Comparison of Faster R-CNN, YOLO, and SSD for Third Molar Angle Detection in Dental Panoramic X-rays

Abstract

The use of artificial intelligence algorithms (AI) has gained importance for dental applications in recent years. Analyzing AI information from different sensor data such as images or panoramic radiographs (panoramic X-rays) can help to improve medical decisions and achieve early diagnosis of different dental pathologies. In particular, the use of deep learning (DL) techniques based on convolutional neural networks (CNNs) has obtained promising results in dental applications based on images, in which approaches based on classification, detection, and segmentation are being studied with growing interest. However, there are still several challenges to be tackled, such as the data quality and quantity, the variability among categories, and the analysis of the possible bias and variance associated with each dataset distribution. This study aims to compare the performance of three deep learning object detection models-Faster R-CNN, YOLO V2, and SSD-using different ResNet architectures (ResNet-18, ResNet-50, and ResNet-101) as feature extractors for detecting and classifying third molar angles in panoramic X-rays according to Winter's classification criterion. Each object detection architecture was trained, calibrated, validated, and tested with three different feature extraction CNNs which are ResNet-18, ResNet-50, and ResNet-101, which were the networks that best fit our dataset distribution. Based on such detection networks, we detect four different categories of angles in third molars using panoramic X-rays by using Winter's classification criterion. This criterion characterizes the third molar's position relative to the second molar's longitudinal axis. The detected categories for the third molars are distoangular, vertical, mesioangular, and horizontal. For training, we used a total of 644 panoramic X-rays. The results obtained in the testing dataset reached up to 99% mean average accuracy performance, demonstrating the YOLOV2 obtained higher effectiveness in solving the third molar angle detection problem. These results demonstrate that the use of CNNs for object detection in panoramic radiographs represents a promising solution in dental applications.

Keywords: artificial intelligence; convolutional neural networks; dentistry; third molars angle detection.

Figure 1

Conceptual map for a search result for third molar detection literature from 2017 to August 2024.

Figure 2

History of published documents based on the implementation of AI in dentistry from 2017 to August 2024.

Figure 3

Relationship between published documents related to AI and dentistry and countries of origin from 2017 to August 2024.

Figure 4

Proposed methodology for third molar angle detection in X-rays using Winter’s criterion. We used Faster R-CNN, YOLO V2, and SSD combined with ResNet-18, ResNet-50, and ResNet-101 feature extractor CNNs.

Figure 5

Proposed methodology Workflow for third molar angle detection in X-rays using Winter’s criterion.

Figure 6

Distribution of third molar angle to Data Acquisition. (a) Distoangular, (b) Horizontal, (c) Mesioangular, (d) Vertical.

Figure 7

YOLO V2 Pipeline for third molar detection in panoramic X-rays.

Figure 8

Precision and recall representation for the third molar angle detection in dental panoramic X-rays context.

Figure 9

Precision-Recall (PR) Curve for the third molar angle detection in dental panoramic X-rays context. (a) Results for the third molar angle detection at different threshold values. (b) A sample of the Precision-Recall (PR) Curve for one category (vertical).

Figure 10

Best precision-recall curves for the obtained algorithms using ResNet-18. Training results (left), validation results (center), and testing results (right). (a) Test 2—Faster R-CNN, (b) Test 10—YOLO V2, (c) Test 12—SSD. YOLO V2 using ResNet-18 obtained the best result in test 10 (training: 99%, validation: 90%, and testing: 96%).

Figure 11

Third molar angle detection in dental panoramic X-rays with ResNet-18 as feature extractor for Faster R-CNN, YOLO v2, and SSD.

Figure 12

Best precision-recall curves for the obtained algorithms using ResNet-50. Training results (left), validation results (center), and testing results (right). (a) Test 11—Faster R-CNN, (b) Test 18—YOLO V2, (c) Test 9—SSD.

Figure 13

Third molar angle detection in dental panoramic X-rays with ResNet-50 as feature extractor for Faster R-CNN, YOLO V2, and SSD.

Figure 14

Best precision-recall curves for the obtained algorithms using ResNet-101. Training results (left), validation results (center), and testing results (right). (a) Test 5—Faster R-CNN, (b) Test 11—YOLO V2, (c) Test 12—SSD.

Figure 15

Third molar angle detection in dental panoramic X-rays with ResNet-101 as feature extractor for Faster R-CNN, YOLO V2, and SSD.