Gaze-Driven Pneumonia Detection: Enhancing Diagnostic Accuracy with Eye-Tracking Insights

Pneumonia, a prevalent respiratory infection, continues to pose a significant global health threat, underscoring the imperative for precise diagnostic approaches. This research aims to enhance pneumonia classification from chest X-ray images by incorporating gaze data and comparing its performance to a baseline model solely reliant on chest X-ray images. Leveraging gaze data along with chest X-ray images, our study proposes two distinct methodologies, both of which demonstrate improved performance. Our findings showcase compelling results, with the model integrating gaze data achieving a notable test accuracy of 81.32%. Additionally, the precision, recall, and F1 scores for this model stand at 0.84, 0.82, and 0.80 respectively, surpassing those of the baseline model utilizing only chest X-ray images. These results highlight the effectiveness of incorporating gaze data in pneumonia classification tasks. By showcasing the efficacy of integrating gaze data, this research will aid in the development of more robust and reliable decision support systems for medical imaging analysis.