Automatic Segmentation and Reconstruction of Extraocular Muscles from Mri Using Superpixel and Normalized Cut

Extraocular muscles (EOMs) enlargement affects the biomechanics of eye movement and is a key factor of several orbital diseases [1]. Identification of EOM enlargement is important for clinical diagnosis and treatment, however, accurate and efficient quantification of EOM anatomy is challenge. We present a fully automatic method to segment and reconstruct 3D model of the EOMs. We design a novel algorithm which uses superpixels (i.e. clusters of pixels) as the basic units for segmentation. After obtaining the segmented EOM boundaries, we reconstruct 3D models of the EOMs. Our proposed method on automatically reconstructing patient-specific EOM models can be applied in clinical diagnosis and surgical planning.