Approach towards Automatic Segmentation of Diaphragm from Ultrasound Images

Diaphragm helps in diagnosing various abnormalities and diseases involving other human organs. Ultrasound imaging being cheap, portable and does not involve ionized radiation, is preferred over other imaging modalities for imaging diaphragm. In this paper method for the automatic segmentation of diaphragm from ultrasound images is proposed. Difference in echogenicity of the diaphragm with respect to neighboring tissues and the position of the diaphragm in the ultrasound images are the two facts which are used in the proposed method for the automatic segmentation of Diaphragm. Proposed method creates a set of binary images using different threshold values that lies between minimum and maximum intensity present in the ultrasound image. Difference between any two threshold values is in the multiples of predefined constant, a. From each binary image initially all possible diaphragm look -alike curved objects are detected and finally based on the positions of the detected objects, exact position of the diaphragm is obtained automatically. Proposed method, tested on ultrasound images of fatty liver and healthy volunteers, is able to segment diaphragm from ultrasound images with high accuracy and without any user input.