Effectiveness of lightweight convolutional neural networks for detecting the relationship between the mandibular third molar and the inferior alveolar canal on panoramic radiographs

Objective: This study aimed to develop and evaluate lightweight convolutional neural networks (CNNs) capable of automatically localizing the mandibular third molar (M۳) and classifying its relationship with the inferior alveolar canal (IAC) on panoramic radiographs.Methods: A total of ۶۰۹ panoramic radiographs (containing ۸۹۹ M۳s) were analyzed in two stages. First, ۸۲ panoramic images (۱۳۴ M۳s) were used to fine-tune a pre-trained EfficientDet model for automatic M۳ localization. The detected regions were standardized to include the IAC and preprocessed through resizing, contrast enhancement, and mirroring. For ground-truth labeling, the presence or absence of M۳–IAC contact was determined by an experienced oral and maxillofacial radiologist based on established panoramic radiographic criteria. Second, a custom lightweight CNN was trained on ۵۲۷ panoramic radiographs (۷۶۵ M۳s) to classify M۳–IAC contact (contact = ۱, no contact = ۰). Model performance was compared with a pre-trained ResNet۵۰ architecture using accuracy, sensitivity, specificity, precision, and F۱ score.Results: The detection model achieved ۱۰۰% accuracy with an intersection-over-union (IoU) of ۸۷.۹%. Compared to the ResNet۵۰ benchmark model, the lightweight CNN demonstrated comparable overall accuracy (۸۷.۵%). However, the lightweight CNN outperformed ResNet۵۰ in specificity (۹۰.۴% versus ۸۶.۹%) and precision (۹۳.۴% versus ۸۸.۷%), while ResNet۵۰ exhibited a slightly higher mean sensitivity (۸۸.۳% versus ۸۶.۲%).Conclusions: Lightweight CNNs can achieve diagnostic performance comparable to large pre-trained networks while requiring less training time and computational power. The proposed model enables automated, efficient, and clinically feasible detection of the M۳–IAC relationship on panoramic radiographs.

کلیدواژه ها:

Artificial intelligence ، Computer-assisted image processing ، Convolutional neural networks ، inferior alveolar canal ، mandibular third molar ، panoramic radiographs