Description:Complete Specification:

The following specification describes and ascertains the nature of this invention and the manner in which it is to be performed.
Field of the invention
[0001] The invention is related to a control unit for estimating the value of a blood parameter in a bio-analyte device and a method thereof.

Background of the invention

[0002] An accurate monitoring of bioanalyte levels in patients is vital to a patient's health. Monitoring of glucose in blood sample of diabetic patients, for example, is valuable in order to prevent blindness, kidney diseases, necrosis of nerve tissue, as well.
as other complications. On a broader basis, it has become increasingly important in analytical and clinical chemistry to have the capability of remote (or noninvasive) sensing of chemical and physical parameters. Some methods of performing this type of
sensing is known in the state of the art, such as potentiometry, amperometry, Pulse Oximeters, Continuous Hb monitors. In addition to these methods, optical techniques can
be used for remote sensing of analytics and other substances.

[0003] A US patent application 20220188701 discloses an interpretation of machine learning classifications in clinical diagnostics using shapely values and uses thereof. The interpretation discloses that shapely values (SVs) have become an important tool to further the goal of explainability of machine learning (ML) models. However, the computational load of exact SV calculations increases exponentially with the number of attributes. Hence, the calculation of SVs for models incorporating large numbers of interpretable attributes is problematic. Molecular diagnostic tests typically seek to leverage information from hundreds or thousands of attributes, often using training sets with fewer instances. Methods are described for evaluate SVs using Monte Carlo sampling or exact calculation in polynomial time (i.e., reasonably quickly and efficiently) using the architecture of a ML model designed for robust molecular test generation, and without requiring classifier retraining.

Brief description of the accompanying drawings
[0004] Figure 1 illustrates a control unit for estimating the value of a blood parameter in a bio-analyte device according to one embodiment of the invention; and
[0005] Figure 2 illustrates a flowchart of a method of estimating the value of the blood parameter in the bio-analyte device according to the present invention.

Detailed description of the embodiments
[0006] Figure 1 illustrates a control unit for estimating the value of a blood parameter in a bio-analyte device according to one embodiment of the invention. The control unit 10 extracts multiple features from at least one light intensity generated in said bio-analyte device 12 wherein each of the light intensity value is referred as a data point. The control unit 10 builds and trains an intelligence model 14 based on the extracted multiple features and data points. The control unit 10 identifies plurality of combinations of the extracted features and record a corresponding impact on an output of the intelligence model 14. The control unit 10 estimates the value of the blood parameter in the bio-analyte device 12 based on the recorded corresponding impact on the output of the intelligence model 14.

[0007] Further the construction of the bio-analyte device and the control unit and the working modes of both are explained in detail. The bio-analyte device 12 is a medical device that is used to measure/determine multiple body parameters using a body appendage exposed to a light source. According to one embodiment of the invention, the body parameter is a Hemoglobin value present in the body of the human being using the device and the bio-analyte device 12 is a Hemoglobin monitoring device. The bio-analyte device 12 comprises four light sources 20 and each light source 20 produces a light intensity value when the light from the light source 20 is made to pass through the body appendage of the human being. However, it is to be understood, that, the body parameters can be any other parameters that is known to a person skilled in the art.

[0008] The control unit 10 is chosen from a group of control units comprising a microcontroller, a microprocessor, a digital circuit, an integrated chip and the like. According to one embodiment of the invention, the control unit 10 is integrated in the device. In another embodiment, the control unit 10 is connected via a cloud connection using any one of the communication means known in the state of the art. The control unit 10 extracts multiple features from corresponding light intensities produced during the operating conditions of the bio-analyte device 12. The bio-analyte device 12 comprises four light sources 20, through which corresponding four light intensities are generated . Each light intensity is referred to as a data point. From each of the data point, multiple features are extracted. The control unit 10 stores all these data (multiple data points and corresponding multiple features) for building and developing an intelligence model 14.

[0009] The control unit 10 an estimating module adapted to add/remove the extracted features for recording the output of the intelligence model 14. The intelligence module 14 uses any one of the neural networks that are used to build and train the models 14 and on the other hand, the intelligence model 14 can be any one of the intelligence model 14 comprising an artificial intelligence model, a deep learning model and a machine learning model and the like. The estimating module 16 determines an average marginal contribution value of each feature of each of the data point, on the model output and adapted to assign a credit value to each feature to detect an impact on the model output. The estimating module 16 of the control unit 10 uses multiple techniques to determine the body parameter of the human being. The control unit 10 comprises a memory 18 to store data related to the intelligence model, the data comprises an average marginal contribution value of each feature, a noise value that has been added, original values of each of the feature.

[0010] Figure 2 illustrates a flowchart of a method of estimating the value of the blood parameter in the bio-analyte device 10 according to the present invention. In step S1, multiple features from at least one light intensity generated in the bio-analyte device 12 are extracted, wherein each of the light intensity value is referred as a data point. In step S2, an intelligence model 14 is built and trained based on the extracted multiple features and data points.

[0011] In step S3,plurality of combinations of the extracted features are identified and a corresponding impact on an output of the intelligence model 14 is recorded. In step S4, the value of the blood parameter in the bio-analyte device 10 is estimated based on the recorded corresponding impact on the output of the intelligence model 14.

[0012] The method of estimating the value of the blood parameter in the bio-analyte device 10 is explained in detail. The bio-analyte device 10 comprises four light sources 20 , wherein each light source 20 produces a corresponding light intensity value during the operation of the device 10. Each light intensity value is considered as one data point. And from each of the data point multiple features are extracted. The information comprising multiple data points and the corresponding multiple extracted features are stored in the control unit 10 memory 18 for building and training the intelligence model 14.

[0013] The intelligence model 14 in communication with the estimating module 16, estimates the body parameter in the bio-analyte device 12 using at least one estimating technique. In one scenario, the estimating module 16 uses the multiple data points and the associated multiple features of the device 12. The estimating module 16 adds/removes the multiple features of the data points for identifying a possible combination of the features and records the change in the output of the intelligence model 14. The estimating module 16 considers all the stored data points and corresponding extracted features of the data points. The control unit 10 records the change in the output when the features are combined in plurality of ways.

[0014] The estimating module 16 of the control unit 10 determines an average marginal contribution value of each feature on the output (which is the estimated Hb value) and the same is stored in the memory 18 of the control unit 10. The control unit 10 associates each feature with a value for assigning a credit number. The higher assigned value indicates a higher impact on the output of the model/ device 14/12.

[0015] With this kind of methodology, one can identify the most affecting parameter and the modification required for obtaining the required output of the intelligence model 14. The combination of the various extracted features helps the control unit 10 to understand the high impacting features and the variations that are required to obtain a near output value in the intelligence model 14.

[0016] In another scenario, the estimating module 16 uses another methodology in estimating the Hb value in the bio-analyte device 12. When this methodology is employed in the control unit the estimating module 16 approximates the behavior of a complex model in a local neighborhood around a specific prediction, and provides an interpretation that is based on simple, interpretable features. Ie., the features are simple and interpretable. The estimating module 16 uses one data point and the corresponding extracted features of the data point at a time. For the considered data point, the features are perturbed by adding a small amount of random noise by the estimating module 16 of the control unit 10.

[0017] The control unit 10 with every of the altered featured by adding the noise, a prediction associated with the altered featured is recorded for determining the impact on the output of the intelligence model 14. The variation on each of the feature associated with the data point is considered and the variation in the output is determined , such that, the feature that is more impacting can be identified. Thus, identified feature is modified ,such that, the output of the intelligence model 14 which is the estimated Hb value is close to the real-time Hb value during the operating condition of the device 12. The control unit 10 further considers the approximate linear explanation model for computation of the estimated Hb value/output of the intelligence model 14 using the predictions and the coefficients which acts as explainers.

[0018] Yet in another scenario, the estimating module 14 uses another methodology in estimating the Hb value of the bio-analyte device 12. In this process, the estimating module 14 uses one data point and the extracted features associated with the data point. From the stored data, the original value of the each of the extracted feature is compared with the generated output. The outputs here are generated by generating a hypothetical inputs to the model 14. By varying at least one or two features, the desired output is modified.

[0019] The estimating module 16 insights into causal relationships between input features and models output by generating hypothetical inputs resulting in a different output. This helps in identifying minimum changes in the input, that would result in a different output /desired output). The estimating module 16 keeps a set of features fixed and perturb one or more input features to observe the changes in the model’s output. For instance, a set of counterfactual explanations which displays the suggested values for the features (in comparison to the original feature values) that if changed accordingly, will lead to the desired prediction.

[0020] With the above method, the bio-analyte device 12 estimates the total hemoglobin value based on the features or characteristic measures derived from multiple signals (ie., light intensities). These signals contains information about the absorption characteristics of hemoglobin components in the human being blood. The estimation of that hemoglobin is indirect, because of the other underlying interfering compounds apart from the blood like tissues, skin thickness, etc. The correlation of the features with respect to desired/estimated hemoglobin value may be not higher, indicating that it has a non-linear representation between them. The above disclosed methodology deciphers the non-linearity relationship to give the estimated value of hemoglobin. The variation of one feature or multiple features that contributes to the estimated value can be complex.

[0021] A quantitative measure of how these features have impact on the estimated hemoglobin value gives transparency and trust for the clinicians to interpret the results and validate the decisions. This can be achieved by the above disclosed methodology. An intelligence model 14 build and developed, is hosted on a cloud environment. The bio-analyte device 12 having this intelligence model 14, which is capable of data transmission over cloud is passed through the intelligence model 14 that identifies key input parameters influencing the output of the intelligence model 14. The same output will be communicated to the subject through an interface hosted on the device 12. However, the same methodology/above disclosed process can be implemented via a mobile application.

[0022] It should be understood that embodiments explained in the description above are only illustrative and do not limit the scope of this invention. Many such embodiments and other modifications and changes in the embodiment explained in the description are envisaged. The scope of the invention is only limited by the scope of the claims.
, Claims:We Claim:
1. A control unit (10) for estimating the value of a blood parameter in a bio-analyte device (12), said control unit (10) adapted to :
- extract multiple features from at least one light intensity generated in said bio-analyte device (12), wherein each of the light intensity value is referred as a data point;
- build and train an intelligence model (14) based on the extracted multiple features and data points;
- identify plurality of combinations of the extracted features and record a corresponding impact on an output of the intelligence model (14);
- estimate the value of the blood parameter in the bio-analyte device (12) based on the recorded corresponding impact on the output of the intelligence model (14).

2. The control unit (10) as claimed in claim 1, comprises an estimating module (16) adapted to add/remove the extracted features for recording the output of the intelligence model (14).

3. The control unit (10) as claimed in claim 2, wherein the estimating module (16) determines an average marginal contribution value of each feature of each of the data point, on the model output and adapted to assign a credit value to each feature to detect an impact on the model output.

4. The control unit (10) as claimed in claim 1, wherein the estimating module (16) adapted to consider at least one data point and the corresponding extracted features and perturbates the extracted features by adding a noise.

5. The control unit (10) as claimed in claim 4, wherein the altered featured by adding the noise, a prediction associated with the altered featured is recorded by the control unit (10) for determining the impact on the output of the intelligence model (14) .

6. The control unit (10) as claimed in claim 1, wherein the control unit (10) comprise a memory (18) to store data related to the intelligence model (14), the data comprises an average marginal contribution value of each feature, a noise value that has been added, original values of each of the feature.

7. The control unit (10) as claimed in claim 1, wherein the estimating module (16) adapted to compare original values of each of the feature with a corresponding output of the intelligence model (14) based on the generated hypothetical inputs.

8. The control unit (10) as claimed in claim 7, wherein the estimating module (16) varies atleast one extracted feature to modify the corresponding output of the intelligence model (14) and the modified output is compared with the original value of the extracted feature.

9. The control unit (10) as claimed in claim 1, wherein the blood parameter is a hemoglobin value, and the bio-analyte device (12) is a Hemoglobin monitoring device.

10. A method for estimating the value of a blood parameter in a bio-analyte device (12), said method comprising the steps of
- extracting multiple features from at least one light intensity generated in said bio-analyte device (12), wherein each of the light intensity value is referred as a data point;
- building and training an intelligence model (14) based on the extracted multiple features and data points;
- identifying plurality of combinations of the extracted features and record a corresponding impact on an output of the intelligence model (14);
- estimating the value of the blood parameter in the bio-analyte device (12) based on the recorded corresponding impact on the output of the intelligence model (14) .