Radiomics-Driven Machine Learning Models for Diagnosis of Pancreatic Adenocarcinoma

This repository contains the complete source code and dataset for the manuscript "The Application of Radiomics and Machine Learning in Diagnosis of Pancreatic Cancer" submitted to the Iranian Journal of Medical Sciences. The study develops and validates machine learning models using CT-based radiomics features to distinguish pancreatic adenocarcinoma from normal pancreatic tissue. The repository enables full reproducibility of all analyses presented in the manuscript, including: Feature Selection: Implementation of three feature selection methods (Mutual Information, LASSO, Recursive Feature Elimination) Machine Learning Modeling: Three classifiers (Random Forest, Support Vector Machine, Logistic Regression) evaluated using 5-fold stratified cross-validation Performance Evaluation: Comprehensive metrics including Accuracy, Precision, Sensitivity, F1-score, PPV, NPV, and AUC-ROC Model Interpretability: SHAP (SHapley Additive exPlanations) analysis for feature importance interpretation