JMIR Publications

ABSTRACT

Breast cancer is the most common cancer in women worldwide, and early detection and accurate diagnosis are crucial for improving patient outcomes. In Saudi Arabia, it is the most prevalent cancer type among women, with a projected increase by the year 2040. In this research, we aimed to apply Explainable Artificial Intelligence (XAI) learning approaches to predict benign and malignant breast cancer using various clinical and pathological features of breast cancer patients in Saudi Arabia. Six different models were trained, and their performance was evaluated using several common metrics, including accuracy, precision, recall, F1 score, and AUC-ROC score. To improve transparency and interpretability, we applied Local Interpretable Model-Agnostic Explanations (LIME) and SHapley Additive exPlanations (SHAP) to interpret the proposed models. Our analysis reveals that the Random Forest model exhibited the highest performance metrics compared to all other models, demonstrating its proficiency in accurately predicting benign or malignant breast cancer diagnoses. The model achieved an accuracy of 0.72, along with precision, recall, F1 score, and AUC-ROC score values of 0.69, 0.77, 0.73, and 0.72, respectively. Conversely, the Support Vector Machine model demonstrated the poorest performance metrics among all models, with an accuracy of 0.59, indicating its limited ability to accurately predict breast cancer diagnoses. Moreover, the application of XAI approaches revealed notable discrepancies in the rankings of feature importance across the proposed models, highlighting the need for further investigations. These findings provide valuable insights to healthcare providers regarding the diagnosis and interpretation of machine learning results, as well as the potential integration of such technologies in healthcare practices.