Quantifying the Impact of Clinical Features On MRI-Based Stroke Diagnosis using Machine Learning

Abstract

Accurate diagnosis of ischemic stroke is critical for timely intervention and improved patient outcomes. While MRI imaging is central to stroke diagnosis, relying solely on imaging data may overlook important clinical factors that contribute to patient-specific outcomes. This paper explores the impact of integrating clinical features—such as demographics, medical history, and stroke-specific metrics—with MRI-based features to enhance diagnostic accuracy. Using machine learning techniques, we evaluate and quantify the contribution of clinical data in improving model performance for stroke classification. We develop and compare three models: one using only MRI features, one using only clinical features, and a combined model incorporating both. The results demonstrate that the fusion of clinical and imaging data significantly boosts classification accuracy and model interpretability, proving that clinical features play a vital role in improving stroke diagnosis. This research underscores the importance of a multimodal approach in medical diagnostics, where clinical and imaging data together provide a more comprehensive understanding of ischemic stroke.