Multi-Region Weighted Registration Method for Facial Deformation in Transcranial Magnetic Stimulation Navigation

In image-guided TMS navigation, patients’ posture and facial expression result in facial deformation, which finally affects positioning accuracy of the stimulation coil. This paper proposes a multi-region weighted registration method to mitigate the impacts. First, a facial partition model is built. The facial surface is divided into 9 parts based on facial muscles’ distribution. Second, a coarse registration is implemented automatically based on facial anatomical landmarks. Finally, a fine registration is implemented by an improved ICP algorithm where the reciprocal value of part’s surface distance is assigned as the weight value for each part. Experimental results show that this method effectively improves registration accuracy in the presence of local facial deformation.