Machine learning for craniomaxillofacial landmark digitization of 3D imaging

Craniomaxillofacial (CMF) deformities (including congenital and acquired deformities of the head and face) seriously affect patients' daily lives. For computer-aided diagnosis and treatment planning, cone-beam computed tomography (CBCT) is typically used to scan CMF patients, where anatomical landmarks are digitized to quantitatively assess the CMF anatomy. This chapter presents the latest machine learning methods for CMF landmark digitization of 3D CBCT images. Specifically, four methods for landmark digitization are described, including (1) a multi-atlas-based method, (2) a regression forest-based approach, (3) a segmentation-guided regression forest method, and (4) a deep learning approach. Experimental results demonstrate that machine learning-based methods help boost the digitization performance of anatomical landmarks for CMF patients.