Atlas-based non-rigid image registration to automatically define line-of-action muscle models: A validation study |
| |
Authors: | Lennart Scheys Dirk Loeckx Arthur Spaepen Paul Suetens Ilse Jonkers |
| |
Affiliation: | 1. School of Allied Health Sciences, Menzies Health Institute Queensland, Griffith University, Gold Coast, Australia;2. Centre for Musculoskeletal Research, Menzies Health Institute Queensland, Griffith University, Gold Coast, Australia;3. Queensland Children''s Gait Laboratory, Queensland Paediatric Rehabilitation Service, Children''s Health Queensland Hospital and Health Service, Brisbane, Australia;4. Queensland Cerebral Palsy and Rehabilitation Research Centre, School of Medicine, The University of Queensland, Brisbane, Australia |
| |
Abstract: | Research has raised a growing concern about the accuracy of rescaled generic musculoskeletal models for estimating a subject's musculoskeletal geometry. Information extracted from magnetic resonance (MR) images can improve the subject-specific detail and accuracy of musculoskeletal models. Nevertheless, methods that allow efficient, automated definition of subject-specific muscular models for use in biomechanical analysis of gait have not yet been published to the best of our knowledge. We report a novel method for automated definition of subject-specific muscle paths using non-rigid image registration between an atlas image and the subject's MR images. We validated this approach quantitatively by measuring the distance between automatically and manually defined coordinates of muscle attachment sites. Data was collected for 34 muscles in each lower limb of 5 paediatric subjects diagnosed with diplegic cerebral palsy and presenting varying degrees of increased femoral anteversion. Distances showed an overall median Euclidean error of 6.1 mm: 2.0 mm along the medio-lateral direction, 1.8 mm along the anterior–posterior direction and 3.8 mm along the superior–inferior direction. A qualitative validation between automatically defined muscle points and the muscular geometry observed in the subject's medical image data corroborated the quantitative validation. This automated approach followed by visual inspection and, if needed, correction to the muscle paths, reduced the time required for defining 34 lower-limb muscle paths from around 3.5 to 1 h. Furthermore, the method was also applicable to aberrant skeletal geometry. Using the proposed method, defining MR-based musculoskeletal models becomes a time efficient and more accurate alternative to rescaling generic models. |
| |
Keywords: | |
本文献已被 ScienceDirect 等数据库收录! |
|