Computer-based echotexture comparison of muscles. Exploring innovative analyses of ultrasound images

Image analysis can reliably quantify median nerve echogenicity and texture changes in patients with carpal tunnel syndrome

OBJECTIVE

To study the ability of image analysis measures to quantify echotexture changes of median nerve in order to provide a complementary diagnostic tool in CTS.

METHODS

Image analysis measures (gray level co-occurrence matrix (GLCM), brightness, hypoechoic area percentage using max entropy and mean threshold) were calculated in normalized images of 39 (19 younger and 20 older than 65y) healthy controls and 95 CTS patients (37 younger and 58 older than 65y).

RESULTS

Image analysis measures were equivalent or superior (older patients) to subjective visual analysis. In younger patients, GLCM measures showed equivalent diagnostic accuracy with cross sectional area (CSA) (Area Under Curve (AUC for inverse different moment = 0.97). In older patients all image analysis measures showed similar diagnostic accuracy to CSA (AUC for brightness = 0.88). Moreover, they had abnormal values in many older patients with normal CSA values.

CONCLUSIONS

Image analysis reliably quantifies median nerve echotexture alterations in CTS and offers similar diagnostic accuracy to CSA measurement.

SIGNIFICANCE

Image analysis may offer added value to existing measures in the evaluation of CTS, especially in older patients. Its clinical implementation would require incorporation of mathematically simple software code for online nerve image analysis in ultrasound machines.