100 mM CaCh group showcasing the highest mean fluorescence intensity (203.6 Arbitrary Units, AU), starkly contrasting with the 0 mM group (p < 0.0001). This suggests a pronounced fibrin network formation under high CaCl conditions. Furthermore, the emergence of heterogeneously thick fibrin-rich areas post- 100 mM CaCh treatment manifested in more pronounced deviations in the analysis of regions of interest (ROI), with means of 109.5 ± 47.65 for 100 mM compared to 61.02 ± 7.1 for 0 mM. These findings underscore the pivotal role of fibrin in modulating droplet coagulation, reducing diffusion distance, and enhancing overall coagulability.

[0171] Moreover, SEM imaging of the RAT NC membrane provided visual confirmation of these biochemical dynamics. The compaction effect within blood clots is particularly noticeable when comparing whole blood with a recalcification dose of 0 mM and 100 mM CaCh against a control test cassette. The microscopic analysis revealed a significant reduction in the NC membrane pore size with increasing CaCh levels, visually substantiating the biochemical data on the impact of calcium -induced coagulation through platelet-fibrin interactions.

[0172] These observations collectively highlight the nuanced and complex nature of blood coagulation as captured by the innovative repurposing of RAT paper microfluidics. The technology not only enables the distinct separation of blood components under varying coagulation states but also provides a detailed quantitative and qualitative insight into the coagulation process, underscoring the critical interplay between platelets and fibrin in coagulation dynamics.

[0173] The efficacy of an embodiment of the system 100 comprising a random forest classifier was evidenced by its predictive performance metrics. During validation, the random forest model exhibited outstanding discrimination capabilities, with C-index values for intraclass analysis surpassing 0.95 (as illustrated in Figure 16), showcasing the model's precision in classifying coagulation statuses. In Figure 16, Band 2 (Area Under the Curve (AUC) = 1.00) isreferenced by reference numeral 1604, Band 3 (AUC = 0.95) is referenced by reference numeral 1606, Band 4 (AUC = 0.95) is referenced by reference numeral 1608, and Band 5 (AUC = 0.97) is referenced by reference numeral 1610. Multi class ROC analysis further demonstrated the model's effectiveness across different test sizes, achieving C-index values of 0.83, 0.93, and 0.96 for test sizes of 0.7, 0.5, and 0.2 respectively (as illustrated in Figure 17). In Figure 17, Average (AUC = 0.96)(Test Size = 0.2) is referenced by reference numeral 1702, Average (AUC = 0.93)(Test Size = 0.5) is referenced by reference numeral 1704, and Average (AUC = 0.83)(Test Size = 0.7) is referenced by reference numeral 1706. Multiclass ROC analysis demonstrated the model's effectiveness across different test sizes, achieving Precision-Recall values escalating from 0.65 to 0.90 as test sizes decreased (as illustrated in Figure 18). In Figure 17, Average (AUC = 0.90)(Test Size = 0.2) is referenced by reference numeral 1802, Average (AUC = 0.82)(Test Size = 0.5) is referenced by reference numeral 1804, and Average (AUC = 0.65)(Test Size = 0.7) is referenced by reference numeral 1806. These results underscore the model's robust capacity for accurate positive predictions and comprehensive capture of actual positive instances.

[0174] The confusion matrix (as illustrated in Figure 19), reaffirmed the model's superior class discrimination accuracy, with an average success rate of 83.3% in correctly identifying true positives and true negatives. This robust classification underscores the method's reliability and efficiency in determining coagulation status, setting a promising benchmark for potential clinical application.Antithrombotic medication

[0175] The optimization of antithrombotic medication dosages is a pivotal element in the management of thrombotic risks and bleeding during medical interventions. Achieving an optimal balance requires precise adjustment of anticoagulant levels.

[0176] The application of paper-based LFAs for refining antithrombotic medication dosing strategies was investigated. Specifically, the effects of administering a fixed 20U dose of enoxaparin on whole blood samples, which were preconditioned with varying concentrations of CaCT, to simulate different coagulation states was investigated. These experiments aimed to compare the coagulation dynamics in the presence and absence of enoxaparin.

[0177] Figure 20 illustrates the results of an analysis of RBC diffusion distance (wicked diffusion distance) comparing increasing concentrations of CaCL with 20 U of enoxaparin, as abar chart represented as mean ± SD, in accordance with an embodiment. Figure 21 illustrates the results of an analysis of RBC diffusion distance (e.g., wicked diffusion distance) comparing increasing doses of enoxaparin, as a bar chart represented as mean ± SD, in accordance with an embodiment. For both Figure 20 and 21, n = 4, ns = not significant; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001, assessed by one-way ANOVA.

[0178] Experimental results revealed that administering 20U of enoxaparin significantly altered the coagulation profile, as evidenced by increased wicked diffusion distances in samples treated with 25 mM and 50 mM CaCh, compared to controls without enoxaparin (as illustrated in Figure 20). Specifically, enoxaparin augmented the wicked diffusion distances from 0.981 cm to 1.14 cm for the 25 mM CaCh group, and from 0.715 cm to 0.877 cm for the 50 mM group (Figure 20), indicating a notable anticoagulant effect at these concentrations (p < 0.05). Conversely, at higher CaCh concentrations (75 mM and 100 mM), the addition of 20U enoxaparin did not significantly extend the wicked diffusion distance, suggesting a plateau in the efficacy of this enoxaparin dosage under conditions of more pronounced coagulation.

[0179] Further experiments were conducted to evaluate the dose-response relationship of enoxaparin in modulating coagulation, particularly at the 50 mM CaCh concentration, identified as a critical threshold for assessing the sensitivity and effectiveness of enoxaparin in the paperbased LFA system (Figure 21). Incremental enoxaparin dosages ranging from 0 U to 100 U were administered, demonstrating a positive correlation between enoxaparin dosage and wicked diffusion distance. Specifically, wicked diffusion distances progressively increased with higher enoxaparin doses, recording measurements of 0.897 cm, 0.962 cm, 1.023 cm, and 1.125 cm for 20 U, 50 U, 75 U, and 100 U of enoxaparin, respectively (Figure 21). Statistical analysis confirmed the significance of these findings, with all enoxaparin-treated groups showing substantial increases in wicked distances compared to the control, achieving statistical significance at p < 0.05 for the 20 U group, and p < 0.001 and p < 0.0001 for the 50 U, 75 U, and 100 U groups, respectively (Figure 21).

[0180] These results illuminate the potential of utilizing paper-based LFAs as an innovative and practical tool for clinicians to fine-tune antithrombotic dosages. By providing a rapid, quantitative method to assess the anticoagulant effect of varying enoxaparin doses in real-time, this approach promises to contribute significantly to personalized patient care, optimizing therapeutic outcomes while minimizing bleeding risks.

[0181] Advantageously, embodiments of methods described herein demonstrate the capability to discern distinct levels of coagulation using minimal blood volumes akin to those obtained from a finger-prick. Furthermore, the adaptation of LFAs for classifying blood coagulability not only showcases the versatility of LFAs beyond their original purpose but also highlights the potential for rapid, on-site blood coagulability assessments.

[0182] To avoid obscuring the inventive subject matter with unnecessary detail, various functional components (e.g., modules, devices, databases, etc.) that are not germane to conveying an understanding of the inventive subject matter have been omitted from Figure 9. However, a skilled artisan will readily recognize that various additional functional components may be supported by the system 900 to facilitate additional functionality that is not specifically described herein. Furthermore, the various functional components depicted in FIG. 4 may reside on a single computing device or may be distributed across several computing devices in various arrangements such as those used in cloud-based architectures.

[0183] It will be appreciated by persons skilled in the art that numerous variations and / or modifications may be made to the above-described embodiments, without departing from the broad general scope of the present disclosure. Furthermore, it will be appreciated by persons skilled in the art that embodiments disclosed herein can be combined with one or more other embodiment disclosed herein, without departing from the broad general scope of the present disclosure. The present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive.

[0184] It will be appreciated by persons skilled in the art that any suitable distribution of functionality between different functional units may be used without detracting from the invention. For example, functionality illustrated to be performed by separate computing devices may be performed by the same computing device. Likewise, functionality illustrated to be performed by a single computing device may be distributed amongst several computing devices. Hence, references to specific functional units are only to be seen as references to suitable means for providing the described functionality, rather than indicative of a strict logical or physical structure or organization. For example, operations described as being performed by the pre-processing module 150 may be performed in part or in full by the evaluation module 190 (or a module thereof). Similarly, operations described as being performed by the evaluation module 190 (or a module thereof) may be performed in part or in full by the pre-processing module 150.

[0185] It will be appreciated by persons skilled in the art that, for processes and methods disclosed herein, the operations performed in the processes and methods may be implemented in differing order. Furthermore, the outlined steps and operations are only provided as examples, and some of the steps and operations can be optional, combined into fewer steps and operations, or expanded into additional steps and operations without detracting from the essence of the disclosed embodiments.

[0186] References herein to software or executable instructions are to be understood as referring to executable instructions stored in volatile or non-volatile memory. The memory can include any data storage device that can store data which can thereafter be read by a processor. Examples of memory include read-only memory (ROM), random-access memory (RAM), magnetic tape, optical data storage device, flash storage devices, or any other suitable storage devices.

[0187] Throughout this specification the word ‘comprise’, or variations such as ‘comprises’ or ‘comprising’, will be understood to imply the inclusion of a stated element, integer or step, or group of elements, integers or steps, but not the exclusion of any other element, integer or step, or group of elements, integers or steps.

[0188] As used herein, any reference to “one embodiment” or “an embodiment” means that a particular element, feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment. The appearances of the phrase “in one embodiment” in various places in the specification are not necessarily all referring to the same embodiment. Similarly, use of “a” or “an” preceding an element or component is done merely for convenience. This description should be understood to mean that one or more of the element or component is present unless it is obvious that it is meant otherwise.

[0189] Unless expressly stated to the contrary, “or” refers to an inclusive or and not to an exclusive or. For example, a condition A or B is satisfied by any one of the following: A is true (or present) and B is false (or not present), A is false (or not present) and B is true (or present), and both A and B are true (or present).

[0190] Moreover, in the following claims, the terms “first,” “second,” and “third,” etc. are used merely as labels, and are not intended to impose numerical requirements on their objects.References[1] L. Labajos-Broncano, M. L. Gonzalez -Martian, and J. M. Bruque, “Washburn’s Equation Facing Galileo’s Transformation: Some Remarks,” J. Colloid Interface Sci., vol. 253, no. 2, pp. 472-474, Sep. 2002, doi: 10.1006 / jcis.2002.8521.[2] S. Deutsch, “A preliminary study of the fluid mechanics of liquid penetrant testing,” J. Res. Natl. Bur. Stand., vol. 84, no. 4, p. 287, Jul. 1979, doi: 10.6028 / jres.084.012.[3] A. Atangana, “Principle of Groundwater Flow,” in Fractional Operators with Constant and Variable Order with Application to Geo-Hydrology, Elsevier, 2018, pp. 15-47. doi: 10.1016 / B978-0-12-809670-3.00002-3.

Claims

CLAIMS:

1. A computer-implemented method for determining a coagulability of a sample of blood, the method comprising: obtaining a visual representation of a test membrane to which a blood sample has been applied and has diffused along the test membrane via capillary action; applying a trained machine learning model to the visual representation of the test membrane, the trained machine learning model trained on a plurality of training visual representations of training membranes to determine an indication of coagulability of a training blood sample applied to a respective training membrane; generating, by the trained machine learning model, an indication of coagulability of the blood sample; and issuing a notification comprising the indication of coagulability of the blood sample.

2. The method of claim 1, wherein the trained machine learning comprises a deep learning convolutional neural network classifier, configured to extract one or more features from the visual representation of the test membrane.

3. The method of claim 1, further comprising: extracting one or more features from the visual representation of the test membrane; and providing the one or more features to the trained machine learning model.

4. The method of any one of claims 2 or 3, wherein the one or more features comprise one or more of: the red blood cell diffusion front; a red blood cell diffusion distance; a red blood cell diffusion area; a red blood cell diffusion rate; the plasma diffusion front; a plasma diffusion distance; a plasma diffusion area; a plasma diffusion rate;an area ratio of the plasma diffusion area to the red blood cell diffusion area; an distance ratio of the plasma diffusion distance to the red blood cell diffusion distance; a colorimetric parameter indicative of a change in a blood component composition of the blood sample; a colorimetric parameter indicative of a change in a pathology of the blood sample; and a colorimetric indicator of coagulability of the blood sample.

5. The method of any one of claims 2 to 4, wherein extracting one or more features from the visual representation of the test membrane comprises at least one of: calculating the red blood cell diffusion area based on the red blood cell diffusion front; calculating the red blood cell diffusion rate based on a first position of the red blood cell diffusion front at a first time, and a second position of the red blood cell diffusion front at a second time; calculating the plasma diffusion area based on the plasma diffusion front; calculating the plasma diffusion rate based on a first position of the plasma diffusion front at a first time, and a second position of the plasma diffusion front at a second time; calculating a ratio of the plasma diffusion rate to the red blood cell diffusion rate; calculating the red blood cell diffusion distance based on the red blood cell diffusion front; calculating the plasma diffusion distance based on the plasma diffusion front; and calculating the distance ratio of the plasma diffusion distance to the red blood cell diffusion distance.

6. The method of any of claims 4 to 5, wherein the trained machine learning model is configured to, perform one or more of: in response to the area ratio exceeding an area threshold, determine that the indication of coagulability comprises high coagulability;in response to the area ratio not exceeding the area threshold, determine that the indication of coagulability comprises normal coagulability; and in response to the distance ratio exceeding a distance threshold, determine that the indication of coagulability comprises high coagulability.

7. The method of any one of claims 4 to 6, wherein extracting the colorimetric parameter comprises determining a color intensity of a colorimetric region within the visual representation.

8. The method of claim 7, wherein the colorimetric region comprises a region of the test membrane to which analytic dye has been applied, and through which the blood sample has diffused.

9. The method of claim 8, wherein the analytic dye in the dyed region is configured to effect a color change of the dyed region in response to the presence of one or more of: free haemoglobin; platelet biomarkers; and fibrin biomarkers, in the blood sample.

10. The method of any of claims 4 to 9, wherein the trained machine learning model is configured to, in response to the colorimetric parameter indicating color intensity exceeding a color intensity threshold, determine the indication of coagulability comprises high coagulability.

11. The method of any one of claims 1 to 10, further comprising: mapping the indication of coagulability to a medication recommendation; and and outputting the medication recommendation to a user.

12. The method of any of claims 1 to 11, further comprising preprocessing the visual representation, wherein preprocessing the visual representation comprises one or more of: normalizing the visual representation for color intensity; normalizing the visual representation for brightness; normalizing the visual representation for contrast; and performing region of interest alignment.

13. The method of any of claims 1 to 12, wherein the trained machine learning model comprises a classifier, configured to classify the visual representation of the test membrane into one of a plurality of blood coagulability classifications.

14. The method of claim 13, wherein the classifier has been trained, using weak supervision, wherein the test visual representations comprise a plurality of videos, and wherein labels are provided at the video level, and not for individual frames of a video of the plurality of videos.

15. The method of any of claims 1 to 14, wherein the trained machine learning model comprises a cross entropy loss function.

16. The method of any of claims 1 to 15, wherein the trained machine learning model comprises a center loss algorithm, the center loss algorithm configured to apply a distance query to features extracted from the image of the test membrane to determine a second indication of coagulability.

17. The method of claim 16, wherein the trained machine learning model is configured to adjust the indication of coagulability based on the second indication of coagulability.

18. The method of any one of claims 1 to 17, wherein the test membrane is a component of a lateral flow assay.

19. A non-transitory storage medium storing machine-readable instructions which, when executed by one or more processors, individually or in combination, cause the one or more processors to perform the method of any one of claims 1 to 18.

20. A system comprising: one or more processors; andmemory comprising computer executable instructions, which when executed by the one or more processors, individually or in combination, cause the system to perform the method of any one of claims 1 to 18.

Images