This paper reviews methods for computer assisted medical intervention using statistical models and machine learning technologies, which would be particularly useful for representing prior information of anatomical shape, motion, and deformation to extrapolate intraoperative sparse data as well as surgeons' expertise and pathology to optimize interventions. Firstly, we present a review of methods for recovery of static anatomical structures by only using intraoperative data without any preoperative patient-specific information. Then, methods for recovery of intraoperative motion and deformation are reviewed by combining intraoperative sparse data with preoperative patient-specific stationary data, which is followed by a survey of articles which incorporated biomechanics. Furthermore, the articles are reviewed which addressed the used of statistical models for optimization of interventions. Finally, we conclude the survey by describing the future perspective.

The accuracy of non-rigid 3D face recognition is highly influenced by the capability to model the expression deformations. Given a training set of non-neutral and neutral 3D face scan pairs from the same subject, a set of Fourier series coefficients for each face scan is reconstructed. The residues on each frequency of the Fourier series between the finely aligned pairs contain the expression deformation patterns and PCA is applied to learn these patterns. The proposed expression deformation model is then built by the eigenvectors with top eigenvalues from PCA. Recognition experiments are conducted on a 3D face database that features a rich set of facial expression deformations, and experimental results demonstrate the feasibility and merits of the proposed model.

To segment a shape into parts is an important problem in shape representation and analysis. We propose in this paper a novel framework of shape segmentation using deformation models learned from multiple shapes. The deformation model from the target image to every other image is then estimated. Finally, normalized-cut graph partition is applied to the graph constructed based on the similarity of local patches in the target image, and a segmentation of the shape is carried out. Experimental results for images from MPEG7 shape database show the effectiveness of the proposed method.