Executive Summary : | Liver Cirrhosis (LC) is a prevalent global condition caused by various factors such as obesity, non-alcoholic fatty liver disease, high alcohol consumption, hepatitis B or C infection, autoimmune diseases, cholestatic diseases, and iron or copper overload. Electronic Health Records (EHR/EMR) data can improve care documentation, communication of clinical information across sites, and decision support for individual health improvement. AI-based data analysis can be helpful in diagnosing LC and predicting mortality in terms of prognostics indicators. This study proposes an EMR registry of LC patients in collaboration with the Gastroenterology Department of SCB Medical College, Cuttack and KIMS, Bhubaneswar, and AI-ML-based data study to develop a model predicting time-to-event (TTE) for liver cirrhosis diagnosis. The objective will involve creating an EMR registry, collecting approximately 1000 sample data, and extracting features based on clinical and biomarkers data. The inclusion of biomarkers-based features will improve performance in prognosis of LC. Deep learning models (DNN) will be compared with machine learning models like Random Forest (RF) and Logistic Regression (LR) to predict mortality and compare this to the MELD-Na model using four variables. The study will use libraries like Scikit-learn, Keras, Scipy, and Matplotlib with Python for analysis. The receiver operating curve will be used to evaluate the performance of the models for each prediction case. The proposed objective will help develop AI-assisted models, such as LC-Pred software, to guide clinical decisions about patients unlikely to have liver cirrhosis without the need for invasive methods. |