Deformable contour models for digitizing a printed brainstem atlas

摘要

The brainstem is a part of the brain that is connected to the cerebrum and the spinal cord. Ten out of twelve pairs of cranial nerves emerge from the brainstem. The cranial nerves transmit information between the brain and various parts of the body. Due to its anatomical and physiological relevance, a descriptive digital brainstem is important for neurosurgery planning and simulation. For both of these neurosurgical applications, the complexity of the brainstem requires a digital atlas approach to segmentation that maps intensities to tissues rather than less descriptive voxel or surface-based approaches. However, a descriptive brainstem atlas with adequate details for neurosurgery planning and simulation has not been developed to date. Fortunately, various textbooks contain 2D representations of the brainstem at various longitudinal coordinates. The aim of this thesis is to describe a minimally supervised method to segment sketches coinciding with slices of the brainstem featuring labeled contours. This thesis also describes a deformable contour model approach, emphasizing a 1-simplex framework, to reconstruct a 3D volume from 2D slices.