This study introduces an innovative explainable AI (XAI) pipeline designed to predict breast cancer survival by integrating clinical, socioeconomic, and geographic data. Using data from 10,172 patients treated at hospitals in the Memphis, Tennessee metropolitan area, the pipeline identifies key survival determinants and reveals significant survival disparities affecting Black women. Advanced machine learning models combined with SHapley Additive exPlanations (SHAP) provide actionable and interpretable insights into the role of tumor stage, socioeconomic conditions, and access to preventive care. This framework facilitates personalized survival predictions and targeted equity-focused interventions, demonstrating the potential of multi-source data integration to address health inequities and improve patient outcomes.

This study reports on the progress in designing and developing a framework for integrating heterogeneous datasets—structured, semi-structured, and unstructured—into an AI-ready multimodal data pipeline aimed at predicting radiation therapy interruptions (RTI) and enhancing patient care navigation. The AI-Ready dataset incorporates a broad set of information, including patient demographics, health data, clinical notes, medical imaging, and data on social determinants of health. Preliminary results indicate that this pipeline effectively integrates diverse distributed data sources, providing a foundation for training AI models capable of generating reliable and actionable predictions.

The proliferation of high-dimensional data in the era of big data has presented significant challenges for machine learning models. Feature selection methods have emerged as essential preprocessing techniques to address these challenges. However, most existing feature selection techniques primarily rely on correlations or associations between features and the target variable, overlooking the consideration of causal relationships. This study introduces a novel causal feature selection (CFS) algorithm that leverages causal structure learning to identify a subset of causal features. Our approach involves employing a causal graph discovery method to represent the causal relationships among variables and the causal effects of features on the target variable. To evaluate the effectiveness of our proposed CFS algorithm, we introduce a new evaluation criterion based on causal metrics, offering a principled and rigorous approach to assess the performance of causal feature selection methods. We empirically evaluate our algorithm using synthetic and real-world datasets, demonstrating that the truncated subsets of features selected by the CFS algorithm exhibit comparable or improved performance compared to baseline methods while utilizing fewer causal features.