Explainable radiomics for prostate cancer diagnosis

Prostate cancer diagnosis can be challenging. Artificial intelligence (AI) and radiomic models are emerging as valuable tools for identifying malignant cells. However, the black-box nature of these models hinders their clinical use. Explainable AI (XAI) is gaining traction to clarify how machine learning models make decisions, enhancing interpretability for users. This study presents preliminary results from explainable radiomics models trained on a dataset from ARNAS Ospedali Civico Di Cristina, consisting of PET-CT DICOM images of patients with suspected prostate cancer who underwent standard PET/CT examinations. Our analysis focuses on the retrospective reconstruction of tumor diagnosis. AI models for radiomic feature extraction were implemented using the PyRadiomics toolbox. An explainability study was conducted to understand the decision-making processes of the AI framework. Results indicate that our model can predict disease diagnosis while identifying influential features. The radiomics model shows promise in detecting cancerous cells, with added explainability enhancing its potential for clinical implementation