Deep learning models compared for dental X ray image segmentation
Deep Learning-based Semantic Segmentation of Dental Radiographs: Comparative Evaluation of PSPNet, DeepLabv3, UNet, and UNet++
Computer Vision and Pattern RecognitionArtificial Intelligence
Summary
Analyzing dental X rays helps dentists find problems, but the images can be hard to interpret because they are unclear or noisy. The authors compared four popular deep learning methods that can automatically outline important parts in dental images to see which works best. They tested the methods fairly on 1269 images, adjusting image size and training batch size to find the best settings. They found that a model called UNet++ performed the best in accurately segmenting parts of the dental images. This research helps improve automated tools for dental care by showing which methods give the best results.
What this means in practice
- •For dental imaging software developers: Improve automated dental radiograph analysis by incorporating the most accurate deep learning segmentation model identified here.
- •For medical device manufacturers: Develop enhanced dental diagnostic tools that automatically segment X rays for better clinical decision support using the optimal model configuration found.$Commercial implications: Enables commercial dental diagnostic devices with improved image analysis for dentists by leveraging superior deep learning models.
Authors
Malithi Nawarathne, Nadeesha Herath Mudiyanselage Lihini Sangeetha
Abstract
Accurate and efficient analysis of dental radiographs has become increasingly important for disease diagnosis and clinical decision making. In particular, dental radiographs segmentation plays a critical role as it is challenging due to several reasons, such as low contrast, noise in the images, variations in the image quality, etc. Previous studies have primarily addressed these challenges using convolutional neural networks (CNN)-based models. However, direct comparison of these segmentation models is still challenging due to the variations among the experimental settings. Therefore, we mainly compared four segmentation models, including Pyramid Scene Parsing Network (PSPNet), DeepLabV3, UNet, and UNet++, through a fair, controlled experimentation setting under varying conditions such as image sizes and batch sizes. For this, we used 1269 dental radiography images, and prior to training, we implemented a series of preprocessing techniques. To assess the models, we used several metrics, including Dice coefficient (DICE), Intersection over Union (IoU), accuracy, precision, training and validation losses. According to our results, UNet++ demonstrated the highest IoU score of 0.6719, and a DICE value of 0.7869, showing superior results in terms of segmentation, specifically when configured with an image size of 300x300 and a batch size of 16. In addition, we also explored the effect of image resolutions and batch sizes on model performance and accuracy under different architectures. Therefore, this study provides insights into the strengths of different segmentation models and the potential of deep learning-based semantic segmentation methods for improving automated dental radiograph analysis.