This manuscript presents an automated classification approach to classify lesions into four categories of liver diseases, based on Computer Tomography (CT) images. The four diseases types are Cyst, Hemangioma, Hepatocellular carcinoma (HCC), and Metastasis. The novelty of the proposed approach is attributed to utilizing the difference of features (DOF) between the lesion area and the surrounding normal liver tissue. The DOF (texture and intensity) is used as the new feature vector that feeds the classifier. The classification system consists of two phases. The first phase differentiates between Benign and Malignant lesions, using a Support Vector Machine (SVM) classifier. The second phase further classifies the Benign into Hemangioma or Cyst and the Malignant into Metastasis or HCC, using a Naive Bayes (NB) classifier. The experimental results show promising improvements to classify the liver lesion diseases. Furthermore, the proposed approach can overcome the problems of varying intensity ranges, textures between patients, demographics, and imaging devices and settings.
|Title of host publication||Lecture Notes in Engineering and Computer Science|
|Publication status||Published - 1 Jul 2016|
|Event||World Congress on Engineering 2016 - London, United Kingdom|
Duration: 29 Jun 2016 → 1 Jul 2016
|Conference||World Congress on Engineering 2016|
|Period||29/06/16 → 1/07/16|
- Centre for Intelligent Visual Computing Research