Fig 1.
The Block diagram of the proposed framework for muscles/fat segmentation and quantification based on MRI 3-D volumes.
Fig 2.
An Example for applying LCDG algorithm MRI 3-D volumes.
LCDG algorithm output on (a) exemplary 3D FS-MRI image data; (b) probability density functions of the image voxels in Fig 2A, as determined empirically, and as approximated via LCDG using two dominant DGs; (c) the deviations (standard and absolute) between the empirical and estimated marginal probability density functions in Fig 2B; (d) LCDG algorithm output on the dominant and subordinate DGs in the image data in Fig 2A; (e) the final estimated LCDG model of the empirical density function; and (f) the final LCDG output of the conditional probability density functions of light tissue (muscle) and dark tissue (fat) intensities and the empirical density function.
Fig 3.
Segmentation accuracy measures.
(a) In the segmentation quality measurements, there are 4 regions to be considered as: True positive (TP), false positive (FP), true negative (TN), and false negative (FN). (b) The calculation of the HD between the red line X and the blue line Y.
Fig 4.
Examples for the utilization of LCDG to segment the soft tissue volumes.
(a) From left to right: gray scale MR images for FS+WS, WS and FS; (b) From left to right: binary mask of total thigh area, total fat and total muscle area; (c) From left to right: steps for segmenting the bone and bone marrow; (d) 3-D representations of the segmentation results for SCI (left) and ND (right) thigh; Grey: Muscle area, Yellow: SAT, Blue: IMAT, Red: bone.
Table 1.
Accuracy measures for adipose tissue and total muscle area using LCDG method.
Fig 5.
Examples of muscle group segmentation algorithm for four SCI subjects.
(a) original cross sectional MR image; (b) automatic segmentation of muscle groups: blue area is extensor, red is flexor and yellow presents the medial compartment; (c) manually segmented muscle groups (cyan lines) overlaid on automatic segmentation for comparison; and (d) 3-D representation of automatic segmentation of muscle groups: blue volume is extensor, red is flexor and yellow presents the medial muscle group.
Table 2.
Accuracy measures for segmenting three muscle compartments using the proposed method, ANTs and STAPLE.
Fig 6.
The boxplot representation of the calculated volumes and ratios for manual (black) and automatic (red) segmentation results.
(a) Extensor volume; (b) Flexor volume; (c) Medial volume; (d) IMAT volume; (e) SAT volume; and (f) Total muscle volume.