Muscular atrophy and dysfunction commonly follow surgical intervention and prolonged immobilization. Consequently, quantitation of muscle volume and cross-sectional area from magnetic resonance images could prove useful in the diagnosis and therapy of muscle injury or disease. The particular aim of our study was to obtain volumes and cross-sectional areas from the muscles of the human calf, specifically the lateral gastrocnemius, medial gastrocnemius, and the soleus. In order to measure volumetric data of human skeletal muscle, we have developed an interactive computer program to segment three-dimensional MRI data according to distinct muscle groups. The program permits the interactive selection of standard image processing operations, including pixel intensity threshholding, seed growing, boundary tracing, and morphological erosion and dilation. Segmentation of a data set can be performed either as a sequence of 2D operations on each slice, or with 3D operations on the entire volume or specified sub-volume. In order to validate the accuracy of the algorithm, phantom studies were imaged on a GE Signa console at 1.5T. The segmentation results from multiple operators (n=10) were compared to the true volumes as determined by weight. Agreement was within 1.90% (1.17%). The same imaging protocol was then repeated on two human volunteers and segmented by multiple operators (n=4) for the muscle groups detailed above. The maximum variation between operators in estimated muscle volume was found to be 5.19%.
Showing the most recent 10 out of 414 publications
