Segmentation and Analysis of Diaphragm from Ultrasound Images

Abstract

Main aim of image processing is to extract important data from images. Using this extracted information we can describe, interpret and understand the scene provided by the machine. The main aim of image processing is to modify images in desired manner using several techniques. It is generally found that size of medical images is so large due to high resolution so it is difficult to analyze them. Also, radiologists find it difficult to segment the desired part out of ultrasound images. This process is highly cumbersome and time consuming. In order to simplify the task we can design several algorithms to simplify the tasks. In this project, we are dealing with segmentation of diaphragm from ultrasound image which is a muscle that separates the chest (thoracic) cavity from the abdomen. The diaphragm is the main muscle of respiration. Image processing involves lot of steps i.e. image acquisition, image pre-processing and then segmentation take place. We have used one global thresholding technique i.e. Otsu method and one local thresholding technique i.e Niblack method. Finally we used combination of both techniques to yield desired result. We observed the segmented images after implementing these techniques. In order to check the performance we calculated and analyzed the SNR values and proposed the best method for the segmentation.