Improving diagnostic trust: an explainable deep learning framework for genitourinary cancer prediction

Genitourinary malignancies significantly impact global health and present substantial physical and economic challenges.To improve diagnostics, there has been a growing reliance on machine learning and deep learning algorithms in clinical settings but it is slow to gain wide acceptance due to the complex nature of these "black-box" models.In this paper, we introduced an explainable deep-learning framework for the diagnosis of Genitourinary Cancers utilizing laboratory data to enhance trust and transparency for healthcare staff and patients.Firstly, we trained various approaches namely Deep Neural Network (DNN), Random Forest (RF), Support Vector Machine (SVM), Logistic Regression (LR), K-Nearest Neighbour (KNN) and Naive Bayes (NB) on patient's data and evaluated their performance through standard metrics.The experimental results shows that our proposed DNN performed better than other state-of-the-art achieving an accuracy of 90.64%, precision of 90.58%, recall of 90.70% and F1-score of 90.55%.Secondly, to improve the transparency and explainability of the DNN, we integrated explainable AI approaches including Local Interpretable Model-agnostic Explanations (LIME), SHapley Additive exPlanations (SHAP) and Partial Dependence Plots (PDPs) to empower healthcare staff to make more precise and informed decisions in clinical practice.