Multilevel Thresholding for Medical Image Segmentation Using TeachingLearning Based Optimization Algorithm

Abstract

Medical image segmentation is the basic pre-processing step to infer information from the input image with RGB color space. In this paper, multilevel thresholding (MLT) with most optimistic objective functions such as Kapur and Otsu are used for image segmentation. But the MLT suffers from high execution time with the increase in number of threshold levels while exploring for optimal threshold. This difficulty is eased by the robust teachinglearning based optimization (TLBO) algorithm. It mimics the classroom environment where the student gains knowledge from the teacher. The main aspect of the TLBO is the use of less algorithm specific parameters in search process and it avoids premature convergence and getting trapped with sub-optimal solutions. The performance of TLBO algorithm is compared with cuckoo search (CS) algorithm at 4, 5, 6 and 7 threshold levels. Experimental results confirm that the Otsu based MLT outperform the Kapur objective function. Exploration and exploitation reveal the fast convergence and are confirmed by metrics such as computational time, peak signal to noise ratio (PSNR) and structural similarity index (SSIM). This affirms the inclusion of TLBO algorithm for precise medical image segmentation.

Random Forest-Based Survival Analysis for Predicting the Future Progression of Brain Disorder from Mild Cognitive Impairment (MCI) to Alzheimer’s Disease (AD)

“…Medical imaging and its applications have been widely attracted the attention of researchers for decades. Different types of scanning techniques and the analysis of them can help physicians in accurately diagnose the disease, not only that but also aid in early detection of dangerous diseases like tumours and AD [9][10][11][12]. The emergence and development of machine learning and deep learning techniques have also shared in the development of this research area.…”

Implementation of Machine Learning Techniques for detection of Lung Cancer

“…The common method for detecting and assessing lung cancer is depicted in [3]. Expert evaluations on various parameters characterizing a nodule's morphology and shape are frequently used in clinical practice to assess the malignancy of lung nodules, but these criteria are primarily objective and randomly defined [13] [15].Various Image Processing techniques using MATLAB [14] as a platform, a number of studies on Lung Cancer Recognition and Classification have been conducted for CT images as input [7].A wide evaluation of the most essential algorithms utilized in the CAD application for lung tissue diagnostics, with emphasis on each algorithm's performance [6]. Some systems to utilize Artificial Neutral Networks to detect lung cancer (ANN) to provide insufficient precision [5].…”