Kernel sparse representation based classification of focal liver lesions with contrast-enhanced ultrasound

Contrast-Enhanced Ultrasound (CEUS) is a widespread method for non-invasive diagnosis of Focal Liver Lesions (FLLs). In this paper, an automatic classification algorithm of FLLs based on CEUS imaging is proposed. Firstly, from obtained dynamic CEUS videos of hepatic, Time Intensity Curves (TICs) are extracted. Then TICs' parameters: Peak, TP, Sharpness and AUC are calculated. Both TICs and their parameters are regarded as features in our method. Finally, based on these exeracted features, a Kernel Sparse Representation based Classification (KSRC) algorithm is employed to classify benign FLLs and malignant FLLs. Experimental results show that accuracy, sensitivity and specificity of our proposed method are 98.89%, which outperforms other similar methods.