Liver and Liver Tumor Segmentation Using Modified Encoder Decoder Network and Find the Tumor Geometry
Harini G., R. N. Karthika · 2023
Segmentation of images is necessary for the diagnosis of diseases to know about the detailed content of an image. Liver and liver tumor segmentation is a difficult process due to the complexity of the organ position. In this paper, liver and liver tumor segmentation is performed using architectures like Pyramid Scene Parsing Network (PSP Net) and Modified Encoder Decoder Network (MED Net) using 3D Image Reconstructed for Comparison of Algorithm Database (3D-IRCADb) which includes Computed Tomography (CT) images of the liver. After segmenting the images, preprocessing techniques like median blur, dilation, and erosion is performed on the segmented images. Finally, the area of contours is calculated for finding the tumor geometry of the liver. Therefore, the results obtain for this work are 93.62% and 94.92% accuracy for PSP Net and MED Net respectively in the case of liver segmentation and 95.57% and 96.42% accuracy respectively in the case of liver tumor segmentation.