Deep learning algorithm for automatic assessment of periapical lesion healing on sequential radiographs

Abstract

Deep learning has emerged as a transformative technology in dental imaging, offering automated and objective assessment of treatment outcomes. In endodontics, monitoring the healing of periapical lesions through sequential periapical radiographs is essential for determining the success of root canal therapy. Conventional radiographic evaluation is often limited by subjective interpretation, interobserver variability, and challenges in detecting subtle changes over time. This review explores the application of deep learning algorithms for the automatic assessment of periapical lesion healing using sequential radiographic images. It discusses the principles of image preprocessing, lesion segmentation, feature extraction, and temporal comparison employed by modern artificial intelligence models to quantify healing progression. The review also highlights the potential of deep learning to improve diagnostic accuracy, standardize clinical evaluations, reduce assessment time, and support evidence-based treatment decisions. Furthermore, current challenges, including dataset variability, image quality, algorithm interpretability, and clinical integration, are examined alongside future opportunities for multimodal artificial intelligence and digital workflow implementation. Overall, deep learning-assisted radiographic analysis represents a promising advancement in precision endodontics by providing reliable, reproducible, and efficient monitoring of periapical healing. Continued validation through large-scale clinical studies and integration into routine dental practice will further enhance its role in improving patient outcomes and optimizing long-term treatment success.