Adaptive compression of radiography images by using image processing algorithms

In many health organizations Picture Archiving and Communication System (PACS) has become a standard protocol for storing medical images and transmission from one center to another. High resolution images, stored in DICOM 3 format in PACS, require large storage space. Therefore, reducing the image size by preserving diagnostic information has become a need. In this thesis, regions of interest (ROIs) on pelvis radiography are segmented by image processing algorithms and compressed adaptively by importance with two proposed methods. The first method is an algorithm that detects and adaptively compresses ROIs according to their importance when the regions are contoured manually by the radiologist. The second method is an algorithm that detects and adaptively compresses ROIs from the heat map and focus raw data when the eye tracking system is used. Both methods have a structure that can contribute to the residents who specialize in radiology when representing and correct contouring of regions of interest. The outputs are evaluated by using objective and subjective criteria, obtained results of adaptive compression with more than one ROI in pelvis radiography are explained in details and it has been proved that these kind of images can be compressed without losing diagnostic information.