Abstract:
Anatomical sketching is widely used in medicine. Physicians often draw sketches in patient records to describe pathology. Medical students frequently use sketches in learning anatomy and communicating in group problem solving. If a computer has the ability to recognize such sketches, it would greatly benefit many medical applications such as automated patient records and education software in medicine. In this thesis, we attempt to develop a system that is capable of recognizing anatomical sketches. We present a method for recognizing sketches that consists of two processes. The first process is an algorithm used for comparing the similarity between a sketch and pre-defined set of templates. The second process focuses on the template selection using the principle of classification. The accuracy of the proposed method is evaluated over a number of sketches drawn by medical students. To measure the effectiveness of the approach, the accuracies of the recognition between the system and human experts are compared. Finally, a prototype system is developed to demonstrate the potential use of anatomical sketch recognition in practice.
