Quantification Of Error In The Manual Segmentation Of Ankle CT Scans

Description

This presentation by Subrata Mondal focuses on the quantification of error in the manual segmentation of ankle CT scans. The speaker introduces the significance of finite element analysis in guiding clinical treatment and planning, particularly in the context of three-dimensional modeling derived from CT or MRI images. He emphasizes the need for precise quantification and evaluation of these 3D models to ensure reliability in biomechanical studies.



The research involves segmenting CT scans from three anonymous individuals, where data is segmented multiple times by both novice and experienced users to assess the reliability and accuracy of different measurements, specifically bone volume, surface area, and distances between various bones in the ankle. Intraclass correlation coefficients (ICC) are utilized to analyze intra-observer reliability, revealing high reliability for bone volume and surface area, but lower values for inter-bone distances, notably between the talus and navicular bones.



Moreover, a Bland-Altman plot is generated to evaluate measurement accuracy, further highlighting the challenges faced in assessing smaller distances. The findings suggest the importance of scan resolution in influencing accuracy and reliability, leading to the conclusion that selecting appropriate CT scan resolutions is crucial for obtaining reliable measurements. The speaker encourages audience engagement by inviting them to further discussions at his poster.

DOI: 10.1302/3114-240484

Specialties