ANALYSIS AND REVIEW OF CURRENT DEEP LEARNING TECHNIQUES FOR DENTAL IMAGE SEGMENTATION WITH A NOVEL DEEP NEURAL NETWORK

Abstract

Dental image segmentation is an important task in dental image analysis and plays a crucial role in various dental applications, such as diagnosis, treatment planning, and clinical research. With the advent of different imaging modalities, there has been a significant increase in the complexity and variability of dental images, thus making the dental image segmentation task more challenging. Recent advancements in deep learning techniques have led to significant improvements in dental image segmentation accuracy and efficiency. In this paper, some of the current deep-learning techniques that are used for dental image segmentation on various imaging modalities have been discussed. A comprehensive review is carried out, which provides the details of current deep learning methodologies, including types of deep architecture used, different imaging modalities, and datasets along with their applications and shortcomings. Additionally, a novel deep learning methodology based on a dual stream encoder and decoder architecture is proposed for automatically segmenting panoramic images. The evaluation is carried out on 1000 image dataset and is measured by dice coefficient, jaccard index, accuracy, precision and recall. The proposed methodology performs better than the state-of-the-art deep segmentation models. © The Institution of Engineering & Technology 2023.