Abstrakt

Liver cancer detection and classification based on optimum hierarchical feature fusion with PeSOA and PNN classifier

Rajesh G, Selwin Mich Priyadharson A

Detection and segmentation of abnormalities present in the liver is one of the testing and vital steps to treatment planning which expands the survival of a patient. Liver cancer expands the death rate because the side effects cannot be distinguished even the cancer is in its advanced stage. The early diagnosis and consistent observing is the best way to control the development of the malignancy and to save the lives. Ultrasound imaging is a standout amongst the most frequently utilized diagnosis instruments to distinguish and classify irregularities of the liver which is also a non-invasive, safe technique for patient examination, being easy to apply, economical than the CT, MRI, PET based liver tumor detection. Traditional liver cancer detection strategies have high computation time and multifaceted nature. In order to diminish the complexity in the computational procedure and to enhance the diagnostic exactness in this paper we propose a new optimum hierarchical feature fusion based on Penguin Search Optimization Algorithm (PeSOA). The features resulting from PeSOA is used by a Probabilistic Neural Network (PNN) which classifies the liver cancer tissues. Test results demonstrate that the proposed technique acquires 98.92% classification accuracy and the correlation demonstrates that we obtained superior results than the existing strategies.