Automatic Interpretation of Lung CT Volume Images

Computer-aided systems in medical imaging nowadays play a crucial role to assist clinicians. Research shows that the use of computer-aided systems is indispensable to alleviate the workload of clinicians. In this thesis, a new framework, which interprets lung computed tomography (CT) volume images automatically is proposed to help clinicians. A common interpretation tasks of lung CT volume images involves segmentation and extraction of the organs in the thoracic cavity. The developed framework consists of six main steps to segment and extracts the major parts in the thoracic cavity. In thefirst step, the region of interest (ROI) was determined for following segmentation and extraction tasks. Then in the second step, the large airways were extracted from theROI mask. In the third step, the left and the right lungs were segmented from the ROImask. If the left and the right lungs touched each other, the separation process was performed in the fourth step. After that in the fifth step, mediastinum volume was extracted from the ROI mask. Finally, the vessels tree were segmented from inside of the lungs. The developed framework was tested against 133 computed tomography pulmonaryangiography (CTPA) volume images, and good results have been achieved according to the qualitative evaluation (The average success rate of all steps are over 85% whichgave satisfying results for all segmentation and extraction tasks). Besides, the average execution time for the developed framework is 83.46 seconds per cases and 0.195seconds per slices, which were provided low computational cost according to thecurrent studies and manual interpretation made by clinicians