IEEE Access (Jan 2019)
Automated Muscle Segmentation from Dynamic Computed Tomographic Angiography Images for Diagnosis of Peripheral Arterial Occlusive Disease
Abstract
The purpose of this study was to quantitatively evaluating lower leg muscle ischemia measured from dynamic computed tomographic angiography (dyn-CTA) for patients with peripheral arterial occlusive disease (PAOD). A total of 35 patients with known PAOD underwent a dyn-CTA of the lower leg first with 70 kV tube voltage and 30 mL iodinated contrast media. Five minutes later, a standard CTA (s-CTA) of the peripheral runoff from the diaphragm to the toes was scanned. For each of four lower leg artery segments, a runoff score was given by a radiologist according to s-CTA images as a reference standard. The muscle enhancement measured from the dyn-CTA was analyzed by automated muscle segmentation using curve-based Fuzzy C-means (CBFCM) algorithms with three classes for bone, two classes for muscle and one class for fat and background. The muscle enhancement ratio (MER) was calculated for (i) higher enhanced area over total area; and (ii) corresponding average signal value at higher enhanced are over total area. Lower extremities were diagnosed as a normal group (n = 22) with each vessel segment score ≤ 1 and runoff score ≤ 7, and otherwise as an ischemia group (n = 48). The MER for the ischemia group was significantly different (p <; 0.05) than the normal group. There were weak correlations (|r| = 0.47, p <; 0.05) between runoff scores and the MER values. The receiver operating characteristics (ROC) analysis between the two groups had area under the curve of 0.71-0.73. Our study demonstrated that CBFCM could be used for automated muscle segmentation from the dyn-CTA images for qualitatively evaluation of lower leg muscle ischemia.
Keywords
