Automatic Clinical Image Segmentation Using Pathological Modelling, PCA and SVM

A general automatic method for clinical image segmentation is proposed. Tailored for the clinical environment, the proposed segmentation method consists of two stages: a learning stage and a clinical segmentation stage. During the learning stage, manually chosen representative images are segmented using a variational level set method driven by a pathologically modelled energy functional. Then a window-based feature extraction is applied to the segmented images. Principal component analysis (PCA) is applied to these extracted features and the results are used to train a support vector machine (SVM) classifier. During the clinical segmentation stage, the input clinical images are classified with the trained SVM. By the proposed method, we take the strengths of both machine learning and variational level set while limiting their weaknesses to achieve automatic and fast clinical segmentation. Both chest (thoracic) computed tomography (CT) scans (2D and 3D) and dental X-rays are used to test the proposed method. Promising results are demonstrated and analyzed. The proposed method can be used during preprocessing for automatic computer aided diagnosis.