An artificial intelligence (AI)-based system for assessing liver fibrosis risk in patients and method thereof

Abstract

The present disclosure introduces an AL-based system (100) for assessing liver fibrosis risk in patients with non-alcoholic fatty liver disease (NAFLD). This system integrates seamlessly with the user devices using rest APIs and https for secure communication. Client modules collect and pre-process raw data, ensuring integrity and compatibility. The pre-processed data is sent to the server (104) and forwarded to an input module (112) for authentication and validation. A processing module (116) standardizes the data, where the prediction module (116), utilizing an XGB machine learning model with 85% accuracy, assesses liver fibrosis risk. The clinical pathway module (118) provides personalized protocols, while the risk prediction response module (120) offers risk status and clinical recommendations. Certified by ISO 13485, this system supports clinicians, enhancing decision-making without replacing any diagnostic tests.

Description

4. DESCRIPTION:Field of the invention:

[0001] The present disclosure generally relates to the technical field ofhealthcare and, in specific, relates to an artificial intelligence (Al)-based system and method that assess liver fibrosis risk in patients with non-alcoholic fatty liver disease (NAFLD).Background of the invention:

[0002] Non-Alcoholic fatty liver disease (NAFLD) is a condition where there is an accumulation of excess fat in the liver cells, not caused by alcohol use. It's the most common liver disorder worldwide, affecting about 25% of the global population. NAFLD can progress to nonalcoholic steatohepatitis (NASH), which involves inflammation and liver damage, potentially leading to fibrosis or scarring of the liver. NASH may further progress to cirrhosis, permanent liver damage, and even liver cancer (Ref. Definition & Facts of NAFLD & NASH - NIDDK (nih.gov))

[0003] Non-invasive methods for NAFLD are currently available in the form of clinical signs and symptoms, non-specific laboratory and radiological imaging tests, and combinations of clinical Symptoms and blood test results. Although several of these markers are generally effective for diagnosing a patient with suspected NAFLD, they lack the specificity and sensitivity to differentiate NAFLD from NASH and assess the existence and stage of fibrosis. This is a significant clinical issue because individuals with NASH and fibrosis are likely to require close monitoring and follow-up. To date, liver biopsy, an invasive procedure, is the gold standard for diagnosing steatohepatitis and LiverFibrosis. However, its invasiveness, poor patient acceptability and sampling variability, and impracticality for large-scale screening have posed challenges and fueled the search for non-invasive methods for early diagnosis and staging. As a result, there is a significant need for the development of noninvasive methods that are reliable &accurate for early diagnosis and assessment of steatohepatitis and Liver Fibrosis to facilitate prompt risk stratification and management to prevent disease progression and complications.2

[0004] In existing technology, Non-invasive techniques such as the Fibrosis-4 (FIB-4) Index, NAFLD Fibrosis Score, APRI (AST to Platelet Ratio Index), and Liver Elastography (Fibro Scan) are commonly used in current methodology for assessing liver fibrosis in patients with nonalcoholic fatty liver disease (NAFLD). These tools play an important role in assessing liver damage without requiring invasive procedures. However, they have limitations. For example, they may not always provide an accurate representation of the degree of fibrosis, especially in patients with acute liver injury or inflammation. Certain scores have indeterminate ranges, which can create uncertainty when staging liver fibrosis.

[0005] Furthermore, these scores may not be applicable to all patient demographics, particularly those with concurrent liver diseases. The results may also differ depending on the laboratory performing the tests and the patient's current medical conditions. There is also a risk of overreliance on these scores, which may overshadow the need for a liver biopsy in certain clinical situations. As a result, in current healthcare practices, these scores are commonly used in conjunction with clinical judgment and other diagnostic tests to perform a comprehensive assessment of liver health, with these scores viewed as part of a larger diagnostic framework.

[0006] By addressing all the above-mentioned problems, the study disclosed focuses on the development and validation of an Al-based system. This system is specifically designed to predict the risk of liver fibrosis in patients diagnosed with Non-Alcoholic Fatty Liver Disease (NAFLD). Thus, it provides a solution to assessing liver fibrosis risk in NAFLD patients. This system utilizes a robust machine learning model that integrates clinical features, medication history, and lab reports, achieving a higher level of accuracy compared to traditional methods.

[0007] The disclosed study's design involves building an Al Machine Learning Model using retrospective datasets available at Apollo Hospitals. The goal is to identify advanced fibrosis in NAFLD patients and stratify them into high and low risk for liver fibrosis based on a score. The study is multicentric, prospective, observational, and non-interventional. It includes annual visits for the subjects enrolled and collects data from electronic medical records provided by the site investigator or study coordinator. The data sources include HealthCheck-Up Data, Patient Out-Patient Visit, or In-Patient Visit. The study was conducted in two Apollo Hospitals over a period of 6 months, with a 3-year follow-up.

[0008] The inclusion criteria for the study specify that subjects should be aged between 18- 79 years, willing to provide informed consent, and comply with the study procedure. Subjects who have undergone an elastography scan will also be included. However, subjects aged below 18 or above 79, pregnant female subjects, and subjects with severe debilitating diseases will be excluded.

[0009] The disclosed study involves three groups of participants: Group 1 for model building, Group 2 for validation, and Group 3 for liver biopsy validation. The primary evaluation criterion is building an Al ML Model using the retrospective datasets available at Apollo Hospitals to identify advanced fibrosis in NAFLD patients. Secondary evaluation criteria include stratifying NAFLD patients into high and low risk for liver fibrosis based on the score and determining with better accuracy patients who would require Liver Biopsy as a follow-up.

[0010] The study utilized anonymized data from patients who underwent Elastography at Apollo Gleneagles Hospitals in Kolkata from 2011 to 2018. The initial dataset consisted of over 26,000 patients, 11,597 of which were diagnosed with Non-Alcoholic Steatohepatitis (NASH) or Non-Alcoholic Fatty Liver Disease (NAFLD). After eliminating records with missing data, the final dataset for the model consisted of 5,150 patients. The data was analyzed using the Acoustic Radiation Force Impulse (ARFI) with a cutoff at 1.65.

[0011] The methodology involved grouping and normalizing similar risk factors through Propensity Matching. Further analysis was conducted using the Gradient Boosting Method. Risk factors that did not significantly contribute to the model, as indicated by their p-value, were removed. All patient data was collected with informed consent.For further validation and comparison with the standard Fib4 Score, a prospective validation cohort of 1261 patients were utilized. These patients had undergone elastography at Apollo Gleneagles Hospitals Kolkata between 2018 and 2020. Out of over 3.5K patients, 1760 were diagnosed with NASH / NAFLD. Following the elimination of records with missing data, the finalvalidation dataset was composed of 1261 patients. The model was further validated through 98 liver biopsies, carried out at Apollo Gleneagles Hospitals Kolkata in 2019. Out of over 100 liver biopsies performed, 98 patients were diagnosed with NASH / NAFLD.

[0012] Among the 25 clinical and laboratory parameters examined, 11 variables were found to be significant.The 11 predictors are Age, BMI, History of Diabetes, Albumin, Alkaline Phosphatase, ALT (SGPT), AST (SGOT), Bilirubin, Cholesterol, LDL Cholesterol, and Platelet Count.Notably Age had a Multivariate Odds Ratio (OR) of 3.39 (95%CI 2.99 - 3.84), History of Diabetes Mellitus with an OR of 6.80 (95%CI 5.92 - 7.81), Albumin with an OR of 3.70 (95%CI 3.25 - 4.20), Aspartate aminotransferase (AST) with an OR of 3.65 (95%CI 3.21 - 4.16), Total Bilirubin with an OR of 3.13 (95%CI 2.76 - 3.56), and Platelet Count with an OR of 2.74 (95%CI 2.40 - 3.13 were found to be significant. The performance parameters of the development model are AUC ROC Score of 0.94 and validation cohort had the AUC and accuracy of 0.88. The AUC for 98 liver biopsy validation cohort was 0.83. The model performed better than Fib4 Score with Net Reclassification Improvement (NRI) at 0.499.

[0013] In conclusion, this Al-based system not only addresses the need for a more precise method of predicting liver fibrosis risk in NAFLD patients but also serves as a valuable clinical decision support tool in low-cost settings. By tackling the aforementioned challenges, this system underscores the need for an Al-based solution capable of delivering a comprehensive and holistic risk assessment. Further, it highlights the necessity for a system that can provide personalized recommendations based on the calculated probability of liver fibrosis risk, through an integrated clinical decision support tool.Objectives of the invention:

[0014] The primary objective of the present invention is to devise an Al-driven tool that precisely stratifies the risk of advanced liver fibrosis in NAFLD patients, employing clinical attributes, medication history, and lab records.

[0015] Another objective of the present invention, which is based on Indian data, is to use an Al-predicted score to categorize NAFLD patients into risk groups for liver fibrosis. Furthermore, it offers a Clinical Decision Support System.

[0016] Yet another objective of the present invention is to facilitate the identification and reduction of Liver Fibrosis risk across various healthcare settings, such as preventive gastroenterology and liver disease screening programs, outpatient clinics, and emergency departments.

[0017] Further objective of the present invention is to act as asupplemental tool for physicians, aiding in the comprehensive risk identification for patients, rather than serving as a diagnostic tool for liver fibrosis.Summary of the invention

[0018] The present disclosure proposes an artificial intelligence (Al)-based system for assessing liver fibrosis risk in patients and method thereof. The following presents a simplified summary in order to provide a basic understanding of some aspects of the claimed subject matter. This summary is not an extensive overview. It is not intended to identify key / critical elements or to delineate the scope of the claimed subject matter. Its sole purpose is to present some concepts in a simplified form as a prelude to the more detailed description that is presented later.

[0019] In order to overcome the above deficiencies of the prior art, the present disclosure is to solve the technical problem to provide an artificial intelligence (Al)-based system and method that assesses liver fibrosis risk in patients with nonalcoholic fatty liver disease (NAFLD).

[0020] According to one aspect, the invention provides an artificial intelligence (Al)-based system for assessing liver fibrosis risk in patients with nonalcoholic fatty liver disease (NAFLD).Thesystem supports integration with a user device through various approaches, including the use of REST APIs. This integration method ensures seamless operation and generally employs HTTPS for secure communication.The artificial intelligence (Al) based liver fibrosis risk assessment system can be accessed on the user device connected to a server via a network. The server, housing both processor and memory components, is linked to a database for efficient data management. Within the system, a plurality of modules isexecuted by the processor. These modules include client modules, an input module, a processing module, a prediction module, a clinical pathway module, and a risk prediction response module.

[0021] In one embodiment herein, the client modules are configured to configured to perform multiple functionsfor predicting the risk of liver fibrosis. The client modules comprise a data collection module, a standardization and pre-processing module and an API module. In one embodiment herein, the data collection module is configured to collect input data from various sources such as electronic medical records (EMRs), Data repositories and Databases. The standardization and pre-processing module is configured to standardize and pre-process the collected data to ensure its integrity and compatibility with the system's processing requirements. The API module is configured to transmit the pre- processed data from the data collection module and the standardization and pre-processing module to the server via the network. The Server transmits the preprocessed data to the Input module.

[0022] In one embodiment herein, the input module 112 is configured to receive the pre- processed data from the server (104) to initiate authenticate and validate input data. In one embodiment, the input module initiates authentication and validation of input data, ensuring compliance with required formats. It allows user to facilitateinput of personal parameters, medical history, and laboratory test parameters through the Application programming interface (API) or through web user interface of the user device or other devices.lnput parameters include personal variables such as age, gender, height, weight, and body mass index (BMI) are included, along with medical history details like Liver disease History (Fatty Liver / Non-Alcoholic Steatohepatitis (NASH) / Non-Alcoholic Fatty Liver Disease (NAFLD)), alcoholism, Infective hepatitis, diabetes mellitus and dyslipidemia.

[0023] In one embodiment, the one or more laboratory test parameters include Total bilirubin blood test, Serum Glutamic-oxaloacetic transaminase (SGOT) / aspartate aminotransferase (AST) blood test, Serum glutamic pyruvic transaminase (SGPT) / alanine transaminase (ALT) blood test, alkaline phosphatase (ALP) blood test, albumin blood test,Total protein, Total cholesterol, low-density lipoprotein cholesterol (LDL), and platelet count.

[0024] In one embodiment, the processing module transforms and standardizes data from the input module to meet the necessary input criteria for the subsequent analysis.

[0025] In one embodiment, the prediction module is structured to categorize an individual's risk levels of having Higher Grades of Liver Fibrosis as opposed to No / Low Grades of Liver Fibrosis. The score is predicted using a machine learning model, notably the XGB model, with an accuracy of 85% and above.

[0026] In one embodiment, the clinical pathway module is designed to provide a Recommended Protocol of subsequent actions for at least one patient, based on the determined risk thresholds.

[0027] In one embodiment, the risk prediction response module presents the patient's risk score along with a clinical algorithm for next best actions based on the stratified risk of liver fibrosis. The stratification is into multiple risk levels: low risk (F0 / F1 > 0.80), medium risk (F0 / F1 = 0.20 to 0.80 and F2-F4 = 0.80 to 0.20), and high risk (F0 / F1 < 0.20 and F2-F4 > 0.80). The risk prediction response module provides an overview including the patient's risk status (e.g., high, medium, low), Risk Score for probability of High fibrosis, Al score (ranging from 1 to 10), and a clinical algorithm with recommendations for diagnostic tests, referrals, treatment goals, educational materials, and revisit guidelines.

[0028] According to another aspect, the invention provides a method for assessing liver fibrosis risk in the patient using the Al-based system. At one step, the data collection module collects the input data from various sources such as electronic medical records (EMRs), Data repositories and Databases. At one step, the standardization and preprocessing module standardizes and pre-processes the collected data to ensure its integrity and compatibility with the system's processing requirements.

[0029] At one step, the API module transmit the pre-processed data from the data collection module and the standardization and pre-processing module to the server via the network.At one step, the input module receives the pre-processed data from the server to initiate authenticate and validate input data of the patient, including personal parameters, medical history, and lab test parameters. At one step, the processing module standardizes the collected data to fit the necessary input criteria for subsequent analysis. At one step, the prediction module receives the processed data from the processing module to predict the risk score using the XGB machine learning model to categorize the patient's risk of having higher grades of liver fibrosis.

[0030] At one step, the Clinical pathway module creates a personalized protocol of subsequent actions based on the determined risk. At one step, the risk prediction response module displays a patient's risk status score i.e. high, medium and low risk, risk score probability of High fibrosis, Al Score and a clinical algorithm suggesting next steps based on the stratified risk of liver fibrosis.

[0031] Further, objects and advantages of the present invention will be apparent from a study of the following portion of the specification, the claims, and the attached drawings.Detailed description of drawings:

[0032] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate an embodiment of the invention, and, together with the description, explain the principles of the invention.

[0033] FIG.l illustrates a block diagram of an artificial intelligence (Al)-based system for assessing liver fibrosis risk in patients with nonalcoholic Fatty Liver Disease (NAFLD), in accordance to an exemplary embodiment of the invention.

[0034] FIGs. 2A-2B illustrate a flow chart detailing the criteria for patient screening and clinical parameters for predicting liver fibrosis risk, along with the subsequent clinical algorithm for determining next actions, in accordance with an exemplary embodiment of the invention.

[0035] FIGs. 3A-3B illustrateforest plots of the multivariate odds ratios of the 11 predictors for development and validation cohorts, in accordance to an exemplary embodiment of the invention.

[0036] FIG. 4 illustrates a graphical representation of AUC ROC graphs for 98 liver biopsy validation cohort, in accordance to an exemplary embodiment of the invention.

[0037] FIGs. 5A-5D illustrate Violin graphswith combination of box plots and kernel density plots reflect on different clinical and lab variables and their representation for advanced fibrosis (F2-F4) vs No or low fibrosis in NAFLD patients, in accordance to an exemplary embodiment of the invention.

[0038] FIG. 6 illustrates a flowchart of a method for assessing liver fibrosis risk in the patient using the Al-based system, in accordance to an exemplary embodiment of the invention.

[0039] FIG. 7 illustrates the system architecture for the method designed to develop and deploy the Al-based liver fibrosis risk assessment tool for NAFLD patients, in accordance with an exemplary embodiment of the invention.Detailed invention disclosure:

[0040] Various embodiments of the present invention will be described in reference to the accompanying drawings. Wherever possible, same or similar reference numerals are used in the drawings and the description to refer to the same or like parts or steps.

[0041] The present disclosure has been made with a view towards solving the problem with the prior art described above, and it is an object of the present invention to provide an artificial intelligence (Al)-based system and method that assesses liver fibrosis risk in patients with nonalcoholic fatty liver disease (NAFLD).

[0042] According to one exemplary embodiment of the invention, FIG. 1 refers to a block diagram of an artificial intelligence (Al)-based Liver Fibrosis Risk Assessment system 100 for assessing liver fibrosis risk in patients with nonalcoholic fatty liver disease NAFLD).

[0043] In one embodiment herein, the system 100 supports integration with a user device through various approaches, one of the methods is use of REST APIs. This integration method ensures seamless operation and generally employs HTTPS for secure communication. The Al-based Liver Fibrosis Risk Assessment system 100 can be accessed on theuser device connected to a server 104 via a network 102. The server 104, housing both processor 106 and memory 108 components, is linked to a database 110 for efficient data management.Within the system 100, a plurality of modules is executed by the processor 106. These modules include client modules, an input module 112, a processing module 114, a prediction module 116, a clinical pathway module 118, and a risk prediction response module 120.

[0044] In one embodiment herein, the client modules are configured to configured to perform multiple functionsfor predicting the risk of liver fibrosis. The client modules comprise a data collection module 122, a standardization and pre-processing module 124 and an API module 126. In one embodiment herein, the data collection module 122 is configured to collect input data from various sources such as electronic medical records (EMRs), Data repositories and Databases. The standardization and pre-processing module 124 is configured to standardize and pre-process the collected data to ensure its integrity and compatibility with the system's processing requirements. The API module 126 is configured to transmit the pre-processed data from the data collection module 122 and the standardization and pre-processing module 124 to the server 104 via the network 102. The Server 104 transmits the preprocessed data to the Input module 112.

[0045] In one embodiment herein, the input module 112 is configured to receive the pre- processed data from the server (104) to initiate authenticate and validate input data, ensuring compliance with required formats. It allows Authorized users to facilitate personal parameters, medical history, and laboratory test parameters of at least one patient through a user interface of the user device or other devices. Clinical parameters such as age, gender,height, weight, and body mass index (BMI) are included, along with medical history details like Liver disease, Alcoholism, History of hepatitis Infections, Diabetes Mellitus, and Dyslipidemia.

[0046] In one embodiment herein, the laboratory test parameters include Total bilirubin blood test, Serum Glutamic-oxaloacetic transaminase (SGOT) and aspartate aminotransferase (AST) blood test, Serum glutamic pyruvic transaminase (SGPT) and alanine transaminase (ALT) blood test, alkaline phosphatase (ALP) blood test, albumin blood test, Total protein, Total cholesterol, low-density lipoprotein cholesterol (LDL), and platelet count. The data collected by the input module 112 is transmitted to the processing module.

[0047] In one embodiment herein, the processing module 114 transforms and standardizes data from the input module 112 to meet the necessary input criteria for the subsequent analysis.

[0048] In one embodiment herein, the prediction module 116 is structured to categorize an individual's risk levels of having Higher Grades of Liver Fibrosis as opposed to No / Low Grades of Liver Fibrosis. The score is predicted using a machine learning model, notably the XGB model, which has shown anaccuracy rate of 85%.

[0049] In one embodiment herein, the clinical pathway module 118 is designed to provide a personalized protocol of subsequent actions for at least one patient, based on the determined risk thresholds.

[0050] In one embodiment herein, the risk prediction response module 120 presents the patient's risk score along with a clinical algorithm for next best actions based on the stratified risk of liver fibrosis. The stratification is into multiple risk levels: low risk (F0 / F1 > 0.80), medium risk (F0 / F1 = 0.20 to 0.80 and F2-F4 = 0.80 to 0.20), and high risk (F0 / F1 < 0.20 and F2-F4 > 0.80). The risk prediction response module 120 provides an overview including the patient's risk status (e.g., high, medium, low), risk score of probability of significant fibrosis, Al score (e.g., 1 to 10), and a clinical algorithm with recommendations for diagnostic tests, referrals, treatment goals, educational materials, and revisit guidelines.

[0051] According to another exemplary embodiment of the invention, FIGs. 2A-2B depict flowcharts(200, 202) of the clinical algorithm for liver fibrosis assessment in patients with non-alcoholic fatty liver disease (NAFLD). This algorithm recommends individual protocols based on risk thresholds, categorized as low risk (F0 / F1 > 0.80), medium risk (F0 / F1 = 0.20 to 0.80 and F2-F4 = 0.80 to 0.20), and high risk (F0 / F1 < 0.20 and F2-F4 > 0.80). The flowchart 200 encompasses 24 clinical input parameters broadly categorized as patient parameters, medical history, lifestyle, and lab values. It includes steps such as generating an Al-based liver fibrosis risk score and recommending protocols such as lab investigations, diagnostics & imaging, treatment goals, repeat testing, and referrals for each risk category.

[0052] In one embodiment herein, the system 100 considers 24 clinical and lab parameters collected by the input module 112. The clinical and lab parameters include patient parameters, patient medical history, lifestyle, and lab values. The patient parametersinclude the patient's gender, height, weight and body mass index (BMI). The medical history includes etiology-liver disease, hypertension, diabetes mellitus, Dyslipidemia, history of liver disease and history of jaundice.The lifestyle attributesinclude the patient's diet, alcohol intake, and smoking habits. The Lab Values includes the patient's blood tests, such as total bilirubin, SGPT, SGOT, total proteins, cholesterol, HDL, platelets, alkaline phosphatase, and albumin.

[0053] In one embodiment herein, processing modulell4 standardizes data from the input module 112 for analysis, the prediction module 116 utilizes a machine learning model like XGB, with an accuracy of 85% and above, to categorize individuals' liver fibrosis risk levels. The clinical pathway module 118, in another embodiment, offers personalized protocols for subsequent actions, tailored to determined risk thresholds. Additionally, the risk prediction response module 120 presents the patient's risk score and stratified risk levels, along with recommendations for diagnostic tests, referrals, treatment goals, educational materials, and revisit guidelines, enhancing intervention efficacy.

[0054] In one embodiment herein, the minimal risk is considered when the F0 / F1 > 0.08. The moderate risk is considered when the F0 / F1 varies between 0.20 and 0.80, and F2-F4varies between 0.80 and 0.20, as shown in FIG 2A. The high risk is considered when the F0 / F1 < 0.20, and F2-F4 > 0.8, as shown in FIG 2B.

[0055] In one embodiment herein, the lab investigations are recommended based on a patient's risk score. Patients with minimal and moderate risk scores are advised to undergo tests such as a complete blood count, fasting and postprandial blood sugar, liver function tests, and HBA1C. Those with high-risk scores require additional testing, including prothrombin time and alpha-fetoprotein.

[0056] In one embodiment herein, the recommended diagnostic and imaging tests for minimal and moderate risk patients include ECG, 2D Echo for cardiovascular assessment, abdominal ultrasound, elastography, and upper Gl endoscopy. High-risk patients require these tests and additional ones such as hepatocellular carcinoma screening and liver biopsy.

[0057] In one embodiment herein, the treatment goals for NAFLD patients focus on treating underlying conditions, optimizing metabolic risk factors, and managing complications. Repeat testing is suggested every three years for minimal risk, every six months for moderate risk, and every three months for high risk.

[0058] In one embodiment herein, the referral to a gastroenterologist is recommended based on the patient's risk level: minimal risk patients do not require a referral, moderate risk patients should be referred routinely, and high-risk patients need urgent referral. The flowchart 200 provides a general framework, but specific treatments and further tests should be determined by the physician.

[0059] According to another exemplary embodiment of the invention, FIGs. 3A-3B refer to forest plots (300, 302) of the multivariate odds ratios of the 11 predictors for development and validation cohorts. The data of 5150 NAFLD patients (Advanced Fibrosis F2-4 - 30.34%) is collected from various regions of Apollohospitals for certain time period using standardized template and electronic medical records (ICD 10 codes). The 25 clinical and laboratory test parameters are studied along with patients' Elastography reports and ARFI values. In one embodiment herein, considering 11 variables out of the 25 clinical andlaboratory test parameters including Age [Multivariate Odds Ratio (OR)- 3.39; 95%CI 2.99 - 3.84], History of Diabetes Mellitus [OR - 6.80, 95%CI 5.92 - 7.81], Albumin [OR- 3.70, 95%CI 3.25 - 4.20] , Aspartate aminotransferase (AST) [OR- 3.65, 95960 - 3.21 - 4.16} , Total Bilirubin [OR-3.13, 95%CI 2.76 - 3.56] and Platelet Count [OR-2.74, 95%CI 2.40 - 3.13]are found to be significant. The Machine Learning (ML) modelling is performed using extreme Gradient Boosting (XGB) algorithm. The prospective validation cohort is selected of 1261 patients (F2-4 - 31.24%) for certain time period and compared with Fib4 Score. This is further validated with 98 Liver Biopsies from validation cohort. Theforest plots (300, 302) for Odds ratios of development and validation cohorts are shown in FIGs 3A and FIGs 3B respectively.

[0060] According to another exemplary embodiment of the invention, FIG. 4 refers to a graphical representation 400 of AUC ROC Curves for 98 liver biopsy validation cohort. In one embodiment herein, the performance parameters of the development model are AUC ROC Score of 0.94 and validation cohort had the AUC and accuracy of 0.88. The AUC for 98 liver biopsy validation cohort is 0.83 (shown in FIG. 4). The model performed better than Fib4 Score with Net Reclassification Improvement (NRI) at 0.499.

[0061] According to another exemplary embodiment of the invention, FIGs.5A-5D refer to the Violin graphs (500, 502, 504, 506) with combination of box plots and kernel density plots reflect on different clinical and lab variables and their representation for advanced fibrosis (F2-4) vs No or low fibrosis in NAFLD patients.

[0062] According to another exemplary embodiment of the invention, Figures 5A to 5D pertain to Violin graphs (500, 502, 504, 506) which combine box plots and kernel density plots. These graphs reflect various clinical and laboratory variables, illustrating their representation for advanced fibrosis (F2-4) versus no or low fibrosis in patients with Non- Alcoholic Fatty Liver Disease (NAFLD).These violin graphs provide a comprehensive visualization of data distribution, highlighting the differences in clinical and laboratory variables between patients with varying degrees of fibrosis. The combination of box plots and kernel density plots allows for a detailed analysis of the central tendency, spread, and density of the data, facilitating the identification of significant patterns and trends.

[0063] Specifically, Figure 5A depicts the distribution of Age for patients with advanced fibrosis (F2-4) compared to those with no or low fibrosis. Figure 5B illustrates the distribution of Albumin levels for patients with advanced fibrosis (F2-4) compared to those with no or low fibrosis. Figure 5C compares the distribution of AST (Aspartate Aminotransferase) levels, and Figure 5D represents the distribution of Platelet counts, providing a clear comparison between patients with advanced fibrosis and those with no or low fibrosis.

[0064] According to another exemplary embodiment of the invention, FIG. 6 refers to an example flowchart 600 of a method forassessing liver fibrosis risk in the patient using the Al- based system 100. At step 602, the data collection module 122 collects the input data from various sources such as electronic medical records (EMRs), Data repositories and Databases. At step 604, the standardization and pre-processing module 124 standardizes and pre- processes the collected data to ensure its integrity and compatibility with the system's processing requirements.

[0065] At step 606, the API module 126 transmit the pre-processed data from the data collection module 122 and the standardization and pre-processing module 124 to the server 104 via the network 102.At step 608, the input module 112 receives the pre-processed data from the server 104 to initiate authenticate and validate input data of the patient, including personal parameters, medical history, and lab test parameters. At step 610, the processing module 114standardizes the collected data to fit the necessary input criteria for subsequent analysis. At step 612, the prediction module 116receives the processed data from the processing module 114 to predict the risk score using the XGB machine learning model to categorize the patient's risk of having higher grades of liver fibrosis.

[0066] At step 614, the Clinical pathway module 118createsa personalized protocol of subsequent actions based on the determined risk. At step 616, the risk prediction response module 120 displays a patient's risk status score i.e. high, medium and low risk, risk score probability of High fibrosis, Al Score and a clinical algorithm suggesting next steps based onthe stratified risk of liver fibrosis. The risk score of probability of high Fibrosis and Al predicted score and provides Recommended protocol based on the risk threshold or risk status.

[0067] In accordance with another exemplary embodiment of the present invention, FIG. 7 provides a schematic depiction of the system architecture 700 for a method devised to develop and deploy the Al-based Liver Fibrosis Risk Assessment Tool. The diagram highlights a sequence of operations that are integral to the system's development and deployment.

[0068] As depicted in step 702, we have the data sourcing phase where information is collected from various sources, including but not limited to health checks, Inpatient Records, Laboratory& Diagnostic data from Electronic Medical Records (EMR) and servers 714. Additional sources include clinical knowledge bases 716 like Standard Operating Protocols (SOPs), literature from research papers, journals, and taking help of clinicians 718.

[0069] In step 704, we move to the data ingestion phase. Here, data is collected and put into a centralized database 720. This data then traverses through a data pipeline 722and is then stored in a Data Repository 724 in structured formats. This phase is crucial for organizing the data and preparing it for further pre-processing stages.

[0070] In step 706, we include the data pre-processing phase, which encompasses metadata management 726, ETL processes 728, and data transformation and harmonization 730. These processes improve data quality and consistency, setting the stage for subsequent analysis and model training.

[0071] In step 708, the data analysis phase, a range of statistical tools are harnessed to extract insights and correlations from the gathered data. Propensity matching 732 unveils hidden patterns, while descriptive statistics 734 provide a summarized overview of the data. Correlation coefficients 736 establish relationships between various risk factors, and odds / hazard ratios 738 compare the possibility of outcomes. KM plots and survival charts 740 visualize time-to-event data, and visualization tools 742 simplify complex datarepresentation. These tools, in concert, contribute to feature selection and model building of the liver fibrosis risk assessment system 100.

[0072] In step 710, we move towards model development, employing a neural network model created with the TensorFlow library, specifically designed for predicting the likelihood of liver fibrosis development in NAFLD patients.

[0073] The model's structure is built of four layers, each with a specific number of neurons. The first layer has 256 neurons, followed by the second and third layers with 128 and 64 neurons respectively, and finally a binary classification layer with one neuron. The activation functions 'relu' and 'sigmoid' are used in the first three layers and the last layer respectively, effectively tuning the model for binary classification. By training the model with domain insights, data insights, and observing various iterations of the model performance, we have arrived at the optimal performance.

[0074] In further step 712, involves the deployment of the model for integration and usage through REST API 752 protocols. This architecture involves using an API Management Service 744 and an application service resource 750 for model inference. The model inference code is developed in Python programming language 754. All the resources are hosted in a Virtual Private Network 748securely. The API service 746 acts as the interface for the REST API, facilitating communication between the API Management Service and the web application. The API Management Service 744 includes storage capabilities 758 for securely managing and storing model artifacts, data inputs, and results. In one such application of the integration, this model inference API is integrated into a web application code 756 to be deployed and used as a Web Application 760. This web application serves as a user-friendly platform for users or clinicians, enabling them to input data and obtain liver fibrosis risk scores. Therefore, this phase makes the model operational and accessible, facilitating the prediction of liver fibrosis risk.

[0075] Numerous advantages of the present disclosure may be apparent from the discussion above. In accordance with the present disclosure an artificial intelligence (Al)- based system 100 for assessing liver fibrosis risk in patients and method thereof is disclosed.The proposed invention provides an artificial intelligence (Al)-based system 100 that assesses liver fibrosis risk in patients with Nonalcoholic fatty liver disease (NAFLD). The Al- based systemis developed usingethically sourced deanonymized Indian population data by incorporating comprehensive parameters such as clinical features, medication history, and lab reports.

[0076] The proposed invention providestheAI-based Liver Fibrosis Risk prediction system 100 provides comprehensive and holistic risk assessment. The Al-based system 100 assists in personalized recommendations based on stratified risk for liver fibrosis through an integrated clinical decision support tool.

[0077] It will readily be apparent that numerous modifications and alterations can be made to the processes described in the foregoing examples without departing from the principles underlying the invention, and all such modifications and alterations are intended to be embraced by this application.19

Claims

5. CLAIMS: l / We Claim:

1. An artificial intelligence (Al)-based system (100) for predicting the risk of liver fibrosis in patients suffering from Nonalcoholic Fatty Liver Disease (NAFLD), comprising: a server (104) having a processor (104) and a memory (106) for storing one or more instructions executed by the processor (104), wherein the server (104) is in communication with a database (110) for efficient data management, wherein the server (104) is in communication with a user device via a network (102) through various methods, including REST APIs for ensuring smooth operation, and is configured to communicate securely typically using HTTPS; wherein the processor (104) is configured to execute plurality of modules (108) for performing multiple operations, wherein the plurality of modules comprises: client modules configured to perform multiple functionsfor predicting the risk of liver fibrosis, wherein the client modules comprise: a data collection module (122) configured to collect input data from various sources such as electronic medical records (EMRs), Data repositories and Databases; a standardization and pre-processing module (124) configured to standardize and pre-process the collected data to ensure its integrity and compatibility with the system's processing requirements; an API module (126) configured to transmit the pre-processed data from the data collection module (122) and the standardization and pre-processing module (124) to the server (104) via the network (102); an input module (112) configured to receive the pre-processed data from the server (104) to authenticate and validate input data, thereby ensuring compliance with required formats from various sources, and allowing users to facilitate data relatedto one or more personal parameters, medical history and one or more laboratory test parameters of the patient received through the Application programming interface (API) or through web user interface of User device or other devices; a processing module (114) configured to transform and standardize data from the input module (112) to meet the necessary input criteria for the subsequent analysis; a prediction module (116) configured to categorize an individual's risk levels of having higher grades of liver fibrosis as opposed to no and low grades of liver fibrosis, wherein the prediction module (116) configured to predict the risk score using a machine learning model, notably the XGB model, with an accuracy rate of at least 85%; and a clinical pathway module (118) configured to provide a personalized protocol of subsequent actions for at least one patient based on the determined risk thresholds; and a risk prediction response module (120) configured to provide a patient's risk score and a clinical algorithm for next steps based on their liver fibrosis risk level, thereby offering a comprehensive view of the patient's risk status stratified as low, medium, and high Risk, Risk Score for probability of High fibrosis and Predicted Al Score and providing clinical algorithm that suggests diagnostic tests, referrals, treatment goals, educational materials, and guidelines for future visits and tests follow-up, with the capability to generate a PDF report, and whereby the Al-based system (100) acts as a complementary tool for clinicians, and enhances informed decision-making without replacing clinicians or any diagnostic tests. The system adheres to ISO 13485 standards, ensuring patient safety and reliability, and the Al-based system (100) functions as a software as a medical device (SaMD) and certified by ISO 13485, and integrates algorithms for assessing risk of Liver Fibrosis in patients with Non-alcoholic Fatty Liver Disease, thereby enhancing patient outcomes and supporting clinicians.

2. The artificial intelligence (Al)-based system (100) as claimed in claim 1, wherein the one or more clinical parameters include age, gender, height, weight and body mass index (BMI).

3. The artificial intelligence (Al)-based system (100) as claimed in claim 1, wherein the medical history includes liver disease (fatty liver, non-alcoholic steatohepatitis (NASH) and non-alcoholic fatty liver disease (NAFLD)), alcoholism, Infective hepatitis, diabetes mellitus and dyslipidemia.

4. The artificial intelligence (Al)-based system (100) as claimed in claim 1, wherein the one or more laboratory test parameters include total bilirubin blood test, serum Glutamic- oxaloacetic transaminase (SGOT) and aspartate aminotransferase (AST) blood test, serum glutamic pyruvic transaminase (SGPT) and alanine transaminase (ALT) blood test, alkaline phosphatase (ALP) blood test, albumin blood test, total protein, total cholesterol, low- density lipoprotein cholesterol (LDL), and platelet count.

5. The artificial intelligence (Al)-based system (100) as claimed in claim 1, wherein the machine learning model is trained to predict liver fibrosis risk in (non-alcoholic fatty liver disease) NAFLD patient.

6. The artificial intelligence (Al)-based system (100) as claimed in claim 1, wherein the Al- based system (100) comprises a clinical pathway module (118) configured to provide recommended protocol of next based actions for at least one patient based on the determined risk thresholds.

7. The artificial intelligence (Al)-based system (100) as claimed in claim 1, wherein the recommended protocol includes diagnostic tests, pulmonology referral, treatment goals, educational materials or information and revisit guidelines.

8. The artificial intelligence (Al)-based system (100) as claimed in claim 1, wherein the at least one risk categories include a low-risk category, a moderate risk category and a high risk category, risk score for probability of high fibrosis and Al score.

9. The artificial intelligence (Al)-based system (100) as claimed in claim 1, wherein the processor (106) is in communication with an application server (104) through a network (102).

10. A method for assessing liver fibrosis risk in a patient using an Al-based system (100), comprising: collecting, by a data collection module (122), input data from various sources such as electronic medical records (EMRs), Data repositories and Databases; standardizing and pre-processing, by a standardization and pre-processing module (124), the collected data to ensure its integrity and compatibility with the system's processing requirements; transmitting, by an API module (126), the pre-processed data from the data collection module (122) and the standardization and pre-processing module (124) to the server (104) via a network (102); receiving, by an input module (112), the pre-processed data from the server (104) to initiate authenticate and validate input data of the patient, including personal parameters, medical history, and lab test parameters; standardizing, by a processing module (114), the collected data to fit the necessary input criteria for subsequent analysis; receiving, by the prediction module (116), the processed data from the processing module (114) to predict the risk score using the XGB machine learning model to categorize the patient's risk of having higher grades of liver fibrosis; creating, by a clinical pathway module (118), a personalized protocol of subsequent actions based on the determined risk; and displaying, by the risk prediction response module (120), a patient's risk status score of probability of High fibrosis, Al Score and a clinical algorithm suggesting next steps based on the stratified risk of liver fibrosis.DATE AND SIGNATURE:Dated this 28thday of August, 2024

Images