Explainable artificial intelligence (XAI) in medical imaging: a systematic review of techniques, applications, and challenges
Abstract
Explainable Artificial Intelligence (XAI) is crucial for enhancing transparency and trustworthiness, as well as for developing AI-based diagnostic systems in medical imaging to achieve clinical acceptability and reliability. This synthesis aligns with the current context of XAI in medical imaging on four key dimensions: trends in techniques, their application to clinical use cases or with human subjects, and the associated problems. In radiology and pathology, we report on image analysis based on current literature from credible databases. This review extends on the existing surveys by explicitly addressing the focus on feature selection (FS), graph neural networks (GNNs), and multimodal transformers and combining them to a cohesive XAI taxonomy and matches the techniques to the specific impact points in the clinical workflow in radiology and pathology. We present saliency maps, attention mechanisms, and gradient-based and rule-based elucidations of deep learning (DL) models in today’s healthcare environment. Finally, the results show that XAI significantly enhances overall clinical decision-making by making the high-level reasoning of the model readily available to users, thereby increasing their confidence in clinical decisions despite the remaining obstacles, including standardization, interpretability, data bias, and complicated data integration. From 980 records, 289 duplicates were removed, 691 screened, 209 excluded, 482 full texts assessed, 263 excluded with 219…
How strong is the evidence?
Evidence Hub reads this as a systematic review, done in people. Its evidence level is 1 of 5 (Highest · pooled evidence). Pooled evidence from several studies sits at the top of the evidence pyramid.
Read the paper
Related studies
- Active Learning during Federated Learning for Object Detection
- Fairness-aware federated learning
- Decentralized Federated Learning
- Building a Personally Identifiable Information Recognizer in a Privacy Preserved Manner Using Automated Annotation and Federated Learning
- Preservation of Patient Level Privacy: Federated Classification and Calibration Models
- Efficient explainable federated learning
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.