Title:
Development of an Automatic Region of Interest Extraction Model for Lung Cancer Candidates in CT Scan Images

Authors:
Sri Widodo*, Mohammad Faizuddin Bin MD Noor, Zalizah Awang Long, Indonesia

Abstract:

Lung cancer is the uncontrolled proliferation of cancer cells in lung tissue.  Recently, numerous researchers have started working on lung cancer detection.  The technique for detecting lung cancer begins with selecting the Region of Interest (ROI).   ROI extraction in lung cancer diagnosis requires two steps: lung field segmentation and cancer candidate segmentation. Earlier research on lung cancer detection used manual ROI extraction, generally through cropping.  Automatic segmentation can be hard, especially when separating the lung field from adjacent tissues. If the anomaly is significant and located near the lung's edge, the lung's edge may become obscured.  When segmentation is conducted, the image suspected of being cancerous is excluded from the lung image (the diseased portion of the lung is gone).  Thus, lung field segmentation is regarded to have failed. The suggested study intends to create a Region of Interest (ROI) extraction algorithm for automatic cancer candidate segmentation on CT scan images utilising the Active Shape algorithm and mathematical morphology techniques. The proposed research is separated into two stages: lung field segmentation using the Active Shape Model (ASM), and lung candidate segmentation using the mathematical morphology method. The Active Shape Model technique achieves 97.28% lung segmentation accuracy, with 96% sensitivity and 97.4% specificity.  Meanwhile, the Morphology technique achieves 99.4% accuracy in lung cancer candidate segmentation, with 96.2% sensitivity and 99.7% specificity. 

DOI: http://dx.doi.org/10.51505/ijaemr.2025.1308