Spatial probabilistic distribution map based 3D FCN for visual pathway segmentation
Abstract—Image-guided surgery has become an important aid in sinus and skull base surgery. In the preoperative planning stage, vital structures, such as the visual pathway, must be segmented to guide the surgeon during surgery. However, owing to the elongated structure and low contrast in medical images, automatic segmentation of the visual pathway is challenging. This study proposed a novel method based on 3D fully convolutional network (FCN) combined with a spatial probabilistic distribution map (SPDM) for visual pathway segmentation in magnetic resonance imaging. Experimental results indicated that compared with the FCN that relied only on image intensity information, the introduction of an SPDM effectively overcame the problem of low contrast and blurry boundary and achieved better segmentation performance.
