Description:FIELD OF THE INVENTION:
The present invention generally relates to the Method for Plasma Glucose Monitoring System. It relates to In vitro Diagnostics (IVD) tests. Particularly it relates to, any optical lens/camera-based/ image capturing means integrated with programmed application including but not limited to smartphone-based system, methods, and components thereof. More particularly, it relates to the method for Plasma Glucose Monitoring System which uses integrated paper strip kit and an application installed on any camera with either offline or online programmed application backed by machine learning algorithms for object detection, segmentation, feature extraction, classification and quantification combined together as a technology platform coined as Digital Image Colorimetry (DIC) for the detection of the concentration of glucose present in the plasma, where the results are not sensitive to the camera specifications, background settings and user-skill.

BACKGROUND OF THE INVENTION
Diabetes describes a group of metabolic diseases that cause high blood sugar levels. In recent years, diabetes has become one of the leading causes of death worldwide. According to the World Health Organization, around 1.5 million people worldwide died due to diabetes in 2019. It is estimated that 537 million people are currently living with diabetes all over the world. By 2045, projections show this number rising to some 783 million diabetics globally. India is the second-largest country of diabetics worldwide with 101 million diabetics and 136 million Prediabetics as per the latest study published by Indian Council of Medical Research in June 2023.
Blood sugar monitoring is an important aspect of diabetes management. In India, approximately only 1% of diabetics have their own blood glucose monitors, while more than 90% of diabetics in developed countries like Europe and America use their own portable blood glucose monitors.
In addition to the large number of diagnosed diabetic patients, there is a significant proportion of the population that is undiagnosed for various reasons or at potentially high risk. Timely screening and continued monitoring of the glucose level holds the key to manage the diabetes. Primarily following methods are used for Glucose Monitoring –
a) Traditional Phlebotomy Based Invasive Procedure – It requires a blood sample to be drawn out of the vein and analysing the same in the laboratory where trained technician/pathologist process the sample in equipment by first separating plasma from the whole blood and then detecting glucose using accredited biochemistry analysis equipment. The results are accurate, but the procedure is challenged by the need of specialized equipment, expert technicians, delayed availability of the results and the expenses associated.
b) Minimally Invasive Procedure – It requires a tiny amount of ‘capillary’ or subcutaneous blood that is commonly collected by finger-prick using a lancet or microneedle patches or other equivalent means. The sample may then be processed in-situ, typically using a portable analyser. The results are reasonably accurate for point of care applications. Common examples include Glucometers that are widely used to measure the glucose level for at-home or point of care applications. These methods commonly require an analytic device, performing optical or colorimetric analysis. Since plasma separation commonly requires additional embodiment, whole-blood based analysis is more common in case of glucometers, where further mathematical correlations are invoked to correlate the whole blood glucose with the plasma glucose level that is a clinically referenced benchmark. However, because of scientifically established artefacts, these correlations may be questioned at situations where the measurement itself may be interfered due to other blood analytes such as the haematocrit (i.e., volume fraction of the red blood cells). Irrespective of this aspect, a more compelling technical bottleneck is the compulsive need of an analytic instrument that hinders portability, ease of analysis, along with the accompanying adverse cost propositions.
c) Non-Invasive Procedure – These procures commonly pertain to ‘bloodless blood tests’, where correlations are developed between the blood analytic parameter and another signal that is generated non-invasively by interfacing a sensor with the patient. These sensors commonly deploy different optical methods such as near-infrared reflectance spectroscopy (NIRS), polarized optical rotation, Raman spectroscopy, fluorescence, optical coherence tomography (OCT) and so on, requiring controlled incident waves (infrared, radio frequency etc.). As a typical example, a monitor shines a beam of light through body skin, which is then reflected and scattered by the target molecules in the blood. The sensor then detects signals from the light that is reflected back. The particular reflection pattern is then calibrated with the analyte concentration to arrive at the test result. This calibration, however, is commonly interfered with other analytes in the blood, resulting in signals that lack the desired level of specificity. Further, the signal to noise ratio is often very weak for the specific target molecules, requiring specialized filtering techniques with no assurance of approaching the laboratory gold standard values.

Many of the minimally-invasive and all the non-invasive procedures commonly target the interstitial fluid as against the free-flowing blood. While this is attractive from circumventing sample collection issues and needle-phobia, there is a lack of adequate physiologically-established evidence on the correlation between the blood and interstitial fluid analytes, in particular, for patients suffering from severe generalized edema, such as those with hypoalbuminemia and hepatic failure, and similar other ailments.

Also, the general challenges in Smartphone based optical sensing of analytes are lack of standardization of the smartphone cameras, illumination sources and backgrounds and human error due to unprofessional handling of the device. Firstly, the response of Smartphone cameras of the same colorimetric response varies significantly due to the small camera sensor size, variations in the circuitry, poor repeatability and random noise due to the inherent randomness of light sensing. In addition, the commercial LED lights used for controlled illumination in a 3D enclosure introduce some new challenges like error due to uncontrolled light reflection, decay of the illumination with time, non-uniformity in illumination etc. All of the above-mentioned problems impose a significant variation in sensing of the colour response. Furthermore, the spectral densities of different commercial LED lights also vary significantly and thus a 3D enclosure calibrated for a particular LED is not suitable for a new LED and needs complicated correction procedure. Additionally, the different Smartphone camera parameters like gamma correction, auto white balance correction matrix, different colour constancy features and other non-linearity functions are pre-calibrated for a particular spectrum of light and in general not disclosed by the manufacturers. Thus, colour and white balance correction without the prior knowledge of these parameters are a difficult blind inverse problem which in most cases is intractable. Moreover, these 3D enclosures and their maintenance also incorporate an extra complexity for the end user and increase the unavoidable human error due to manual handling of the device.

Considering the advantages and shortcomings of various methodologies, it is thus imperative to develop a technology that harnesses the combined rigor of gold-standard reaction chemistry and the simplicity and user-friendliness of common POC tests in an amalgamated format. Previous developments in this connection were successful in translating the gold-standard biochemistry to the framework of portable-device based procedures using optical signal based quantitative readout technologies. However, the need of a device unit could not be circumvented, adding to both cost and complexity in the procedure, and inhibiting their use for seamless personalized care or monitoring of vulnerable patients, particularly when they are traveling or in other engagements when the access to any specialized device remains unrealistic, leaving apart the operation and maintenance issues associated. Despite such compelling demands, the need of a device unit or analytic instrument could not be completely circumvented thus far for the same, as attributable to the challenges associated with the standardization of signal acquisition framework that may otherwise be adversely affected by aberrations due to lack of control thereof. The present invention aims to circumvent the same by arriving at a robust technology that interfaces a paper-strip with a stand-alone camera of any arbitrary specification, without any specialized control on the image acquisition framework.

Patent Literature:
1. IN201931030435: System For Estimation of Plasma Glucose By Integrated Paper-based Device And Method Thereof
IN201931030435 discloses the low-cost paper-based microfluidic device, developed for detection of blood glucose level. It is comprised of an integrated blood plasma separation module and plasma glucose detector module having image processing means and display unit. The system includes a smartphone with the camera means for imaging colour change of plasma on reaction with glucose detecting chemicals in plasma glucose detector of paper based integrated platform and analytic tool for mapping the parameters with plasma glucose level for producing quantitative results. The said system further comprising an enclosure for supporting a cartridge based integrated blood plasma separation module and plasma glucose detector and a smartphone unit said camera means and analytic tool for estimation of glucose.
The system of 201931030435 involves high cost and difficulty to use as it required a controlled environment (a Black colour imaging box) for light exposure and illumination. This Box helps to create a threshold range of light exposure for test results. It needs a constant light source. Further, this box is bulky and not easy to carry.
2. US9957545B2 - Blood glucose measurement unit, blood glucose measurement system comprising same.
US9957545B2 discloses a blood glucose measurement unit comprising a transparent first substrate consisting of a blood inflow region into which blood flows and a reaction region connected to the blood inflow region which is formed on one surface thereof; a transparent second substrate coupled to the first substrate and comprising a blood aperture through which the blood flowing into the blood inflow region passes; and a reagent distributed to the reaction region so as to react with the blood glucose of the blood which has flown into the blood inflow region.
The device of US9957545B2 is not user-friendly as a separate COMS image sensor which is available only in few cameras or needs to be installed on the camera at the time of image capturing; while the proposed invention does not require any such external sensor/ reader/ viewing/image box and image of the test strip can be captured in any light.
Further, the working of the COMS image sensor is based on the number of photons transmitted through the reaction region, on the contrary, proposed invention works on the Digital Image Colorimetry principal.
In US9957545B2, the reagent is to be added manually using a pipette leading to error-prone process as reagent quantity is not controlled whereas in the proposed invention test strip is provided with an inbuilt/ pre-embedded reagent and no manual intervention is required.
3. US9445749B2: Smartphone-based apparatus and method for obtaining repeatable, quantitative colorimetric measurement
The method according to US9445749B2 involves imaging a test strip on which a colorimetric reaction of a target sample has occurred due to test strip illumination by the smartphone. The smartphone includes a smartphone app and a smartphone accessory that provides an external environment-independent/internal light-free, imaging environment independent of the smartphone platform being used. The result can then be presented quantitatively or turned into a more consumer-friendly measurement (positive, negative, above average, etc.), displayed to the user, stored for later use, and communicated to a location where practitioners can provide additional review. Additionally, social media integration can allow for device results to be broadcast to specific audiences, to compare healthy living with others, to compete in health-based games, create mappings, and other applications.
In US9445749B2, the image-capturing strip must be placed in an external device/accessory and then the image can be captured. It works only in a standard/closed environment.
In the proposed invention, there is no requirement of any external device and any pre-printed/pre-defined calibration image. On the other hand, the results are given by the high-end image classification model trained with a large dataset using machine learning.
4. WO2022159570A1: Microfluidic devices and rapid processing thereof
This disclosure relates to paper microfluidic devices for use in combination with a viewing box assembly for imaging and rapid identification and quantification of target analytes in a fluid sample that is deposited onto the device such that one or more target analytes in the sample react with one or more diagnostic components on the paper, causing a detectable reaction. The reacted microfluidic device may then be placed inside an opaque viewing box having an internal light source and top panel viewing aperture through which the microfluidic device may be imaged using a mobile electronic device and graphical user interface for purposes of detecting and quantifying one or more target analytes. In some embodiments, the microfluidic device includes diagnostic paper and a base. In some embodiments, the microfluidic device includes a filter layer on top of the diagnostic paper layer.
The main disadvantage of WO2022159570A1 is a high reaction time of as long as 30min to 50 min as manual operations are involved in reagent processing. The use of an external viewing box is also disclosed in WO2022159570A1 causing inconvenience to the user. Moreover, manual processing leads to inaccurate reading.
5. KR102399939B1: Detection method and detection pad
A detection pad for inspecting a target in a fluid is disclosed in this invention. It includes a machine-readable code displayed with a first reference colour; a second reference colour area used to remove the effect of discoloration due to a fluid; and a detection area including a reagent of which colour is changed in reaction with the target.
KR102399939B1 discloses the use of machine-readable codes (preferably QR codes) for image processing in colorimetry techniques. These machine-readable codes display information on the number of reagent pads included in the detection pad and information on the classification of the detection pads. In addition, the image is compared with a calibration/reference image which is pre-printed/pre-defined on the device.
In the proposed invention, there is no requirement for pre-printed/pre-defined calibration image. Region of Interest is extracted by trained model using large dataset. Proposed inventions separate plasma from blood and then measures glucose.
Non Patent Literature:
L1: https://www.sciencedirect.com/science/article/abs/pii/S0039914021008766
Microfluidic paper-based analytical device by using Pt nanoparticles as highly active peroxidase mimic for simultaneous detection of glucose and uric acid with use of a smartphone
Herein, a simple microfluidic paper-based analytical device (µPAD) by using platinum nanoparticles (Pt NPs) as highly active peroxidase mimic for simultaneous determination of glucose and uric acid was fabricated.
The µPAD consisted of one sample transportation layer, four paper-based detection chips, and two layers of hydrophobic polyethylene terephthalate (PET) films. The four detection chips were immobilized with various chromogenic reagents, Pt NPs, and specific oxidase (glucose oxidase or uricase).
H2O2 generated by specific enzymatic reactions could oxidize co-immobilized chromogenic reagents to produce coloured products by using Pt NPs as an efficient catalyst. The multi-layered structure of µPAD could effectively improve colour uniformity and colour intensity. Total colour intensity from each two detection chips modified with distinct chromogenic reagents was used for quantitative analysis of glucose and uric acid, respectively, resulting in significantly improved sensitivity. The linear range for glucose and uric acid detection was 0.01–5.0 mM and 0.01–2.5 mM, respectively.
The use of platinum nanoparticles makes the device costly, additionally, the use four channel paper on PET film also increases cost. Further, the devices do not separate plasma from blood thereby producing factually incorrect output.
L2 – https://link.springer.com/article/10.1007/s00604-017-2575-7
Microfluidic paper-based device for colorimetric determination of glucose based on a metal- organic framework acting as peroxidase mimetic
This work presents a microfluidic paper-based analytical device (µPAD) for glucose determination using a supported metal-organic framework (MOF) acting as a peroxidase mimic. The catalytic action of glucose oxidase (GOx) on glucose causes the formation of H2O2, and the MOF causes the oxidation of 3,3',5,5'-tetramethylbenzidine (TMB) by H2O2 to form a blue-green product with an absorption peak at 650 nm in the detection zone. A digital camera and the iOS feature of a smartphone are used for the quantitation of glucose with the S coordinate of the HSV colour space as the analytical parameter. Different factors such as the concentration of TMB, GOx and MOF, pH and buffer, sample volume, reaction time and reagent position in the µPAD were optimized. Under optimal conditions, the value for the S coordinate increases linearly up to 150 µmol·L-1 glucose concentrations, with a 2.5 µmol·L-1 detection limit. The µPAD remains stable for 21 days under conventional storage conditions. Such an enzyme mimetic-based assay for glucose determination using Fe-MIL-101 MOF implemented in a microfluidic paper-based device possesses advantages over enzyme-based assays in terms of costs, durability and stability compared to other existing glucose determination methods. The procedure was applied to the determination of glucose in (spiked) serum and urine.
This device determines Glucose using a supported metal-organic framework (MOF). The step of plasma separation from blood is missing which may lead to incorrect output.
L3 – https://www.sciencedirect.com/science/article/abs/pii/S0925400520313848
Machine learning-based colorimetric determination of glucose in artificial saliva with different reagents using a smartphone coupled uPAD
In this study, a new implementation of machine learning classifiers based on a colour change in µPADs was proposed to classify glucose level in artificial saliva. Machine learning classifiers were trained using images captured under seven different illumination conditions with four different smartphones, which improved the robustness of the platform against the illumination variance and camera optics.
This device detects Glucose in artificial saliva whereas the proposed invention detects Glucose in plasma blood. This device has 3 detection zones charged with different detection mixtures (a. KI, b. KI+CHI and c. TMB).
Further, the device is capturing images from 4 different smartphones under 7 different illumination conditions and then with the help of the Glucosensing application the result is displayed. The device needs a controlled environment for proper working i.e. the use of 3D device (external device) for training their system is disclosed. On the contrary, the proposed invention is developed and works in an open environment without the use of any external device.
Hence, in light of the foregoing discussion, there exists a need to overcome the drawbacks associated with existing prior arts. Firstly, the present invention has eliminated the black box/controlled environment while capturing image/video of strip post reaction and has standardized the result accuracy for all exposure and varied light intensities. The result will be the same across all types of light i.e. daylight, yellow light, white light, etc. Secondly, the Design of the test strip has been optimized in the present invention resulting in a quick reaction time less than 3 minutes and there is less to no chance of blood spilling. Thirdly, system is using machine learning algorithm to predict glucose levels based on image data captured from glucose strips through a combination of image pre-processing techniques and trained models. Test & processing time is improved and cost is also reduced by around 50%. Result accuracy is also substantially improved. The present invention delivers more value-added benefits to consumers.
OBJECTIVES OF THE INVENTION
The primary objective of the present invention is to provide a Method for any optical lens/camera-based device/ Smartphone-Based, point-of-care based, Plasma Glucose Monitoring System (PGMS) that eliminates separate controlled environment/imaging box/viewing box/3D enclosure or any such external device while capturing image/video of strip post reaction and that has standardized the result accuracy for all exposure and varied light intensities.
Furthermore, objective of the present invention is to provide a Plasma Glucose Monitoring System (PGMS) and Method that uses paper test strips and a mobile app backed by machine learning algorithm for Digital Image Colorimetric detection of the concentration of glucose present in the plasma.
Another objective of the present invention is to provide a Plasma Glucose Monitoring System which can be used online and offline i.e. with and without internet.
Another objective of the present invention is to provide a Plasma Glucose Monitoring System which can work on all type of cameras integrated with programmed application in different environment without using any external device while capturing image/video of strip post reaction.
Another objective of the present invention is to provide a Plasma Glucose Monitoring System which can be carried out in any types of light luminosity i.e. daylight, yellow light, white light etc.
Yet another objective of the present invention is to provide a first of its kind ultra-high accuracy plasma-separation integrated paper strip kit for doing Glucose tests using a drop of blood.
Yet another objective of the present invention is to provide a Plasma Glucose Monitoring System which firsts separate]es plasma from blood and then measures Glucose, which is more accurate, reliable, user friendly and affordable to any common man.
Yet another objective of the present invention is to provide a Plasma Glucose Monitoring System with microfluidic separation, capillary transport and analyte detection in a data-science empowered digital image colorimetric analytics format that works universally across all the smartphones without any quality compromise and at ultra-low cost.
Yet another objective of the present invention is to provide a Plasma Glucose Monitoring System which has an improvised strip design, leading to ease of use for patients/user.
Yet another objective of the present invention is to provide a Plasma Glucose Monitoring System which uses method of advanced image processing that properly identifies the specific region of interest on the test strip.
Yet another objective of the present invention is to provide a Plasma Glucose Monitoring System based on point-of-care method, which can prove useful in demand-supply scenarios of more patient’s vs lesser pathologists.
Yet another objective of the present invention is to provide a Plasma Glucose Monitoring System by capturing image in an application from any type of camera.
Yet another objective of the present invention is to provide a Plasma Glucose Monitoring System at ultra-low cost.
Yet another objective of the present invention is to provide a Plasma Glucose Monitoring System with small strip having inbuilt reagents which are stabilized at room temperature.
Yet another objective of the present invention is to provide a Plasma Glucose Monitoring System with unique strip design leading to ease of usage. The design solved the issues to be able to take an apt amount of blood sample, with less to no chance of blood spillage, for reaction to take place. It also improves reaction time.
Yet another objective of the present invention is to provide a Plasma Glucose Monitoring System unlike the standard glucometers, does not require any complex instrument.
Yet another objective of the present invention is to provide a Plasma Glucose Monitoring System backed with machine learning algorithms to predict glucose levels based on image data captured from glucose strips through a combination of image pre-processing techniques, camera and trained models, the application accurately predicts glucose concentration, providing users with real-time and reliable information about their glucose levels.

SUMMARY OF THE INVENTION:
Embodiments of the present disclosure present technological improvements as solution to one or more of the above-mentioned technical problems recognized by the inventor in conventional practices and existing state of the art.
Accordingly, present invention is a Plasma Glucose Monitoring System and method thereof. The present invention is based on the paper test strips and programmed application installed on any camera based smart device for colorimetric detection of the concentration of glucose present in the plasma. This application is integrated with machine learning algorithms for more accurate result.
The point-of-care diagnostic system of present invention uses any optical lens/camera-based device/ smartphone camera for capturing the image/video of the test strip after reaction. The aperture of a camera affects the amount of light that enters the sensor. In different environments, lighting conditions can vary significantly, which can result in variations in image quality and exposure. Present invention overcome this challenge by adding a filter while capturing the raw image/video. By using a filter, one can maintain consistent illumination across different phone devices, which helps in comparing the number of images captured in the image/video. This ensures that the lighting conditions are standardized, regardless of the camera's aperture.
In one example, the system and method of the present invention uses mobile application/software integrated with Machine Learning model(herein after referred as ML model) for image processing and prediction, which is used for the estimation/monitoring of plasma glucose level. The ML model accurately predicts glucose concentrations, providing users with real-time and reliable information about their glucose levels. This enables proactive health management and empowers users to make informed decisions about their lifestyle and diet.
The plasma glucose level is a significant input for assessing the severity of the diabetes. It is an important pathological factor which can be used to understand the metastasis of the diabetics. This in turn serves as an input to decide the treatment of the affected patients. Effectively, it improves the diagnosis in diabetic patients.
Estimating and monitoring plasma glucose levels using test strips and smartphone-based application provides an alternate method which does not involve complexities, patient-discomfort, and cost involved in lab tests and manual and/or computational resources required in analyzing images and other problems in the detection kits.
Many smartphones come with a built-in camera feature that automatically enhance and beautify the captured images. These features may introduce artificial changes to the image, altering its true colours and details. Present invention overcome this challenge by applying post-processing techniques by extracting the true colours of the Region of Interest (ROI), which are similar to the ground truth colour of the image captured from different phones, the images then can be restored to their original state. This can be achieved by equalizing the histogram of the RGB colour space, which helps in preserving the original colour information.
Different smartphones have varying zoom capabilities, ranging from optical zoom to digital zoom. This difference in zoom levels can affect the composition and level of detail captured in the images. To overcome this challenge, it is essential to consider the zoom levels while comparing images taken from different phones. This can be achieved by either capturing images at a standardized zoom level or by adjusting the image composition during post-processing. The present invention ensure that the zoom levels are consistent across all images, a fair comparison can be made between different phone cameras.
The approach for optical accessory free Smartphone camera based colorimetric sensing needs to be data-driven because of the non-invertibility nature of the in-camera pipeline (ICP) data processing. The spatial image information as well as temporal information could be useful to reduce the unwanted variations in pixel intensities due to the camera-specific and ambient environment-imposed noises. Some specific image degradation operations due to shadow, flickering of light, wrong white balance and colour constancy need to be handled algorithmically in the real-time. The information encoded in multiple images taken in different camera parameters like ISO, white balance etc. may thus be useful to arrive at a generic technology via a suitable Neural Network based training methodology, so as to mask the variations due to non-uniform illumination or erroneous camera settings. Considering the same, the present invention adapts a hybrid approach of combining different Neural Network models for unsupervised learning of the task-specific deep image features that are finally optimized for the particular analyte concentration estimation for a new test set. The present invention model can also handle the noise and other variations present in the dataset due to a plethora of physiological and other factors, impossible to decode otherwise, in an automated framework without requiring manual intervention and interpretation. Moving further forward, it enables the detection technology to learn the inherent variations present in the dataset without any statistical bias. ML model in accordance with present Invention uses a cross-validation approach for the latter, and develop a framework where progressively added data are automatically incorporated for dynamic adaptation in the predictive methodology, so that in the limit as more and more data are fed and analysed by the system, the predictions would approach the gold standard laboratory benchmarks with a statistically acceptable confidence limit.
Yet another inventive outcome of the technology is the establishment of an ultra-low-cost framework as the first of its kind for blood plasma based biochemical analysis by strictly adhering to the WHO set RE-ASSURED criteria (Real Time – Ease of specimen collection - Affordable - Sensitive- Specific- User Friendly- Rapid - Equipment Free and Deliverable to end users) for Point-of-care testing. To this end, no other reported technologies for the same purpose could strictly adhere to all the requirements in tandem. In contrast, the present invention integrates all the features of easy sample collection (finger-pricked blood directly absorbed onto a paper strip without needing any dispensing or metering accessory, highly sensitive and specific because of in-situ plasma separation and its analysis by the established benchmarked gold standard in a miniature format and interlacing data science with the optical signal analysis to an extent that highly accurate quantitative results are deliverable to the end user in a complete equipment free format in real time within a few minutes of the sample collection). The user does not need any special skill except for pricking the finger akin to any other standard at-home blood sugar tests and absorbing a drop from the same onto a paper strip. The rest is taken care of by the fully automated albeit device-free analytic procedure that does not require any specialized skill or training of the user.
The main embodiments of the present invention are the test strip and the camera application. Test strip is already lyophilized with a mix of three reagents, Glucose Oxidase (GOD), Peroxidase (POD) and Potassium iodide (KI), in a proportionate recipe. The user has to add drop of blood sample on test strip. Plasma gets separated in-situ from the whole blood by the action of polyvinyl alcohol bound glass based Whatman® LF1 membrane paper. The separated plasma then flows in a microchannel printed in the paper strip in a pump-free manner due to surface tension and accumulates at that site of the strip. where the reaction with pre-loaded reagents occurs. The reaction between the blood plasma and the reagents generates a colour signal, which may be captured by the camera (with no particular specification subject to adherence of a minimum configuration of 1 mega pixel that is covered by all the Smartphones available commercially) and analysed by the application developed to give a readout of the test results within less than 3 minutes of sample collection in any intended format, including optional features of data encryption and cloud integration with no additional cost penalty.
The ML model in accordance with present invention covers various stages, including data collection, pre-processing, model development, training, evaluation, and discussions of the results. By systematically exploring different algorithms and techniques, present invention aims to enhance the accuracy and reliability of glucose level prediction, ultimately benefiting healthcare professionals and individuals in managing and monitoring glucose levels effectively.
Training of ML model in accordance with present invention:
Region of interest (ROI) detection from the real-time camera captured colorimetric digital images is a really challenging task and needs either supervised or unsupervised algorithmic implementations. Here, a very simple test strip with reagent, in which even any unskilled person can identify the colourful reaction pad, blood drop region and the rest test strip very easily. Thus, the specially designed strip is useful for easy annotation of the ROI and quick pre-processing for the supervised Machine Learning (ML) algorithms. Usually, the state-of-the-art ROI segmentation techniques can be divided into two approaches; a) automatic ROI thresholding-based segmentation and b) supervised pre-annotated foreground mask and background based supervised classification techniques. The automatic ROI thresholding methods are based on some a priori information on the foreground and the background region and are unsupervised in nature. On the other hand, the supervised Machine Learning (ML) methods need background and foreground mask pair as the training set of the algorithm which can extract the ROI of any test sample.
The unsupervised ROI segmentation techniques suffer from poor generalization accuracy of the algorithm for the real-time images, poor performance in case of varying illumination and camera properties compared to the supervised techniques. Consequently, annotations in supervised segmentation techniques are labour intensive, costly and time consuming but have the potential for producing highly accurate region of interest in real-time scenario, which is the most important pre-processing part of the complete prediction algorithm. Hence, present invention uses some Object detection-based identification of different segments like blood drop, colourful ROI and the strip itself. Additionally, from the detected ROI region, for final fine-tuning of the ROI, the a priori knowledge of the colour features like average hue and saturation of the colorimetric response of the available range of analyte values are utilized. However, this object detection and thresholding-based technique is used for automatically generating foreground and background masks for the training set of the applied supervised segmentation technique. Additionally, some algorithmic work for exposure, white balance and ISO adjustment have been utilized to further mitigate the impact of environmental changes. Moreover, the complete image segmentation and pre-processing algorithms are deployed in an offline manner in Smartphones to overcome the requirement of internet and other computing facilities.
The image analytic module includes the step of extracting the ROI and resize/crop the image using a series of end-to-end deep learning models designed for fast object detection, via predicting a class of an object and the bounding box, containing the desired colorimetric feature as an outcome of the on-strip reaction chemistry, which is defined by its center, width and height. The output prediction conforms to a number that is unique to the predicted class, along with the corresponding probability. This approach is adapted in preference to the other common variants of the neural network models, because of obviating the needs of a large amount of annotated training data that would otherwise require a continued access to high-performance computing during the dynamic training phase which is prohibitive for applications developed to work for low-resource settings. The bounding box identifier number, when formalized as a function, includes arguments as the probability of the said object class, the coordinates of the centre of the bounding box relative to the cell, width and height of the bounding box relative to the entire image, and a binary identifier that delineates the class represented in the said bounding box. If there are total N arguments of the function, the same creates an N-dimensional vector space. If there are M such bounding boxes constituting the whole image, then the entire image gets represented with M vectors of size N, alternatively 3×3×N tensor. This enables a computationally efficient yet minimally data-hungry training and testing cohort for the convolutional network. Generalizing, if each cell is described using a vector with dimension B × 5 + C where B is the number of bounding boxes, and C is the number of classes. If the image is divided into R × R grid then its representation would thus require R × R × (B × 5 + C) tensor. In case of highly overlapping colorimetric features inevitable for on-field applications, several bounding boxes could potentially map to one class, for which the algorithm is trained to select only one box per class conforming to the highest probability. If there are multiple objects of one class in the image, the same may be addressed by first taking the box with the maximum probability. After that, this box is compared with all other boxes of the same class using an index that is the ratio of the intersection and the union of the two bounding boxes under consideration. In the event that this index is greater than a predefined threshold probability value, the box with lower probability is excluded. This prediction is repeated in a loop until all the boxes are assessed for inclusion or exclusion. These measures altogether ensure the finalization of an optimal ROI in an automated manner from an image captured without any special control. Subsequently, a CNN architecture is applied on this ROI that can intrinsically support several thousands of convolutional layers, which is a significant upscaling of the previously used CNNs for analogous purposes. The reported methodologies for colorimetric image analytics from biochemical reactions are mostly undertaken with machine learning features alone with only restrictive capabilities of their deep learning frameworks. This deficit stemmed from the fact that their neural networks were mostly trained through a backpropagation process pivoting on the gradient descent. In case of too many layers, repeated multiplicative effect in the loss function tended to attenuate the gradient to an extent that it tended to disappear altogether on adding a threshold number of layers, thereby restricting the training bandwidth. The present invention could overcome this constraint by stacking multiple identity mappings and skipping those layers that do nothing at first, thereby speeding up the initial training by compressing the network into fewer numbers of layers. Subsequently, a re-training of the network ensures that all layers get expanded and the residual (remaining) parts of the network that were previously skipped are also explored to exploit more of the feature space of the input image, enhancing the sensitivity of the detection. The method is further trained to learn and implement dynamically the requisite number of layers to skip, thereby optimizing the detection accuracy further.
Further various features and advantages of the present invention will become apparent from the detailed description provided herein. This detailed description is intended for the purpose of illustration only and not intended to limit the scope of the invention.

BRIEF DESCRIPTION OF THE DRAWINGS
The embodiments can be better understood with reference to the following drawings and descriptions. The components in the figures are not necessarily to scale, the emphasis instead being placed upon illustrating the principles of the embodiments. Moreover, the figures, like reference numerals designate corresponding parts throughout the different views.
Reference will be made to embodiments of the invention, examples of which may be illustrated in the accompanying figures. These figures are intended to be illustrative, not limiting. Although the invention is generally described in the context of these embodiments at various places, it should be understood that it is not intended to limit the scope of the invention to these particular embodiments.
The above and other objects, features, and advantages of the present invention will be more clearly understood from the following detailed description taken in conjunction with the accompanying drawings, in which:
FIG. 1 Block diagram of an exemplary system in accordance with the present invention
FIG. 2A. Structure of the test strip
FIG. 2B Test Strip after reaction with reagent
FIG. 3 Exemplary process flow diagram of the programmed application installed on Smart Device(102)

Fig. 4 Flow diagram of an exemplary process in accordance with present invention

DETAILED DESCRIPTION OF THE INVENTION
Embodiments of the present disclosure is described herein. It is to be understood, however, that the disclosed embodiments are merely examples and other embodiments can take various and alternative forms. The figures are not necessarily to scale; some features could be exaggerated or minimized to show details of particular components. Therefore, specific structural and functional details disclosed herein are not to be interpreted as limiting, but merely as a representative basis for teaching one skilled in the art to variously employ the embodiments.
As those of ordinary skill in the art will understand, various features illustrated and described with reference to any one of the figures can be combined with features illustrated in one or more other figures to produce embodiments that are not explicitly illustrated or described. The combinations of features illustrated provide representative embodiments for typical applications. Various combinations and modifications of the features consistent with the teachings of this disclosure, however, could be desired for particular applications or implementations.
In the following description, for the purpose of explanation, specific details are set forth in order to provide an understanding of the present invention. It will be apparent, however, to one skilled in the art that the present invention may be practiced without these details. One skilled in the art will recognize that embodiments of the present invention, some of which are described below, may be incorporated into a number of systems.
The various embodiments of the present invention provide a Plasma Glucose Monitoring System (PGMS) and method that uses paper test strips and a mobile application backed by machine learning algorithm for digital image colorimetric detection of the concentration of glucose present in the plasma. Therefore, the disclosed system provides an accurate glucose concentration, providing users with real-time and reliable information about their glucose level.
Furthermore, connections between components and/or modules within the figures are not intended to be limited to direct connections. Rather, these components and modules may be modified, re-formatted, or otherwise changed by intermediary components and modules.
The systems/devices and methods described herein are explained using examples with specific details for better understanding. However, the disclosed embodiments can be worked on by a person skilled in the art without the use of these specific details.
Throughout this application, with respect to all reasonable derivatives of such terms, and unless otherwise specified (and/or unless the particular context clearly dictates otherwise), each usage of
“a” or “an” is meant to read as “at least one.”
“the” is meant to be read as “the at least one.”
References in the present invention to “one embodiment” or “an embodiment” mean that a particular feature, structure, characteristic, or function described in connection with the embodiment is included in at least one embodiment of the invention. The appearances of the phrase “in one embodiment” in various places in the specification are not necessarily all referring to the same embodiment.
Embodiments of the present invention include various steps, which will be described below. The steps may be performed by hardware components or may be embodied in machine-executable instructions, which may be used to cause a general-purpose or special-purpose processor programmed with the instructions to perform the steps. Alternatively, steps may be performed by a combination of hardware, software, firmware and/or human operators.
If the specification states a component or feature "may", "can", "could", or "might" be included or have a characteristic, that particular component or feature is not required to be included or have the characteristic.
As used in the description herein and throughout the claims that follow, the meaning of "a'', ''an," and "the" includes plural reference unless the context clearly dictates otherwise. Also, as used in the description herein, the meaning of "in" includes "in" and "on" unless the context clearly dictates otherwise.
Exemplary embodiments will now be described more fully hereinafter with reference to the accompanying drawings, in which exemplary embodiments are shown. This invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. These embodiments are provided so that this invention will be thorough and complete and will fully convey the scope of the invention to those of ordinary skill in the art. Moreover, all statements herein reciting embodiments of the invention, as well as specific examples thereof, are intended to encompass both structural and functional equivalents thereof. Additionally, it is intended that such equivalents include both currently known equivalents as well as equivalents developed in the future (i.e., any elements developed that perform the same function, regardless of structure).
While embodiments of the present invention have been illustrated and described, it will be clear that the invention is not limited to these embodiments only. Numerous modifications, changes, variations, substitutions, and equivalents will be apparent to those skilled in the art, without departing from the spirit and scope of the invention, as described in the claim.
Table-1 The reference numerals used in Figures:
Ref. Numeral Particulars
101 Test Strip
102 Smart Device/Smart Phone
103 Glucose Reading/Output
104 Filter Paper
105 Reaction Pad
106 Machine Learning Model
107 Image Capturing Means/Camera
108 Image/Video of Test Strip after reaction with reagent
109 QR Code/Bar Code
110 Mobile Application
111 Image Processing Tool
112 User Registration
113 User Authentication
114 Test History/Report

As illustrated in Fig.1 the System for Plasma Glucose Monitoring is comprising of Test strips(101), Smart Device(102) having programmed application(110) integrated with Machine Learning model(herein after referred as ML model)(106) and image capturing means/camera(107) etc. The Test Strip(101) used herein is any disposable test platform suitable for digital image colorimetric test. In an exemplary embodiment, the material used for test strip(101) is selected from: PVC (90 microns to 430 micron), PP (Polypropylene) (90 microns to 430 micron), PET, PETG, Hips, Paper (coated, uncoated and craft - 210gsm to 400 gsm) or the like. In preferred form microfluidic paper based test strip is used in the present invention.
As shown in Fig. 2A and 2B, the test strip(101) is provided with QR Code/Bar Code(109), Filter Paper(104) and Reaction Pad(105). The QR Code(109) is used to identify Kit, Batch No. & Shelf Life. The Filter Paper(104) and Reaction Pad(105) are sticked with adhesive on Test Strip(101) as shown in Fig. 2A. The said Filter Paper(104) is used for Plasma separation from Whole Blood; in preferred from Whatman® LF1 Paper is used in present invention. In an exemplary embodiment the present invention may use different Plasma Separation techniques, which are listed below:
1. Vivid Membrane® by Pall Corporation
a. Separation by size exclusion and capillary action
b. Poly(ethyl sulfone) membrane
2. Blood Separators by GE
a. Fusion 5
b. GF/DVA - bound glass fiber filter
c. LF1 - bound glass fiber
d. MF1 - bound glass fiber
e. VF2 - Bound glass fiber
Present invention may use Whatman® Grade 1 Cellulose Paper treated with Glucose Oxidase (GOD), Peroxidase (POD) and Potassium iodide (KI) as Reaction Pad(105) for Plasma Glucose Reaction. Present invention uses stabilized reagents at room temperature with shelf life of around 2 years.
The Smart Device(102) used herein may be a mobile apparatus that is capable of running a programmed application(110) for executing instructions in accordance with process of present invention. Further this programmed application(110) is integrated with ML model(106) and image capturing means(107). The image capturing means(107) may be an inbuilt mobile/smart phone camera or any optical lens/camera-based device which can be further connected to smart device(102) or is embedded with said mobile application(110). In an exemplary embodiment, the Smart Device(102) may include any smart computing device that could be made ‘portable’ or ‘mobile’; any digital mobile devices such as smartphones, tablets, phablets, smart watches, and other current or future ‘smartphone’ platforms having similar minimal functionality. (hereinafter the terms ‘smartphone’ and ‘smart device’ are used interchangeably). Further said programmed application(110) eliminates need of separate controlled environment/imaging box/viewing box/3D enclosure or any such external device while capturing the image of test strip(101) after reaction. The application(110) has standardized the result accuracy for all exposure and varied light intensities. The image capturing is thus carried out in any types of light luminosity i.e. daylight, yellow light, white light etc.
The Test strip(101) is already lyophilized with reagents and is capable of receiving a drop of blood sample. After blood sample is added by the user on test strip, the Plasma gets separated from whole blood by polyvinyl alcohol bound glass based Whatman® LF1 membrane paper(104), then glucose present in the plasma gets reacted with reagents present on reaction pad(105) of the strip(101). Based on concentration of glucose in plasma, the said reaction causes change in colour of the reaction pad(105). With the image/video capturing means(107) available in Smartphone(102) the user then captures the image/video(108) of test strip(101). The image/video pre-processing tool(111) & processing technique of the model(106) extracts Region of Interest (ROI) for further analysis. Machine learning algorithm & analytical tool then predicts a glucose value(103) based on analysed ROI. The smart application(110) then displays the predicted Glucose level(103) on mobile display. The predicted Glucose level(103) is also saved for future reference.
As illustrated in Fig. 3, in another embodiment of the present invention, the programmed application(110) installed on Smart Device(102) includes a registration process for accessing the information through the system. This registration process(112) includes making of application, profile compilation. The registration/enrolment has to be done only at initial/entry level after downloading the app.
According to another embodiment of the present invention, system and method of the present invention includes authentication/confirmation process. The authentication process(113) starts after registration process. During authentication, system will generate unique login id and password for each user and will send them to their email address. System will also send a welcome email along with the detailed instructions about accessing the system. User can access the mobile application(110) through this login credentials generated in registration and authentication step. User can perform new test glucose test or can access his/her previous test history(114) through this mobile application.
As illustrated in Fig. 4, an exemplary process in accordance with present invention comprises following steps:
1. Providing a test strip(101) and Smart Device(102) having programmed application(110) integrated with ML model(106) and image capturing means/camera(107) to the user;
2. User log-in and selecting the test option;
3. Pricking a finger tip of the user for drop of blood sample;
4. Receiving a drop of blood sample on test strip(101) which is pre-embedded with GOD (Glucose Oxidase), POD (Per Oxidase) and Potassium Iodide (KI) on reaction pad(105);
5. Separating the plasma form blood by capillary action and filter paper(104) then reaction of glucose with reagents of reaction pad(105);
6. Capturing the image/video(108) of test strip(101) after reaction with the help of image/video capturing means(107) by user;
7. Using image processing tool(111) for extracting the Region of Interest (ROI) and resizing/cropping the image as per requirement.
8. Deploying and integrating the ML model (106) with identified ROI as an input.
9. Utilizing advanced, deep machine learning techniques for digital image colorimetric detection of the concentration of glucose present in the plasma to make accurate predictions.
10. Displaying/Reporting the estimated plasma Glucose level (103) to the user on application (110);
11. Generating report. The system can announce estimated plasma Glucose level(103) in voice message also.

The advantages provided by the present invention are:
• It separates plasma from whole blood and then measures glucose thereby increases the accuracy of result.
• It has eliminated the need of separate controlled environment/imaging box/viewing box/3D enclosure or any such external device while capturing the image/video of the test strip.
• It has standardized the result accuracy for all exposure and varied light intensities. The result will be the same across all types of light i.e. daylight, yellow light, white light etc.
• It is asset light, making product’s usability easy, user friendly & impacting cost substantially.
• It will be giving results with similar accuracy across any smartphone.
• Its design has been optimized resulting in quick reaction time of ~ 2 to 3 minutes and with less to no chances of blood spilling.
• It has a stabilized reagent at room temperature with shelf life of ~2 years.
• It has an improvised strip design, leading to ease of use for patients. The design solved the issues to be able to take an apt amount of blood sample for reaction to take place.
The system utilizes machine learning algorithms to predict glucose levels based on image data captured from glucose strips. Through a combination of image pre-processing techniques and trained models, the application predicts glucose concentrations using digital image colorimetric detection technique, providing users with real-time glucose levels.
This elusively simple paper-kit, unlike the standard glucometers, does not require any instrument. This game changer technology infuses the complexity of microfluidic separation, capillary transport and analyte detection in a data-science empowered digital image colorimetric analytics format that works universally across all the smartphones without any quality compromise. There is no global parallel of such technology having unique amalgamation of power-free sample analysis in an ultra-low-cost disposable kit, and quantitative readout cum digital image colorimetric image analytics in a device-agnostic framework.

EXAMPLE:
Example 1 – A person diagnoses with hypoglycaemia (a condition characterised by low blood sugar level). As per doctor’s advice, to monitor his blood glucose level and manage his conditions, he started using regular Glucometer. He notices that every time there is lot of variation in the glucose reading when blood is tested in different devices at same time from same sample. Also, there is substantial difference in glucometer result and laboratory result even when same sample is used. This variation is there as plasma is not separated from whole blood which itself is unstable in nature and gives factually incorrect results. He when uses the present invention and noticed increased accuracy and consistency in test result as present invention separates plasma from blood and then measures the glucose concentration.
Example 2 – A type 2 diabetic person who is on insulin has to measures the Glucose level twice a day. When travelling to other city many a times she forgot her glucometer device at home and every time she has to buy new device. Post using the present invention, she is very happy as here there is no need of separate device. She just needs strip and mobile hence it is very simple to use and asset light. She now can check her glucose level anytime anywhere. Also, all the past reading data is available in application so that she can check the trend, improvements etc.
Example 3 – A type 2 diabetic person who is on insulin has to measures the Glucose level twice a day. He needs a Glucometer and strips which is very expensive. Post using the present invention, he is very happy as here there is no need of separate device. Also, price of Strips is very low comparing to current strips available in market. Hence using present invention is very economical and save lot of money. , C , Claims:CLAIMS:
We Claim:
1. A Plasma Glucose Monitoring System (PGMS) which measures blood plasma glucose using paper-based strip, image capturing means integrated with programmed application and machine learning model, said system consisting of:
• Test Strip(101) with QR Code/Bar Code(109) and having Filter Paper(104) and Reaction Pad(105)
• Image Capturing Means/Camera(107)
• Smart Device/Smart Phone(102)
• Machine Learning Model(106)
• Image/Video of Test Strip after reaction with reagent(108)
• Programmed Application(110)
• And Image Processing Tool(111)
characterized by the programmed application (110) is integrated with ML model(106) and image capturing means(107) and it eliminates need of separate controlled environment/imaging box/viewing box/3D enclosure or any such external device while capturing the image/video of test strip(101) after reaction.

2. A method of Plasma Glucose Monitoring using paper-based strip, image capturing means integrated with programmed application such as smartphone and machine learning model comprising steps of:
a. Providing a Test strip(101) and image capturing means like camera (107) and smart device like smartphone (102), to the user;
b. User log-in and selecting the test option;
c. Pricking a finger tip of the user for drop of blood sample;
d. Receiving a drop of blood sample on test strip(101) which is pre-embedded with GOD (Glucose Oxidase), POD (Per Oxidase) and Potassium Iodide (KI) on reaction pad(105);
e. Separating the plasma form blood by capillary action and filter paper(104) then reaction of glucose with reagents on reaction pad(105);
f. Capturing the image/video(108) of test strip(101) after reaction with the help of image capturing means(107) by user;
g. Using image processing tool(111) for extracting the Region of Interest (ROI) and resizing/cropping the image;
h. Deploying and integrating the ML model(106) with identified ROI as an input;
i. Utilizing advanced machine learning techniques for digital image colorimetric detection of the concentration of glucose present in the plasma to make accurate predictions;
j. Displaying/Reporting the estimated plasma Glucose level(103) to the user on application(110);

3. A method of Plasma Glucose Monitoring as claimed in claim-2 wherein the Test Strip(101)
a. used is a microfluidic paper-based test strip suitable for digital image colorimetry test and is capable of receiving a drop of blood sample;
b. is provided with QR Code/Bar Code(109), Filter Paper(104) and Reaction Pad(105);
c. and the material used for test strip(101) is selected from: PVC (90 microns to 430 micron), PP (Polypropylene) (90 microns to 430 micron), PET, PETG, Hips, Paper (coated, uncoated and craft - 210gsm to 400 gsm) etc.

4. A method of Plasma Glucose Monitoring as claimed in claim-2 wherein the QR Code/Bar Code(109) is used to identify Kit, Batch No. & Shelf Life

5. A method of Plasma Glucose Monitoring as claimed in claim-2 wherein the Filter Paper(104) is used for Plasma separation from Whole Blood is Whatman® LF1Paper.

6. A method of Plasma Glucose Monitoring as claimed in claim-2 wherein the Plasma Separation technique is selected from:
1. Vivid Membrane® by Pall Corporation
a. Separation by size exclusion and capillary action
b. Poly(ethyl sulfone) membrane
2. Blood Separators by GE
a. Fusion 5
b. GF/DVA - bound glass fiber filter
c. LF1 - bound glass fiber
d. MF1 - bound glass fiber
e. VF2 - Bound glass fiber

7. A method of Plasma Glucose Monitoring as claimed in claim-2 wherein
a. The Reaction Pad(105) is treated with reagents which are stabilized at room temperature with shelf life of up to 2 years;
b. The reagents used herein are Glucose Oxidase (GOD), Peroxidase (POD) and Potassium iodide (KI).

8. A method of Plasma Glucose Monitoring as claimed in claim-2 wherein the Reaction Pad(105) used is selected from Whatman® Grade 1 Cellulose Paper.

9. A method of Plasma Glucose Monitoring as claimed in claim-2 wherein the Smart Device(102) used is a mobile apparatus that is capable of running a programmed application(110) for executing instructions in accordance with process of present invention.

10. A method of Plasma Glucose Monitoring as claimed in claim-2 wherein the Smart Device(102) used is selected from smartphones, tablets, phablets, smart watches etc.

11. A method of Plasma Glucose Monitoring as claimed in claim-2 wherein the programmed application(110)
a. is executed on smart device(102) and is integrated with ML model(106) and image capturing means(107)
b. and it eliminates need of separate controlled environment/imaging box/viewing box/3D enclosure or any such external device while capturing the image of test strip(101) after reaction.

12. A method of Plasma Glucose Monitoring as claimed in claim-1 wherein User can access the mobile application(110) through this login credentials generated in registration and authentication step and selects options such as: i. perform new glucose test and ii. access his/her previous test history(114) through the mobile application.

13. A method of Plasma Glucose Monitoring as claimed in claim-1,
a. wherein the image capturing means(107) is selected from i. an inbuilt mobile/smart phone camera, ii. optical lens/camera-based device which can be further connected to smart device(102);
b. and wherein the image capturing is carried out in following types of light luminosity: daylight, yellow light, white light etc.

14. A method of Plasma Glucose Monitoring as claimed in claim-1 wherein the Machine Learning Model (106) is consisting of identifying Region of Interest and then further processing it.

15. A method of Plasma Glucose Monitoring as claimed in claim-1 wherein an image processing tool(111) is used for analytics and then for giving the result.

16. A method of Plasma Glucose Monitoring as claimed in claim-2 wherein the estimated plasma Glucose level(103) is announce through voice message.

17. A method of Plasma Glucose Monitoring as claimed in claim-2 wherein complete image segmentation and pre-processing algorithms used in image processing tool (111) is deployed in an offline manner in Smart device(102) to overcome the requirement of internet and other computing facilities.

Dated this 27th day of July 2023

Signature:
Name: Mrs. Rashmi Ganesh Hingmire
Patent Agent Code - IN/PA/1844