Journal DOI 10.68956/journal.human-performance-futures

Creative Commons source reading · Data quality and algorithm evaluation

Automated Detection of Positioning Errors in Bitewing Radiographs Using Deep Learning

Katharina Alves Rabelo; Zimo Huang; Eduardo Delamare; Shwetha Hegde; Jinman Kim; Vesna Miletic

Originally published in Clin Exp Dent Res · 2026-12-01

From the article

Selected source text. JournalOS removes markup and shortens excerpts where indicated. The full article provides the complete context.

Abstract

ABSTRACTObjectivesBitewing radiographs (BWs) are essential diagnostic tools in dentistry, but are often affected by positioning errors, compromising their diagnostic value and increasing retake rates. This study aimed to develop and evaluate a deep learning‐based automated system for detecting and classifying BW type (premolar/molar), side (right/left), and common positioning errors in a controlled mannequin‐based setting.MethodsA total of 403 BWs were collected from dental student assessments acquired from dental radiology mannequins. BWs were consensus‐labeled by dentomaxillofacial radiology experts for type, side, and the most common positioning errors. A balanced data set supported training and validation. Convolutional neural network architectures, including ResNet and DenseNet variants, were trained from scratch and evaluated as…

Results

The performance of the CNN classifier DenseNet169, which distinguishes between MBWs and PMBWs (the BW type classification task), and the performance of the DenseNet121 model, which categorizes BWs as right or left (the BW side classification task), is summarized in Table 2. The CCE was evaluated through two classification tasks: binary and multi‐class. The binary classification task aimed to distinguish between the absence and presence of CCE, while the multi‐class classification task further categorized the extent of CCE in the ROI into three classes: Minimal, Severe, and Critical. The CCE binary classification task using DenseNet161 showed excellent performance, achieving an overall accuracy of 96.3%. For the multi‐class classification, the DenseNet121 model achieved an overall accuracy of 79.3%, with the highest performance…

Discussion

The CNN classification models performed well in binary classification tasks for BW type (premolar and molar), side (right and left), and the positioning errors: CCE (presence and absence) and interproximal overlap (presence and absence), achieving accuracies of 95.8%, 99%, 96.3%, and 93.4%, respectively. Recent studies on quality enhancements of intraoral radiographs using AI algorithms have primarily focused on post‐processing techniques (Moran et al. 2022; Latke and Narawade 2024). However, while such models enhance image clarity, they do not address positioning errors, which remain a significant cause of non‐diagnostic radiographs. This study focuses on positioning errors such as CCE, interproximal overlap, and improper receptor placement, as they are among the most prevalent causes of unacceptable image quality in BW radiographs,…

Editorial reading guide

  • What data created and tested the model, and was evaluation independent?
  • What is the cost of a false positive, false negative or out-of-distribution decision?