Segmentation of Tumor Region in Multimodal Images Using a Novel Self-organizing Map-Based Modified Fuzzy C -Means Clustering Algorithm

Abstract

The significant issue for image segmentation that performs in the realization and analysis of the tumor and lesion region in multimodal such as CT and MRI images lies in the computational time and accuracy values. Recently, tumor region extraction from medical image that executes with quick response of evolution process will result in the aid of clinical surgical application. The automatic process can be ensured from unsupervised image segmentation algorithm, as it provides the clear identification of the tissue and lesion region in CT and MRI image. In specific, the unsupervised image segmentation process comprises of self-organization map (SOM)-based Modified Fuzzy C-Means Clustering (MFCM) algorithm that results in exact tumor identification and clear segmentation of tumor involved in organ such as liver, lung, brain, and thorax region. The proposed SOM-based Modified Fuzzy C-Means Clustering algorithm is an approach that refers to enhance the image quality measures such as mean squared error (MSE), peak signal-to-noise ratio (PSNR), Jaccard index, and dice overlap index (DOI). Modified Fuzzy K-Means (MFKM) algorithm and the self-organization map (SOM) based fuzzy K-Means (FKM) algorithm are evaluated, and it was finalized that better results are obtained from SOM-based Modified Fuzzy C-Means Clustering algorithm.