Harmonizing fusion modeling for accurate liver cancer diagnosis using explainable artificial intelligence: a step toward trustworthy medical AI

Artificial Intelligence (AI) has transformed healthcare by improving prediction, diagnosis, and decision-making processes. Nonetheless, the opaque character of AI presents a considerable obstacle, limiting its extensive utilization, particularly in vital domains such as cancers. This study presents a robust fusion modeling approach for liver cancer detection. It combines ensemble modeling with explainable artificial intelligence, enhancing predicted accuracy while providing clear, interpretable insights into its decision-making process. The proposed methodology combines Random Forest, Gradient Boosting, and Extra Trees classifiers with a voting mechanism to assess the likelihood of liver cancer. The ensemble model improves essential performance metrics, including accuracy, precision, recall, and F1-score, by integrating the capabilities of several methods. The experimental findings indicate that the model attains an accuracy of 0.9682, precision of 0.9824, recall of 0.9787, and an F1-score of 0.9765. The model utilizes SHapley Additive exPlanations (SHAP) to guarantee transparency and foster trust in its predictions, providing local and global interpretability. SHAP delivers a comprehensive analysis of feature contributions, enabling healthcare practitioners to grasp the determinants influencing each prediction. This high-performance prediction and interpretability fusion reconciles AI breakthroughs with clinical practice, providing a more precise and reliable liver cancer detection and therapy method. The proposed fusion modeling demonstrates greater adaptability than current state-of-the-art procedures.