Comparison of Sensitivity and Specificity of Threshold Segmentation Method over Canny Edge Segmentation for Brain Tumor MRI Images

Abstract

To review and explain the performance of currently available threshold segmentation for the segmentation of Brain Tumor MRI Images and calculate the sensitivity and specificity compared with the canny edge segmentation. A total of 300 Brain Tumor MRI images are collected and samples are divided into training dataset <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\boldsymbol{(\mathrm{n}=210\ [70\%]})$</tex> and test dataset <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(\mathbf{n}=90$</tex> [30%]. Segmentation is performed by Threshold Segmentation and Canny Edge Segmentation methods using Sensitivity and Specificity. Brain Tumor MRI image segmentation is done to extract the quantitative information by using Threshold Segmentation and has accuracy of 96% and Canny Edge has attained 93%. Threshold segmentation and Canny Edge Segmentation are statistically significant with the independent sample T-Test value <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(\mathrm{P}\unicode{x00A1}{0.05})$</tex> . The results proved that the threshold segmentation has better efficiency over the Canny Edge for better segmentation of Brain Tumor MRI images.