Measuring Spinous Process Angles on US Spine Images using the GVF Segmentation Method

Spinous process angle (SPA) is an important parameter to evaluate the severity of scoliosis. However, the spinous process cannot be accurately automatic located due to the interference of the muscle layer. The objectives of this study are to apply gradient vector flow (GVF) snake model to segment spinous process (SP) on US transverse vertebral images and to illustrate and measure SPA on US coronal spine images. The snake method could detect SP tip position reliably in this study. Ten spinous process curves were identified on both radiographs and US images. The mean absolute difference (MAD) of SPAs obtained from the two modalities was 2.5±1.9°. It demonstrates the SPA measured from US images via GVF snake model is comparable with the results from the conventional radiographs.