زهرا خسروي

In recent years, global lifestyle changes have contributed significantly to a surge in the preva‎lence of diseases. Among the notable consequences is the escalating incidence of insulin resistance, a precursor to various ailments such as fatty liver and type 2 diabetes. Non-alcoholic fatty liver disease (NAFLD) has emerged as the most preva‎lent liver disorder worldwide, impacting a quarter of the global population. This rate soars to 50-70% among individuals with diabetes. The coexistence of NAFLD and type 2 diabetes not only accelerates the progression of both conditions but also elevates the risk of additional disorders, including cardiovascular diseases. Given the intricate nature of chronic and underlying diseases and the wealth of data generated during the treatment process, the application of machine learning algorithms has proven to be a valuable tool for the timely diagnosis and prediction of these ailments. This study leveraged data from 3762 Iranian patients with type 2 diabetes and employed six classification models—K-Nearest Neighbor, Logistic Regression, Decision Tree, Random Forest, Enhanced Gradient, and Catboost—to predict NAFLD in these patients. The integration of a combined genetic algorithm for feature selection yielded a notable 6% average improvement across the models. Ultimately, employing the genetic algorithm to select 19 features out of 29 available features, the Catboost algorithm demonstrated the most favorable outcome with an area under the receiver operating characteristic curve (AUC) of 89%. According to the algorithm's findings, ALT, PLT, Vitamin D, and HOMA emerged as pivotal features in predicting NAFLD in diabetic patients. Additionally, a numerical index was developed alongside the model using the genetic programming algorithm, aiming to enhance cost-effectiveness, ease of use, and practicality in the diagnostic and decision-making processes for healthcare professionals. Comparative analysis with other indices in the field revealed satisfactory results for this newly developed index.

بيماري كبدچرب مرتبط با اختلال متابوليك , مدل‌هاي پيش‌بيني كننده , الگوريتم ژنتيك انتخاب ويژگي , الگوريتم برنامه‌ريزي ژنتيك , شاخص عددي

Metabolic-associated fatty liver disease , Predictive models , Feature selection genetic algorithm , Genetic programming algorithm , Numerical index

Zahra Khosravi

Dr. Farnaz Barzinpur