DEEP LEARNING APPROACH FOR TOOTH INSTANCE SEGMENTATION ON PANORAMIC DENTAL RADIOGRAPHS USING U-NET
Author(s):
Dr. P. Ammireddy, G. Chiranjeevi, A. Showri Joseph Kumar, A. Siva Kumar, B. Jeevan Harsha
Keywords:
Dental, Panoramic, Segmentation, Counting, Deep learning, U-net.
Abstract
Dental radiography plays a critical role in diagnosing and treating oral health conditions, with panoramic dental radiographs being commonly used for comprehensive assessments. Automating the segmentation of individual teeth within panoramic radiographs is a crucial step towards improving diagnostic accuracy and efficiency. In this study, we explore a deep learning approach tailored specifically for panoramic dental radiographs, aiming to automatically segment teeth using the U-Net architecture. We propose leveraging the U-Net network to achieve precise tooth instance segmentation in panoramic X-ray images. The proposed method achieves an impressive Dice overlap score of 95.4% in overall teeth segmentation. What sets this approach apart is the introduction of a novel post-processing stage that refines the segmentation maps by applying grayscale morphological and filtering operations to the output of the U-Net network before binarization.
The obtained results concludes Deep Learning approach along with innovative post-processing techniques, holds great promise for advancing image analysis in dentistry and beyond, offering potential applications to similar challenges in various domains.
Article Details
Unique Paper ID: 162768
Publication Volume & Issue: Volume 10, Issue 10
Page(s): 959 - 963
Article Preview & Download
Share This Article
Join our RMS
Conference Alert
NCSEM 2024
National Conference on Sustainable Engineering and Management - 2024