Scholarlink

In this paper, we propose a comprehensive comparative study of Convolutional Neural Networks (CNNs) and YOLOv8 detections for the teeth caries detection from the Dataset Ninja Dental AI dataset. Rapidly, deep learning helps with dental caries identification and is now being useful to dental caries detection which is crucial for early diagnosis and treatment (Shenoy and Kotekar in J Healthc Eng 2021: Article ID 8741936, 2021). To the best of our knowledge, we comprehensively benchmarked multiple YOLOv8 variants (YOLOv8n, YOLOv8s, YOLOv8m, YOLOv8l, YOLOv8xl) against a standard CNN model used as comparison in this work. Key metrics such as accuracy, precision, recall, F1-score, sensitivity, specificity, and inference time are evaluated for these models from multi-dimensional perspective to see how they perform these models are evaluated. An accuracy of 85.07%, precision of 78.57%, and specificity of 99.15% demonstrates that CNN has the strength of minimizing false positives. However, its recall was 15.49% which was lower than that possessed by other methods, yet was not able to capture all the positive cases. On the other hand, the YOLOv8xl model performed excellently (with an accuracy of 91.33%, recall of 50.89%, and specificity of 93.91%) at the speed and detection efficiency balance. YOLOv8m was drawn to be a strong contender (93.67% accuracy, 51.67% recall), balancing performance with computational efficiency more than reasonably. It is shown that YOLOv8 models are robust to real tense dental caries detection, while CNN models still have an edge in terms of precision for more precise use cases. This study provides crucial insights into advancing AI based diagnostics in dentistry toward clinical applications worldwide. In this sense, this research will make a significant added value to the International discourse on AI in healthcare (Hossain et al. in Health Inform J 26:1904–1917, 2020) and accelerate the innovations in the automated dental diagnostics.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Benchmarking CNN and YOLOv8 Models for Dental Caries Detection Using the Dataset Ninja Dental AI

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Devananda Reddy,
  • Masku Rahul,
  • Hemanth Thatavarthi,
  • S. Balamithra,
  • Harisudha Kuresan

摘要

In this paper, we propose a comprehensive comparative study of Convolutional Neural Networks (CNNs) and YOLOv8 detections for the teeth caries detection from the Dataset Ninja Dental AI dataset. Rapidly, deep learning helps with dental caries identification and is now being useful to dental caries detection which is crucial for early diagnosis and treatment (Shenoy and Kotekar in J Healthc Eng 2021: Article ID 8741936, 2021). To the best of our knowledge, we comprehensively benchmarked multiple YOLOv8 variants (YOLOv8n, YOLOv8s, YOLOv8m, YOLOv8l, YOLOv8xl) against a standard CNN model used as comparison in this work. Key metrics such as accuracy, precision, recall, F1-score, sensitivity, specificity, and inference time are evaluated for these models from multi-dimensional perspective to see how they perform these models are evaluated. An accuracy of 85.07%, precision of 78.57%, and specificity of 99.15% demonstrates that CNN has the strength of minimizing false positives. However, its recall was 15.49% which was lower than that possessed by other methods, yet was not able to capture all the positive cases. On the other hand, the YOLOv8xl model performed excellently (with an accuracy of 91.33%, recall of 50.89%, and specificity of 93.91%) at the speed and detection efficiency balance. YOLOv8m was drawn to be a strong contender (93.67% accuracy, 51.67% recall), balancing performance with computational efficiency more than reasonably. It is shown that YOLOv8 models are robust to real tense dental caries detection, while CNN models still have an edge in terms of precision for more precise use cases. This study provides crucial insights into advancing AI based diagnostics in dentistry toward clinical applications worldwide. In this sense, this research will make a significant added value to the International discourse on AI in healthcare (Hossain et al. in Health Inform J 26:1904–1917, 2020) and accelerate the innovations in the automated dental diagnostics.