Human-Centred Multimodal Deep Learning Models for Chest X-Ray Diagnosis

Abstract

Chest X-rays are a cornerstone of clinical diagnosis, yet current AI systems often overlook the patient information that radiologists rely on. This thesis develops multimodal and human-centred AI methods that integrate chest X-rays with clinical data to improve diagnostic accuracy and real-world reliability. This thesis introduces a new dataset combining imaging, clinical records and eye-tracking, proposes deep learning models that fuse these modalities, and develops language-model-based techniques to generate missing clinical features. Together, these contributions advance trustworthy, context-aware medical AI for deployment in clinical practice.

Ähnliche Arbeiten