An Automated Computer-aided Diagnosis System for Malignant Tumor Localization from Lung CT Images for Surgical Planning

Locating the cancerous (malignant) tumor is the best way to treat the lung cancer. In-vivo assessment of tumor growth in lungs supports to estimate the cancer threat. This study focused to develop a computer aided detection (CAD) scheme for automatic segmentation of lung lobes and cancerous tumor region from low dose, isotropic computed tomography (CT) images, which may aid the surgical planning for lung cancer treatment. For this retrospective study, CT scan images of 18 cancerous south Indian subjects (confirmed through biopsy test) aged between 22 to 81 years were analysed. Initially, the original CT image was preprocessed and lung lobes were segmented by adaptive fissure sweep and Dual Tree Complex Wavelet Transform (DTCWT). After processing through spatial fuzzy clustering with level set approach, malignant tumor was segmented. Lastly, the segmented malignant tumor was placed over lobes to display its actual position. Two radiologists were appointed to manually segment the lobes and malignant tumor from the CT lung slices of all the 18 cancerous subjects. To validate the result, cancerous tumor in those CT slices were marked manually by the independent radiologists and taken as ground tooth image. The outcomes suggests that the developed CAD system can detect the cancerous tumor location and thereby may help the surgeons to plan for surgery.