Efficient Clustering of Brain Tumor Segments using Level-set Hybrid Machine Learning Algorithms

Abstract

Cluster computing is an essential technology in distributed environments for practical data analysis in complex datasets like tumor segmentation, disease classification etc. Today real-world applications like medicine and transport are needed for big data analytics environments. This research article considers complex image data environments like brain tumor segmentation based on advanced clustering techniques for effective tumor prediction. An a-state-of-art analysis used Hierarchical clustering to extract initial tumor segments from the image. The next segment is further refined using novel Noise detection-based level-setting techniques. The unsupervised Fuzzy C-means and k- means clustering is used to segment the diseases affected region to enhance noise detection used in the level set. Effective features are extracted using gray level co-occurrence matrix and redundant discrete wavelet transform. Finally, classifying malignant and benign brain tumor images is done using deep probabilistic neural networks. Publicly available datasets are used to validate the proposed algorithms. Experimental results prove that proposed pipeline techniques have effective performance in tumor segmentation and classification model.