WO2025173009A1

WO2025173009A1 - Device, system and method for ai-based quantitative colorimetric lateral flow analysis

Info

Publication number: WO2025173009A1
Application number: PCT/IL2025/050158
Authority: WO
Inventors: Rafi BENTAL; Amit ASSA; Adnan Agbaria
Original assignee: Quadpoint Labs Ltd
Current assignee: Quadpoint Labs Ltd
Priority date: 2024-02-14
Filing date: 2025-02-13
Publication date: 2025-08-21
Anticipated expiration: 2026-08-14

Abstract

Disclosed are quantitative lateral flow immunoassay (LFIA), specifically to home-based point-of-care (POC) kits and method for performing quantitative LFIA analysis.

Description

DEVICE, SYSTEM AND METHOD FOR AI-BASED QUANTITATIVE COLORIMETRIC LATERAL FLOW ANALYSIS

TECHNICAL FIELD

This disclosure generally relates to quantitative lateral flow immunoassay (LFIA), specifically to home-based point-of-care (POC) kits and method for LFIA.

BACKGROUND

Recently, the demand for quantitative analysis of lateral flow immunoassays is increasing in various fields.

Home-based LFIA is qualitative and relies on visual detection, by human eyes, of coloration of a test line (the T line) and a control line (the C line).

However, when quantitative results are required, it is practically impossible to do it using the human eye's subjective detection. As a result, a large variety of LFIA reader devices with dedicated software have been developed.

Today’s quantitative LFIA technologies require dedicated readers (referred to herein as “LFIA readers”) which provide full control over parameters, such as environment illumination level, shadowing, orientation, and camera type. LFIA readers include a mechanical part that can store and shield of the LFIA strip, a broadband light source that illuminates the test line and control line of the assay, a sensor that acquires images of the region of interest (ROI) including the test line, and a processor with image-processing software that calculates the presence or concentration of a target analyte from an image. Such devices are expensive and usually require special training for operation.

Therefore, there remains a need for devices, system and methods that enable a cost-efficient home-based POC LFIA. SUMMARY

This disclosure relates to quantitative lateral flow immunoassay (LFIA), specifically to home-based point-of-care (POC) kits and method for quantitative LFIA.

Advantageously, by sequential implementation of Al models on an image of an LFIA strip exposed to a biological sample, the image is initially normalized for environmental differences (such as light exposure etc.) and then a second Al model is applied, the second model trained on various control test mark intensities measured using an LFIA reader, enabling quantitative analysis of the test mark obtained. This advantageously obviates the need for the expensive LFIA reader device for a quantitative analysis of the sample, which in turn enables a point of care or even home testing.

According to some embodiments, the LFIA strip analyzed preferably and advantageously includes more than one test area, each test area including a different concentration of the agent that produces a colorimetric test mark (T) in response to the presence of an analyte in the biological sample. This serves to ensure that the test is performed in a test-window in which the correlation between color intensity and analyte concentration is essentially linear.

In addition, the color intensity of the colorimetric marks of the test area may change over time, measuring the same strip at different times can yield different results, thus making the test results unreliable. Accordingly, it is important to ensure that the image capturing is done within a valid time window and at what specific time within that window. Therefore, the hereindisclosed LFIA strip may further include one or more “elapsed time” control areas (ETC), optionally instead of or in addition to the control-area used for confirmation of correct lateral flow of the sample. The ETC is configured to change (e.g. in color and/or color intensity) in response to time to thereby indicate the biological time of the test, i.e. time elapsed between the sample reaching the ETC and capturing of the image. According to some embodiments, the ETC may include multiple ETCs (also referred to as an “ETC matrix”) each ETC including a different concentration of the agent capturing the control substance and thus changing its color and/or color intensity at different concentrations of a control substance i.e., at different time after reaching the ETC area, e.g. at intervals of between from 30 seconds to 5 minutes. It is further understood that an Al model may be applied on the image, which model is configured to calibrate the color intensity of the test areas (T) over time.

Advantageously, the Al model analyzes the number of ETCs activated, thereby providing a continuous time result, e.g. with a 5-second resolution. Advantageously, the determined test concentration may then be adjusted, based on the exact time elapsed, thereby reducing or even eliminating variations to the test results that are due to differences in image capture times. It is further understood, that if the elapsed time exceeds a predetermined threshold the test may be deemed invalid. Similarly, if insufficient time has passed, an indication may be sent to the user to recapture the image at a later time point.

It is understood that the herein disclosed methods and LFIA strips may advantageously be tailored for use in diagnosis/monitoring of various types of medical conditions and illnesses such as but not limited to: early-stage cancer screening, chronic disease monitoring, infection and parasite detections and the like. Similarly, while it initially is directed to detection/monitoring of human diseases it can likewise be applied in veterinarian and agricultural applications. Moreover, due to the relatively low costs, the LFIA test and associated method, personalization of the test can easily be achieved, taking into consideration factors such as age, sex, medical background etc. in the choice of antibodies and/or test result interpretation.

Anon-limiting example specific example of the hereindisclosed LFIA strip is an LFIA strip for detection of colorectal cancer including a first plurality of test areas (typically 3-6) including different concentrations of an antibody against Hemoglobin, a second plurality of test areas (typically 3-6) including different concentrations of an antibody against Calprotectin and/or a third plurality of test areas (typically 3-6) including different concentrations of an antibody against Serpin.

Advantageously, the herein disclosed LFIA strip and associated method enables home detection/monitoring of medical conditions (such as but not limited to colorectal cancer), and thus increase screening compliance (e.g., as compared to the often-feared endoscopic procedure) and thus increase the rate of early versus late-stage detection (which in the case of colorectal cancer may change 10-year survival rate to 90% as compared to 10%, respectively). Moreover, as compared to LFIA tests conducted in dedicated laboratories, the herein disclosed LFIA strip and associated method enable to conduct the sample in a home-setting, thus avoiding issues such as privacy, discomfort and shame often involved with handing over samples, in particular stool samples. Moreover, this LFIA strip and associated method may advantageously pre-screen subjects (e.g. for colorectal cancer), such that only subjects for which a risk has been identified are sent to further testing (such as endoscopic testing), thus significantly reducing screening costs.

According to some embodiments, the LFIA strip may include a QR-code (or similar element) configured to connect the LFIA strip with a dedicated mobile App, through which the testing will be guided and test results determined.

According to some embodiments, there is provided a computer implemented method for Al-based quantitative colorimetric lateral flow immunoassay analysis (LFIA), the method including: a) receiving, from a user, an image captured with a handheld digital imaging device, the image comprising a picture of a LFIA strip having been exposed to a biological sample, wherein the LFIA strip includes: a test area containing an agent configured to produce a colorimetric test mark (T) in response to the presence of an analyte in the biological sample, and a reference area comprising a colorimetric reference mark (R) having a predefined color and color intensity, the colorimetric reference mark (R) formed irrespective of exposing the LFIA strip to the biological sample; b) removing environmental conditions differences from the captured image (or parts thereof that contain the picture of the LFIA strip), by applying a first Al model on the captured image or parts thereof, the first Al model trained on a plurality of images of the colorimetric reference mark (R) captured under different imaging conditions and normalized to a color intensity of the colorimetric reference mark (R) measured using an LFIA reader, thereby obtaining a normalized image; and c) calculate a concentration of the analyte in the test sample by applying a second Al model on the color intensity of the colorimetric test mark (T) in the normalized image, the second Al model trained on a plurality of various control test mark (CT) intensities measured using an LFIA reader, each of the plurality of control test mark (CT) intensities labeled with its associated concentration of the analyte; and d) transmit the calculated concentration to the user and/or caregiver.

According to some embodiments, the LFIA strip further includes a control area configured to produce at least two colorimetric control marks (C) in response to lateral flow of the test sample therethrough, irrespective of presence or absence of an analyte in the test sample, wherein each of the at least two colorimetric control marks (C) has a different response time indicative of the time that has elapsed between loading of the biological sample and image capturing. According to some embodiments, the method further includes adjusting the calculated analyte concentration, based on time that has elapsed between loading of the biological sample and image capturing.

According to some embodiments, the imaging conditions comprise camera quality, camera configuration/setting, surrounding light condition, shadowing and/or any combination thereof.

According to some embodiments, the method further comprises pre-processing the image to convert all colorimetric marks to a single-color monochromatic scale.

According to some embodiments, the method further comprises applying an object detection algorithm on the image to identify the LFIA strip or parts thereof.

According to some embodiments, the first and second Al models are convoluted neural network (CNN) models. According to some embodiments, the CNN model is a multi-classification model.

According to some embodiments, the multi-classification model is configured to predict a probability that the colorimetric test mark (T) matches with a color intensity of each of the plurality of control test marks (TC). According to some embodiments, the LFIA strip comprises at least two test areas, each test area comprising a different concentration of the agent, such that an essentially linear correlation between color intensity and analyte concentration is obtained at different analyte concentration, for each of the at least two test areas.

According to some embodiments, upon exposure to a given analyte concentration, at least a portion of the at least two test areas provide a different color intensity, thereby increasing linear data point range and accuracy.

According to some embodiments, the multi-classification model is further configured to calculate the concentrations of the analyte in the biological sample, based on matching of the color intensities obtained in each of the at least two test areas to a most similar of the plurality of control test marks (TC).

According to some embodiments, there is provided a lateral flow immunoassay analysis (LFIA) strip comprising: a membrane configured for lateral flow of a test sample, the membrane comprising: a control area configured to produce a colorimetric control mark (C) in response to the lateral flow of the test sample therethrough irrespective of presence or absence of an analyte in the test sample; a test area comprising an agent configured to produce a colorimetric test mark (T) in response to presence of the analyte in the test sample; a reference area comprising: an isolating layer; and a secondary membrane comprising a colorimetric reference mark (R) having a predefined color intensity, the colorimetric reference mark (R) formed before exposing the LFIA strip to the biological sample, wherein the secondary membrane is shielded off from contact with the test sample by the isolating layer, such that the color intensity of the colorimetric reference mark (R) is independent of lateral flow of the test sample through the LFIA strip and , wherein the reference area is positioned proximally to the test area.

According to some embodiments, the LFIA strip comprises at least two test areas (T), each test area comprising a different concentration of the agent, such that a minimal concentration of the analyte sufficient to produce the colorimetric test mark (T) differs in each area.

According to some embodiments, the LFIA strip comprises at least two colorimetric reference marks (R), each having a different preformed and predefined color intensity.

According to some embodiments, the membrane and the secondary membrane are made from the same material. According to some embodiments, the membrane and the secondary membrane are nitrocellulose membranes.

According to some embodiments, the LFIA strip further comprises a sample pad configured to receive the test sample.

According to some embodiments, the LFIA strip further comprises a conjugation pad configured to release a conjugate that binds the analyte if present.

According to some embodiments, the LFIA strip further comprises an absorption pad configured to wick the prevent backflow of the test sample.

According to some embodiments, the LFIA strip further comprises an adhesive pad configured to sequentially adhere the sample pad, the conjugation pad, the membrane and the absorption pad.

According to some embodiments, wherein the control area comprises at least two colorimetric control marks (C), wherein each of the at least two colorimetric control marks (C) has a different response time which is indicative of the time that has elapsed between loading of the biological sample and image capturing.

According to some embodiments, there is provided a method for quantitative analysis of lateral flow immunoassay analysis (LFIA) results, the method comprising: a) loading a biological sample on an LFIA strip including: i. at least two test areas, each test area comprising a different concentration of an agent configured to produce a colorimetric test mark (T) in response to the presence of an analyte in the biological sample, such that an essentially linear correlation between color intensity and analyte concentration is obtained; ii. a reference area comprising at least two colorimetric reference marks (R), each reference mark having a different predefined color and color intensity, formed before exposing the LFIA strip to the biological sample; and iii. a control area configured to produce a colorimetric control mark (C) in response to the lateral flow of the test sample therethrough irrespective of presence or absence of an analyte in the test sample; b) capturing an image of the LFIA strip; c) applying an object detection algorithm on the image to identify the LFIA strip or parts thereof; d) removing image environmental conditions differences from the captured image or parts thereof that contain the picture of the LFIA strip, by applying a first CNN model on the captured image or parts thereof, the first CNN model trained on a plurality of images of colorimetric reference mark (R) captured under different imaging conditions and normalized to a color intensity of the colorimetric reference mark (R) measured using an LFIA reader, thereby obtaining a normalized image; and e) calculating a concentration of the analyte in the biological sample by applying a second CNN model on the color intensity of the colorimetric test mark (T) in the normalized image, the second Al model trained on a plurality of various control test mark (CT) intensities measured using an LFIA reader, each of the plurality of control test mark (CT) intensities labeled with its associated concentration of the analyte; and f) transmitting the calculated concentration to the user and/or caregiver.

According to some embodiments, the control area includes at least two colorimetric control marks (C) each having a different response time indicative of the time that has elapsed between loading of the biological sample and image capturing. According to some embodiments, the method further includes adjusting the calculated analyte concentration, based on the time that has elapsed between loading of the biological sample and image capturing. According to some embodiments, if the time elapsed exceeds a predetermined threshold value, the calculated analyte concentration is deemed invalid.

Certain embodiments of the present disclosure may include some, all, or none of the above advantages. One or more technical advantages may be readily apparent to those skilled in the art from the figures, descriptions and claims included herein. Moreover, while specific advantages have been enumerated above, various embodiments may include all, some or none of the enumerated advantages.

In addition to the exemplary aspects and embodiments described above, further aspects and embodiments will become apparent by reference to the figures and by study of the following detailed descriptions.

BRIEF DESCRIPTION OF THE FIGURES

The invention will now be described in relation to certain examples and embodiments with reference to the following illustrative figures so that it may be more fully understood.

FIG. 1 shows an LFIA strip with embedded reference areas each having a different color intensity, according to some embodiments; FIG. 2A illustratively depicts an exemplary LFIA strip including a single test area and a single reference area, according to some embodiments;

FIG. 2B illustratively depicts an exemplary LFIA strip including a single test area and a plurality of reference area (here six), according to some embodiments;

FIG. 3 is a schematic graph illustrating the color intensity as a function of analyte concentration of a LFIA, according to some embodiments;

FIG. 4A illustratively depicts an exemplary LFIA strip including a plurality of test areas (here three) and a single reference area, according to some embodiments;

FIG. 4B illustratively depicts an exemplary LFIA strip including a plurality of test areas (here three) and a plurality of reference areas (here six), according to some embodiments;

FIG. 5 is a schematic graph illustrating the color intensity as a function of concentration for multiple T-areas (here three) with different cut-off points, according to some embodiments;

FIG. 6 illustratively depicts an exemplary LFIA strip including a plurality of test areas (here three) for more than one analyte (here 2 analytes) and a plurality of reference areas (here six), according to some embodiments;

FIG. 7 is an exemplary flow chart of the herein disclosed method, according to some embodiments;

FIG. 8 illustratively depicts an exemplary LFIA strip including a plurality of test areas (here three) and a plurality of reference areas (here six) and a plurality of lapsed-time control areas (here six), according to some embodiments, according to some embodiments; and

FIG. 9 illustratively depicts, of an LFIA strip as disclosed herein and associated method, according to some embodiments.

DETAILED DESCRIPTION

In the following description, various aspects of the disclosure will be described. For the purpose of explanation, specific configurations and details are set forth in order to provide a thorough understanding of the different aspects of the disclosure. However, it will also be apparent to one skilled in the art that the disclosure may be practiced without specific details being presented herein. Furthermore, well-known features may be omitted or simplified in order not to obscure the disclosure.

For convenience, certain terms used in the specification, examples, and appended claims are collected here. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skills in the art to which this invention pertains.

Disclosed herein are new devices and methods for quantitative LFIA using standard cameras, such as a camera of a smartphone. No special reader is required.

According to some embodiments, there is provided a computer implemented method for Al-based quantitative colorimetric lateral flow immunoassay analysis (LFIA), the method including: receiving an image captured with a handheld digital imaging device, the image capturing an LFIA strip that has been exposed to a biological sample, wherein the LFIA strip includes: a) a test area comprising an agent configured to produce a colorimetric test mark (T) in response to the presence of an analyte in the biological sample, and a reference area comprising at least two colorimetric reference marks (R) each having a predefined color and/or color intensity formed before or irrespective of the exposing of the LFIA strip to the biological sample; removing environmental conditions differences from the captured image or parts thereof that contain the picture of the LFIA strip, by applying a first Al model on the captured image or parts thereof, the first Al model trained on a plurality of images of the colorimetric reference mark (R) captured under different imaging conditions and normalized to a color intensity of the colorimetric reference mark (R) measured using an LFIA reader, thereby obtaining a normalized image; and calculate a concentration of the analyte in the biological sample by applying a second Al model on the color and/or color intensity of the colorimetric test mark (T) in the normalized image, the second Al model trained on a plurality of different control test mark (CT) intensities measured using an LFIA reader, each of the plurality of control test mark (CT) intensities labeled with its associated concentration of the analyte; and transmit the calculated concentration to the user and/or caregiver.

As used herein, the term “lateral flow immunoassay analysis (LFIA)” refers to a diagnostic technique typically used to detect the presence (or absence) of a target substance, such as an antigen, antibody, or other biomolecules, in a sample. It relies on capillary action to move a liquid sample along a test strip, where specific interactions between the target molecule and labeled antibodies produce a visible result. When combined with dedicated reader devices, LFIAs can provide semi-quantitative or fully quantitative results by measuring the intensity of the test area. Examples of readers include: RapidScan ST5-W Lateral Flow Assay Reader, OPERON Rapid Strip Reader, Hamamatsu Lateral Flow Readers, RapidScan ST5-W and more.

As used herein, the term “handheld” with regards to the imaging device refers to refers to imaging devices that can be held in the hand of a user, such as digital cameras, mobile phone cameras, table cameras etc. It is understood that the device may not necessarily be held by a user during the image capturing, it may for example be held by the user via a selfie-stick or be positioned on a tripod or the like, yet this is still referred to as a handheld device.

As used herein, the term “test area” refers to an area of an LFIA strip which contains the agent (e.g. an antibody) configured to detect an analyte (e.g. an antigen) in a biological sample. According to some embodiments, the test area may be rectangular, in which case it is often referred to as a test-line or T-line. In other embodiments, the test area may by circular, in which case it is often referred to as a test-dot or T-dot. Other shapes can also be envisaged and as such are encompassed by this disclosure. As used herein, the term “colorimetric test mark” refers to the color visible as a result to the presence of the analyte in the biological sample. That is, the test area produces a colorimetric test mark/result as a result of the interaction between the agent and the analyte.

As used herein, the term “analyte” refers to a specific substance or component in a sample that is being measured or detected in a medical assay. In LFIA, the analyte binds to specific antibodies or molecules on the test strip, leading to a visible signal (e.g., a test mark appearing).

As used herein, the term “reference area” refers to an area of an LFIA strip which provides signals/coloring in a manner irrespective of the presence/absence of an analyte in the biological sample.

As used herein, the term “colorimetric reference marks” refers to refers to the color visible in the test area irrespective of the presence/absence of an analyte in the biological sample. According to some embodiments, the colorimetric reference marks is formed before exposing the LFIA strip to the biological sample. According to some embodiments, the colorimetric reference marks is formed in response to the LFIA strip being exposed to the biological sample, but in a manner irrespective of the presence/absence of the analyte. For example, the colorimetric reference marks may be formed in response to the sample buffer reaching the reference area.

As used herein, the terms “environmental conditions differences” and “imaging conditions” may be used interchangeably and refer to factors that may influence the color and/or color intensity visualized in the image. Non-limiting examples of imaging conditions include camera quality, camera configuration/setting, surrounding light condition, shadowing and/or any combination thereof. Each possibility is a separate embodiment.

According to some embodiments, the method further includes pre-processing the image to convert all colorimetric marks to a single-color monochromatic scale.

According to some embodiments, the method further includes applying an object detection algorithm on the image to identify the LFIA strip or parts thereof in the image. According to some embodiments, the object detection algorithm is a one-stage detector. Suitable one-stage detectors include: YOLO (Y0L0v3, Y0L0v4, Y0L0v5, Y0L0v6, Y0L0v7, Y0L0v8), Single Shot MultiBox Detector (SSD) and RetinaNet. Each possibility is a separate embodiment. According to some embodiments, the object detection algorithms is a two-stage detector. Suitable two-stage detectors include: Region-based Convolutional Neural Networks (R-CNN), Faster R-CNN, Mask R-CNN, and Region-based Fully Convolutional Networks. Each possibility is a separate embodiment.

According to some embodiments, the first and/or the second Al models may be selected from Mask-RCNN, Unet, DeepLav3+, K-Means and the like. Each possibility is a separate embodiment. According to some embodiments, the Al model is a CNN model. According to some embodiments, the CNN model is a multi-classification model. According to some embodiments, the multiclassification model is configured to predict a probability that the colorimetric test mark (T) matches with a color intensity of each of the plurality of control test marks (TC). Non-limiting examples of suitable multi-classification models include small CNN or a fully connected neural network (FCNN) trained to learn the relationship between color intensity and the probability of a match. Additionally or alternatively, Neural Networks or Support Vector Machines (SVM) may be utilized.

According to some embodiments, the LFIA strip includes at least two test areas, each test area including a different concentration of the agent. According to some embodiments, the LFIA strip includes more than two test areas, such as three, four, six, eight or ten test areas. Each possibility is a separate embodiment. According to some embodiments, the LFIA strip includes more than 10 test areas. According to some embodiments an essentially linear correlation between color intensity and analyte concentration is obtained at different analyte concentration, for at least a portion of the at least two test areas.

Additionally or alternatively, the at least two test areas may include different agents, i.e. different antibodies, each antibody detecting a different protein. According to some embodiments, the different antibodies may detect different proteins involved in a same medical condition, e.g. different proteins associated with colon cancer. According to some embodiments, the different antibodies may detect different proteins involved in different medical conditions.

According to some embodiments, upon exposure to a given analyte concentration, at least a portion of the at least two test areas provide a different color and/or color intensity, thereby increasing linear data point range and accuracy. That is, by including a plurality of test areas, each test area including a different concentration of the agent (and thus different detection cut-off points), the linear range of the strip is advantageously expanded and may match or even exceed the accuracy and range of LFIA readers.

According to some embodiments, the LFAI strip further includes a control area. As used herein, the term control area refers to an area that produces a colorimetric control marks (C) in response to the lateral flow of the test sample through the LFIA strip, irrespective of presence or absence of an analyte in the test sample. It works based on the capillary flow of liquid, antibodyantigen interactions, and colored nanoparticle detection (e.g., gold nanoparticles). That is, once the sample wicks through the LFIA via capillary action, the sample dissolves dried labeled nanobodies, typically gold nanoparticles or latex beads from the conjugate Regardless of whether the analyte is present, excess labeled antibodies (which are not bound at the test area) flow further down the strip. These unbound antibodies are captured at the control area (in the form of lines or dots or other shape) by an anti-species antibody (e.g., an anti-IgG antibody that recognizes the labeled antibody - also referred to as “secondary antibody”), whereby the control mark is formed. It is understood that as time passes by, more and more antibodies reach the control area. Therefore, by including a plurality of control areas, each control area including different concentrations of the secondary antibody, the control areas may change their color/color intensity at different time points.

According to some embodiments, a first of the control areas may include a large concentration of the secondary antibody and a color/color intensity change will occur as soon as the sample reaches the first control area. This first control area is indicative of the test being performed correctly and initiates the biological timer of the test”. According to some embodiments, if the image is captured prior to the color/color intensity of the first control area occurring a message may be issued to the user (e.g. via a dedicated App) that the test is invalid and should be re-executed. According to some embodiments, a second of the control areas may include a concentration of the secondary antibody that ensures that it will only change its color/color intensity upon sufficient time having passed to expect a test-mark color/color intensity change. According to some embodiments, if the image is captured prior to the color/color intensity change of the second control areas, a message may be issued to the user (e.g. via a dedicated App) to re-capture the image.

According to some embodiments, a third of the control areas may include a concentration of the secondary antibody that ensures that it will only change its color/color intensity when too much time has passed to ensure test reliability. According to some embodiments, if the image is captured after the color/color intensity change of the third control area, a message may be issued to the user (e.g. via a dedicated App) that the test is invalid and should be re-executed.

According to some embodiments, additional control areas may also be included, each of the additional control areas including different concentrations of the secondary antibody in a range between that of the second and third control areas, such that their color/color intensity changes are indicative of the amount of time having passed between appearance of the first control area color/color intensity change and the capturing of the image, e.g., in intervals of 5, 10, 20, 30 or 60 seconds. Each possibility is a separate embodiment.

According to some embodiments, the LFIA includes a plurality of control area indicative of the biological age/time of the test.

According to some embodiments, calculating the concentration of the analyte in the biological sample may further include taking into consideration the biological age/time of the test.

According to some embodiments, there is provided a lateral flow immunoassay analysis (LFIA) strip including: a membrane configured for lateral flow of a test sample, the membrane including: a control area configured to produce a colorimetric control mark (C) in response to the lateral flow of the test sample therethrough irrespective of presence or absence of an analyte in the test sample; a test area comprising an agent configured to produce a colorimetric test mark (T) in response to presence of the analyte in the test sample, a reference area including: an isolating layer; and a secondary membrane comprising a colorimetric reference mark (R) having a predefined color intensity, the colorimetric reference mark (R) formed before exposing the LFIA strip to the biological sample, wherein the secondary membrane is shielded off from contact with the test sample by the isolating layer, such that the color intensity of the colorimetric reference mark (R) is independent of lateral flow of the test sample through the LFIA strip.

According to some embodiments, the LFIA strip includes at least two test areas (T), each test area comprising a different concentration of the agent, such that a minimal concentration of the analyte sufficient to produce the colorimetric test mark (T) differs in each area. According to some embodiments, the LFIA strip includes more than two test areas (e.g. 3, 4, 6, 8, 10 or more test areas), least two test areas (T), each test area comprising a different concentration of the agent. Each possibility is a separate embodiment.

According to some embodiments, the LFIA strip comprises at least two colorimetric reference marks (R), each having a different preformed and/or predefined color intensity.

According to some embodiments, the LFIA strip further includes a sample pad configured to receive the test sample. According to some embodiments, the LFIA strip further includes a conjugation pad configured to release a conjugate that binds the analyte if present. According to some embodiments, the LFIA strip further includes an absorption pad configured to wick the prevent backflow of the test sample. According to some embodiments, the LFIA strip further includes an adhesive pad configured to sequentially adhere the sample pad, the conjugation pad, the membrane and the absorption pad.

According to some embodiments, a first of the control areas may include a large concentration of the secondary antibody. This first control area is indicative of the test being performed correctly and initiates the biological timer of the test”. According to some embodiments, if the image is captured prior to the color/color intensity of the first control area occurring a message may be issued to the user (e.g. via a dedicated App) that the test is invalid and should be re-executed.

According to some embodiments, a second of the control areas may include a concentration of the secondary antibody that ensures that it will only change its color/color intensity upon sufficient time having passed to expect a test-mark color/color intensity change. According to some embodiments, if the image is captured prior to the color/color intensity change of the second control line, a message may be issued to the user (e.g. via a dedicated App) to re-capture the image.

Reference is now made to FIG. 1, which shows a schematic outline of an LFIA strip 100 according to some embodiments.

LFAI strip 100 includes a test area 110 which includes an agent configured to change its color and/or color intensity in response to binding an analyte present in a biological sample run through LFIA strip 100. Here test area 110 includes a single test area, here in the form of a T-line. LFIA strip 100 further includes a control area 120 (here in the form of a C-line), which produces a colorimetric control mark in response to lateral flow of the biological sample therethrough, but irrespective of presence or absence of an analyte in the test sample. Similarly, to many LFIA strips the control area serves as a positive control validating the lateral flow of the biological sample through LFIA strip 100. In addition, LFIA strip 100 includes a plurality of reference areas, collectively referred to with the numerical 130, here in the form of reference line each having a different color intensity. In order to overcome differences among cameras, illumination levels, and shadowing, reference areas 130 (here in the form of lines, also referred to as R-lines) are embedded near test-area 110 to ensure that the environmental illumination factors have a same impact on T- line 110 as on R-line 130, wherein reference-lines 130 each have a different pre-defined, known colorimetric levels.

Exemplary configurations of single T-area LFIA strips 200 are shown in FIG. 2A and FIG. LFIA strips 200 include a sample pad 240 configured to receive a biological sample, a conjugation pad 250 configured to release a conjugate that binds the analyte if present, an absorption pad 260 configured to prevent backflow of the test sample, a nitrocellulose membrane 210 containing a test area 210 (here a single T-dot), a nitrocellulose membrane 280 containing a reference area 230 (with a single R-dot in FIG. 2A or a plurality of R-lines in FIG. 2B), an isolator 232 configured to prevent the biological sample from reaching the R-dot(s), a C-line configured to validate the lateral flow of the biological sample and an adhesive pad 280 configured to adhere all elements into a single strip.

A trained Al model can then be applied on the color intensity of reference-dot(s) 230 and their known “real” color intensity to remove the impact of the environmental conditions on the captured image.

Then the real concentration of the T-mark may be determined by applying a second trained Al model on the normalized image based on utilizing a linear region (shaded in grey) of the color intensity as shown in FIG. 3. As a result, the test can be done at home using a standard smartphone by untrained users.

During the training process, the concentration/color intensity of the T-mark and the embedded reference marks concentrations are known. The Al model (e.g. convoluted network model - CNN) learns the color intensity distance between the T-mark and matches it to a concentration distance. The given T-mark is then tagged to the color intensity reference mark (obtained by LFAI reader and denoted herein as “CT”) with the shortest concentration distance.

At inference, for a given T-mark, the model predicts a probability P for the T-mark to match with a specific color intensity. P[i] is the probability of the color intensity L[i] to have the shortest distance to the T-mark. If C[i] is the predefined concentration of the color intensity L[i], the T- mark concentration will be defined by the equation:

T line concentration = P[i] * C[i]

Reference is now made to FIG. 4A and FIG. 4B which show LFIA strips 400 with multiple test areas 410 with different detection cut-off points. This allows expanding the linear range 510 of the strip as illustrated in FIG. 5. Moreover, as seen from FIG. 5, it also improves accuracy in the area where overlapping 520 exists.

According to some embodiments, for a plurality of test areas, e.g. in the form of T-dots with multi-cutoff points, k different concentrations are obtained, {T_dot_concentration[l], ... , T_dot_concentration[k]}. As shown in FIG. 5, for the k T-dots, there are k linear areas, such that each T-dot will have its respective linear area for its color intensity detection. Preferably, the linear area of neighboring T-dots will have a partial overlap (520).

That is, if LA[i] is the linear area (shown as Linear 1 in FIG. 5) corresponding to Tdot[i], r[i] = dist(T_dot_concentation[i], LA[i]), it denotes the distance between T_dot_concentation[i] to the central point at LA[i], During the training phase of the model, a final concentration (FC) is tagged as follows: FC = T_dot_concentation[i], where r[i] = Min{r[j]}, for all j=l,,.,k. It allows the inference process to provide a value P[j] which represents the probability (weight) of a given T-dot[j] value relative to other T-dot[i] values. Hence, the Strip concentration will be defined by the following equation: Strip concentration = P[/] * T _dot_concentration[j].

Reference is now made to FIG. 6, which depicts an LFIA strip 600 that is essentially similar to the LFIA strips described with reference to the previous figures, but which includes a plurality of test areas (single or set), here T-dots 610a and 610b each including a different agent for detecting different analytes. An overall outline of a flow 700 of the herein disclosed method is provided in FIG. 7. It is understood that while the steps are shown as sequential some may be conducted simultaneously. According to some embodiments, at least some of the steps may be executed by applying a single Al model capable of executing various tasks. According to some embodiments, at least some of the steps are executed by applying different Al models.

As seen, initially, in step 710 a user may capture an image of the LFIA strip, for example using a smartphone. According to some embodiments, the image may be captured via a dedicated App. Alternatively, the image may be uploaded to the App after capturing thereof.

In step 720, a first Al model may then be applied, which first model is capable of detecting/framing a region of interest (ROI). Various ROI recognition models may be utilized and specifically trained on LFIA strips. It is understood that during setup, the first algorithm (object detection algorithm, such as YOLO) is trained (step 722) for recognition of LFIA strips. Moreover, as more images are acquired, the training may be continuously updated.

Once the region of interest (ROI) is identified, a second Al model (a trained CNN model) may be applied on the ROI (step 730), which algorithm is configured to normalize the image properties of the ROI, based on one or more reference marks embedded in the LFIA strip. That is, based on the noise level derived from a detected deviance between the expected color intensity and the actual color intensity of the reference mark in the captured image, an inverse noise algorithm is applied (step 734) on the one or more test areas. As before, during setup, the second algorithm is trained (step 732) on a plurality of images captured under various environmental conditions.

Once the noise level of the image captured in step 710 is identified, the concentration of the analyte is determined based on the modified color intensity of the T-marks in step 740. This is achieved by applying a third Al model (CNN model), trained during setup (step 732) for determining a correlation between color intensity and analyte concentration. According to some embodiments, this step also includes taking into consideration the biological age/time of the sample, based on a color/color intensity of two or more control-areas, as essentially described herein. Finally, in step 750 the analyte concentration is outputted and provided to the user and/or a caregiver.

It is understood by one of ordinary skill in the art, that the first, second and third Al models may be separate models or be part of an integrated flow.

Reference is now made to FIG. 8, which depicts an LFIA 800 strip that is essentially similar to the LFIA strips described with reference to the previous figures, but which further includes a modified control area 870 with a plurality of control-areas, also referred to as an “elapsed time” control area or ETC in which each is configured to change its color and/or color intensity according to the time that has passed since the biological sample has reached the control area. That is, each control area changes its color at different intervals and as such reflects the time that has passed between loading of the sample on sample pad 240 and image capturing. It is further understood that in this instance, method 700 may be modified to include a fourth Al model configured to adjust the determined analyte concentration based on the time that has passed. Moreover, in case too much time has passed, and the determined concentration is at risk of being inaccurate, an error message may be issued in step 750 of method 700.

Reference is now made to FIG. 9, which illustrates an LFIA strip 900 and associated AL based analysis 950. It is understood that while specific models are mentioned with respect to this figure, these models are exemplary and may be replaced by others. Those skilled in the art will readily understand which algorithms can be suitable at the various steps.

As shown, LFIA 900 includes a QR-code 902 which links the LFIA strip to a dedicated mobile application. In the mobile application, the user is guided through the testing steps and then directed to capture an image of the LFIA upon completion of the test. The user may then upload the image (alternatively it is automatically uploaded), and the virtual lab running dedicated Al models is initiated. In the first step (step 952), the LFIA strip 900 is identified in the image using ROI detection algorithms such as but not limited to Faster R-CNN. In this step the test validity is also identified by identification of a first control area, here shown as first control line 904 which includes a sufficiently high concentration of a secondary antibody to initiate a color/color intensity change essentially as soon as the sample reaches first control line 904. Once the test is validated, a step of noise removal (step 956) is initiated. That is, the image is calibrated using a matrix of 1 reference marks 906 having a predefined color and/or color intensity, to remove noise from the image resulting from the camera used and/or environment conditions, such as lightning etc. Step 956 includes picture distortion analysis via an analysis of the color/color intensity of reference marks 906, and an inverse distortion picture is created. Suitable Al models for use in step 956 include Mask-RCNN, Unet, DeepLav3+, and K-Means.

In the next step, step 958 an analysis of the protein concentration of various proteins (here 4 different proteins) is conducted based on the color/color intensity of test areas 908a-908d. Step 958 includes utilizing Al models trained on the color/color intensity of a large plurality of controlmarks having different known concentration of the analyte (as further elaborated herein). Suitable Al models for use in step 958 include Mask-RCNN, Unet, DeepLav3+, and K-Means. As seen, each test area includes several test dots, each including different concentrations of test antibody (here 3), thereby ensuring that a window obtained at which there is a linear correlation between color intensity and analyte concentration.

Finally, in step 960, which is optional LLM models may be applied to generate a summary report with the test results and optionally an indication of the subject’s risk and/or a recommendation for further actions (e.g. further testing). Suitable Al models for use in step 960 include MedPalm or GPT-4 tuned.

As used herein, the terms “approximately”, “essentially” and “about” in reference to a number are generally taken to include numbers that fall within a range of 5% or in the range of 1% in either direction (greater than or less than) the number unless otherwise stated or otherwise evident from the context (except where such number would exceed 100% of a possible value). Where ranges are stated, the endpoints are included within the range unless otherwise stated or otherwise evident from the context.

As used herein, the singular forms "a," "an" and "the" include plural referents unless the context clearly dictates otherwise.

As used herein, "optional" or "optionally" means that the subsequently described event or circumstance does or does not occur, and that the description includes instances where said event or circumstance occurs and instances where it does not. While a number of exemplary aspects and embodiments have been discussed above, those of skill in the art will recognize certain modifications, additions and sub-combinations thereof. It is therefore intended that the following appended claims and claims hereafter introduced be interpreted to include all such modifications, additions and sub-combinations as are within their true spirit and scope.

Claims

1. A method for Al-based quantitative colorimetric lateral flow immunoassay analysis (LFIA), the method comprising: a) receiving from a user an image captured with a handheld digital imaging device, the image comprising a picture of a LFIA strip having been exposed to a biological sample of the user, wherein the LFIA strip comprises: at least two test areas, each test area comprising a different concentration of an agent configured to produce a colorimetric test mark (T) in response to the presence of an analyte in the biological sample, such that an essentially linear correlation between color/color intensity and analyte concentration is ensured and a reference area comprising at least two colorimetric reference marks (R), each reference mark having a different predefined color and/or color intensity, formed irrespective of the exposing of the LFIA strip to the biological sample; b) removing image environmental conditions differences from the captured image or parts thereof that contain the picture of the LFIA strip, by applying a first Al model on the captured image or parts thereof, the first Al model trained on a plurality of images of the colorimetric reference mark (R) captured under different imaging conditions and normalized to a color/color intensity of the colorimetric reference mark (R) measured using an LFIA reader, thereby obtaining a normalized image; and c) calculating a concentration of the analyte in the biological sample by applying a second Al model on the color/color intensity of the colorimetric test mark (T) in the normalized image, the second Al model trained on a plurality of various control test mark (CT) intensities measured using an LFIA reader, each of the plurality of control test mark (CT) intensities labeled with its associated concentration of the analyte; and d) transmitting the calculated analyte concentration to the user and/or caregiver.

2. The method of claim 1, wherein the LFIA strip further comprises a control area configured to produce at least two colorimetric control marks (C) in response to the lateral flow of the test sample therethrough irrespective of presence or absence of an analyte in the test sample, wherein each of the at least two colorimetric control marks (C) has a different response time thereby providing an indication of the time that has elapsed between the biological sample reaching the control area and image capturing (biological time of the test).

3. The method of claim 2, further comprising adjusting the calculated analyte concentration based on biological time.

4. The method of claim 2 or 3, wherein if the biological time exceeds a predetermined threshold value, the calculated analyte concentration is deemed invalid.

5. The method of any one of claims 1-4, wherein the imaging conditions comprise camera quality, camera configuration/setting, surrounding light condition, shadowing and/or any combination thereof.

6. The method of any one of claims 1-5, further comprising pre-processing the image to convert all colorimetric marks to a single-color monochromatic scale.

7. The method of any one of claims 1-6, further comprising applying an object detection algorithm on the image to identify the LFIA strip or parts thereof.

8. The method of any one of claims 1-7, wherein the first and second Al models are convoluted neural network (CNN) models.

9. The method of claim 8, wherein the CNN model is a multi-classification model.

10. The method of claim 9, wherein the multi-classification model is configured to predict a probability that the colorimetric test mark (T) matches with a color/color intensity of each of the plurality of control test marks (CT).

11. The method of any one of claims 1-10, wherein upon exposure to a given analyte concentration, at least a portion of the at least two test areas provide a different color intensity, thereby increasing linear data point range and accuracy.

12. The method of claim 11, wherein the multi-classification model is further configured to calculate the concentrations of the analyte in the biological sample, based on matching of the color and/or color intensities obtained in each of the at least two test areas to a most similar of the plurality of control test marks (CT).

13. A lateral flow immunoassay analysis (LFIA) strip comprising: a membrane configured for lateral flow of a test sample, the membrane comprising: a control area configured to produce a colorimetric control mark (C) in response to the lateral flow of the test sample therethrough irrespective of presence or absence of an analyte in the test sample; a test area comprising an agent configured to produce a colorimetric test mark (T) in response to presence of the analyte in the test sample, a reference area comprising: an isolating layer; and a secondary membrane comprising a colorimetric reference mark (R) having a predefined color intensity, the colorimetric reference mark (R) formed before exposing the LFIA strip to the biological sample, wherein the secondary membrane is shielded off from contact with the test sample by the isolating layer, such that the color intensity of the colorimetric reference mark (R) is independent of lateral flow of the test sample through the LFIA strip.

14. The LFIA strip of claim 13, comprising at least two test areas, each test area comprising a different concentration of the agent, such that a minimal concentration of the analyte sufficient to produce the colorimetric test mark (T) differs in each area.

15. The LFIA strip of claim 13 or 14, comprising at least two colorimetric reference marks (R), each having a different preformed and predefined color intensity.

16. The LFIA strip of any one of claims 13-15, wherein the membrane and the secondary membrane are made from the same material.

17. The LFIA strip of claim 16, wherein the membrane and the secondary membrane are nitrocellulose membranes.

18. The LFIA strip of any one of claims 13-17, further comprising a sample pad configured to receive the test sample.

19. The LFIA strip of claim 18, further comprising a conjugation pad configured to release a conjugate that binds the analyte if present.

20. The LFIA strip of claim 19, further comprising an absorption pad configured to wick the prevent backflow of the test sample.

21. The LFIA strip of any one of claims 13-20 further comprising an adhesive pad configured to sequentially adhere the sample pad, the conjugation pad, the membrane and the absorption pad.

22. The LFIA strip of any one of claims 13-21, wherein the control area comprises at least two colorimetric control marks (C), wherein each of the at least two colorimetric control marks (C) has a different response time which is indicative of the time that has elapsed between loading of the biological sample and image capturing.

23. A method for quantitative analysis of lateral flow immunoassay analysis (LFIA) results, the method comprising: g) loading a biological sample on an LFIA strip comprising: i. at least two test areas, each test area comprising a different concentration of an agent configured to produce a colorimetric test mark (T) in response to the presence of an analyte in the biological sample, such that an essentially linear correlation between color intensity and analyte concentration is obtained; ii. a reference area comprising at least two colorimetric reference marks (R), each reference mark having a different predefined color and color intensity, formed before exposing the LFIA strip to the biological sample; and iii. a control area configured to produce a colorimetric control mark (C) in response to the lateral flow of the test sample therethrough irrespective of presence or absence of an analyte in the test sample; h) capturing an image of the LFIA strip a predetermined time after production of the colorimetric control mark (C); i) applying an object detection algorithm on the image to identify the LFIA strip or parts thereof; j) removing image environmental conditions differences from the captured image or parts thereof that contain the picture of the LFIA strip, by applying a first CNN model on the captured image or parts thereof, the first CNN model trained on a plurality of images of colorimetric reference mark (R) captured under different imaging conditions and normalized to a color intensity of the colorimetric reference mark (R) measured using an LFIA reader, thereby obtaining a normalized image; and k) calculating a concentration of the analyte in the biological sample by applying a second CNN model on the color intensity of the colorimetric test mark (T) in the normalized image, the second Al model trained on a plurality of various control test mark (CT) intensities measured using an LFIA reader, each of the plurality of control test mark (CT) intensities labeled with its associated concentration of the analyte; and

1) transmitting the calculated concentration to the user and/or caregiver.

24. The method of claim 23, wherein the control area comprises at least two colorimetric control marks (C) each having a different response time indicative of the time that has elapsed between loading of the biological sample and image capturing.

25. The method of claim 24, further comprising adjusting the calculated analyte concentration based on time that has elapsed between loading of the biological sample and image capturing.

26. The method of claim 24 or 25, wherein if the time elapsed exceeds a predetermined threshold value, the calculated analyte concentration is deemed invalid.