Despeckle filtering: Performance evaluation for malignant focal hepatic lesions
Nimisha Manth, Jitendra Virmani, Vinod Kumar, Naveen Kalra, Niranjan K. Khandelwal · International Conference on Computing for Sustainable Global Development · 2015
The objective of despeckling an ultrasound image is to smoothen the homogeneous areas while preserving the edges and structural information so that the diagnostic features present in the image are highlighted. In the present work, different despeckling methods i.e., methods based on local statistics, median filtering, geometric filtering and anisotropic diffusion have been applied on a comprehensive dataset of 64 B-Mode liver ultrasound images consisting of 19 hepatocellular carcinomas and 45 metastatic carcinomas images. As the sonographic appearance of hepatocellular carcinoma, and atypical cases of metastatic carcinoma are highly overlapping, the differential diagnosis between these malignant focal liver lesions is considered to be a subtle task. Moreover, the subjective analysis is prone to inter and intra observer variability, therefore in the present work, objective analysis has been carried out with respect to (a) Smoothening of homogeneous areas (by computing PSNR metric), (b) Edge preservation (by computing FoM index), and (c) the structure / feature preservation (by computing SSIM index). The exhaustive objective analysis carried out in the present work, indicates that the image processed by Lee Sigma method based on Local Statistics yields the best results with respect to smoothing, edge, and structure/feature preservation.