Application of Geometric Features on Lung Lesion and Non-Lesion Segmentation

• Published in: Proceedings - 13th IEEE International Conference on Control System, Computing and Engineering, ICCSCE 2023
• Main Author: Jafery N.N.; Sulaiman S.N.; Osman M.K.; Karim N.K.A.; Soh Z.H.C.; Abdullah M.F.; Isa I.S.
• Format: Conference paper
• Language: English
• Published: Institute of Electrical and Electronics Engineers Inc. 2023
• Online Access: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85172902165&doi=10.1109%2fICCSCE58721.2023.10237160&partnerID=40&md5=5137268060eb2238572d7489835757c2

Lung cancer is the most prevalent cancer globally, with 1.6 million deaths annually. This is very crucial to develop a new CAD system that will aid medical experts in treating patients and increase the survival rate. Segmenting lung lesions is one of the most important tasks in a CAD system for lung cancer. A new procedure for lung lesion and non-lesion segmentation based on geometric features is presented in this paper with the aim to separate lung lesions and non-lesion from the lung region and to minimise the non-lesion without removing the potential lesion. The procedure's focus is on geometric feature extraction. The procedure was applied to 300 lung CT scan images that were collected from Advanced Medical and Dental Institute (AMDI), Universiti Sains Malaysia. Since the lung CT scans image have low contrast, contrast stretching is used to improve the quality of the image. Diameter and roundness were used as geometric features to provide additional information to reduce the number of non-lesions existing in the segmented image. As a contribution, the average lung lesion and non-lesion segmentation accuracy is 99.70%. When compared to manual delineation by radiologists, the study revealed that 83% of the overall lesion existing in the dataset was successfully segmented using the proposed method. The experiment's findings provided a strong contribution to this research's next phase. © 2023 IEEE.