| MS TSA Meeting

Name of the Speaker: Ms. C SNEHA SREE (EE21S049)
Guide: Prof. Mohanasankar S
Online meeting link: https://meet.google.com/ecr-xgxx-bby
Date/Time: 17th May 2024 (Friday), 9:00 AM
Title: 3D Segmentation Error Estimation in Micro-CT Imaging Using Transformer-Based Graph Structure Learning.

Abstract :

Medical image segmentation plays an important role in extracting meaningful information from complex medical images such as computed tomography (CT), Magnetic Resonance Imaging (MRI), facilitating accurate analysis and diagnosis. It involves partitioning a medical image (2D or 3D) into distinct regions and delineating structures such as organs, tissues, or anomalies, which can be done manually, semi-automatically, or fully-automatically. The primary objective is to enhance the precision of medical image interpretation, aiding clinicians in treatment planning and decision-making processes. Although many segmentation networks have been proposed for 3D volumetric segmentation of tumors and organs, they fail to produce precise results. In case of errors generated by these networks, clinicians would have to edit the generated segmentation maps manually.

This demands a methodology that can deal with the complexities arising from segmentation errors in medical imaging and a robust network to estimate these errors. To this end, we propose a novel methodology designed to identify and quantify inaccuracies within the segmentation map of a given 3D volume. By employing this approach, we can pinpoint and measure erroneous regions within the segmentation map, offering a comprehensive analysis. Furthermore, our method is capable of estimating errors at individual points or nodes within a 3D mesh generated from a potentially erroneous volumetric segmentation map. This approach serves as an effective Quality Assurance (QA) tool, enhancing the reliability and accuracy of segmentation outcomes in medical imaging applications.