Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

An Explainable Two Stage Deep Learning Framework for Pericoronitis Assessment in Panoramic Radiographs Using YOLOv8 and ResNet-50

Ajo Babu George, Pranav S, Kunal Agarwal

TL;DR

This paper tackles the challenge of diagnosing pericoronitis on panoramic radiographs by proposing a two-stage AI framework that first localizes third molars and classifies their Winter’s angulation using YOLOv8, then classifies radiographic signs of pericoronitis with a modified ResNet-50 on ROI crops. Grad-CAM explainability heatmaps accompany predictions to enhance clinical trust, with radiologists reporting 84% alignment between heatmaps and their impressions. The model achieves strong technical performance (Model 1: precision $0.92$, mAP50 $0.925$; Model 2: AUC $=0.94$, F1 normal $=0.88$, pericoronitis $=0.86$) and demonstrates potential to support dental clinicians, particularly in low-resource settings. Limitations include dataset scope, lack of symptom data, and the need for broader validation; future work proposes dataset expansion, multi-class severity predictions, and edge-device optimization to facilitate real-world deployment.

Abstract

Objectives: To overcome challenges in diagnosing pericoronitis on panoramic radiographs, an AI-assisted assessment system integrating anatomical localization, pathological classification, and interpretability. Methods: A two-stage deep learning pipeline was implemented. The first stage used YOLOv8 to detect third molars and classify their anatomical positions and angulations based on Winter's classification. Detected regions were then fed into a second-stage classifier, a modified ResNet-50 architecture, for detecting radiographic features suggestive of pericoronitis. To enhance clinical trust, Grad-CAM was used to highlight key diagnostic regions on the radiographs. Results: The YOLOv8 component achieved 92% precision and 92.5% mean average precision. The ResNet-50 classifier yielded F1-scores of 88% for normal cases and 86% for pericoronitis. Radiologists reported 84% alignment between Grad-CAM and their diagnostic impressions, supporting the radiographic relevance of the interpretability output. Conclusion: The system shows strong potential for AI-assisted panoramic assessment, with explainable AI features that support clinical confidence.

An Explainable Two Stage Deep Learning Framework for Pericoronitis Assessment in Panoramic Radiographs Using YOLOv8 and ResNet-50

TL;DR

This paper tackles the challenge of diagnosing pericoronitis on panoramic radiographs by proposing a two-stage AI framework that first localizes third molars and classifies their Winter’s angulation using YOLOv8, then classifies radiographic signs of pericoronitis with a modified ResNet-50 on ROI crops. Grad-CAM explainability heatmaps accompany predictions to enhance clinical trust, with radiologists reporting 84% alignment between heatmaps and their impressions. The model achieves strong technical performance (Model 1: precision , mAP50 ; Model 2: AUC , F1 normal , pericoronitis ) and demonstrates potential to support dental clinicians, particularly in low-resource settings. Limitations include dataset scope, lack of symptom data, and the need for broader validation; future work proposes dataset expansion, multi-class severity predictions, and edge-device optimization to facilitate real-world deployment.

Abstract

Objectives: To overcome challenges in diagnosing pericoronitis on panoramic radiographs, an AI-assisted assessment system integrating anatomical localization, pathological classification, and interpretability. Methods: A two-stage deep learning pipeline was implemented. The first stage used YOLOv8 to detect third molars and classify their anatomical positions and angulations based on Winter's classification. Detected regions were then fed into a second-stage classifier, a modified ResNet-50 architecture, for detecting radiographic features suggestive of pericoronitis. To enhance clinical trust, Grad-CAM was used to highlight key diagnostic regions on the radiographs. Results: The YOLOv8 component achieved 92% precision and 92.5% mean average precision. The ResNet-50 classifier yielded F1-scores of 88% for normal cases and 86% for pericoronitis. Radiologists reported 84% alignment between Grad-CAM and their diagnostic impressions, supporting the radiographic relevance of the interpretability output. Conclusion: The system shows strong potential for AI-assisted panoramic assessment, with explainable AI features that support clinical confidence.
Paper Structure (20 sections, 6 figures, 3 tables)

This paper contains 20 sections, 6 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Background Study
  3. Methods and Methodology
  4. Dataset Description
  5. Dataset Composition
  6. Model 1: Localisation Model
  7. Model 2: Classification Model
  8. Methodology
  9. Results
  10. Technical Inference
  11. Clinical Inference
  12. Discussion
  13. Conclusion
  14. Declarations
  15. Funding
  16. ...and 5 more sections

Figures (6)

  • Figure 1: End-to-end architecture of the proposed model pipeline showing (A) YOLOv8 localization, (B) ResNet-50 classification, and (C) Grad-CAM explainability.
  • Figure 2: Detection, anatomical localization and pathological classification of pericoronitis of third molars
  • Figure 3: ROC curve of the ResNet-50-based model on the test set, with AUC of 0.94 for classifying pericoronitis in panoramic radiographs.
  • Figure 4: Confusion matrix of Model 2: Classification Model
  • Figure 5: Grad-CAM visualizations of normal cases: (a) Mandibular left, (b) Mandibular right.
  • ...and 1 more figures