3-2-02. Texture Analysis of Muscle Ultrasound can differentiate myopathic conditions

Publication date: June 2017
Source:Clinical Neurophysiology, Volume 128, Issue 6
Author(s): Kazuki Sogawa, Hiroyuki Nodera, Naoko Takamatsu, Atsuko Mori, Yuishin Izumi, Syuji Hashiguchi, Ryuji Kaji
Muscle sonography may show high densities in various myopathic conditions. Further classification among myopathic conditions could be possible by texture analysis, which analyzes relationships between the neighboring pixel densities.The following groups of subjects were assessed: (1) 19 patients with myotonic dystrophy type 1 (MD1), (2) 11 patients with inflammatory myopathy (IM), and (3) 16 normal control. A LOGIQe ultrasonography and an 11-MHz linear probe were used to obtain muscle images of medial gastrocnemius. Texture parameters were obtained by MaZda software. Linear Discriminant Analysis (LDA) was performed for texture-based classification.The mean echodensities of MD1 and IM were higher than those of the control, which yielded high areas under curve (AUC) >0.95. However, MD1 and IM were poorly discriminated by any of the histogram-related parameters (AUC<0.8). LDA yielded 95.7% of correct classification among the three groups. The parameters related to run-length matrix had the highest discriminating power between MD1 and IM. In conclusion, texture analysis of a sonographic muscle image can infer underlying disease mechanisms in myopathic conditions.



http://ift.tt/2qWYHHS