Decision Support System for Liver Cancer Diagnosis using Focus Features in NSCT Domain

Abstract

Diagnosis of liver cancer by medical experts using imaging modalities is found to be sub-optimal as different lesions exhibit similar visual appearance in the spatial domain. Thus computer aided diagnostic tools play a significant role in providing a decision support system for radiologists to minimize the risk of false diagnosis. This paper proposes a different feature set using focus operators for classifying different classes of liver cancer. As computation of focus measure involves the local neighborhood of pixel, focus operator is believed to indirectly measure the intricate texture details of the image. This knowledge of focus operator is exploited in NSCT domain to capture the directional components as feature variables replacing the classic texture features. The results in terms of classification accuracy and kappa coefficient proclaim that the focus operators can be employed as feature variables for classification scenario as it outperforms the state-of-the art texture features.