Spatial statistics and connected component based lung lobe segmentation and thresholded bounding box-based lung nodule extraction in lung CT scan images.

Lung cancer is one of the life-threatening diseases that records a large number of incidences and a high mortality rate and most often is diagnosed at an advanced stage. Computed Tomography (CT) scans are widely used in lung cancer diagnosis. The image segmentation algorithms help to segment the Region of Interest (RoI) from the lung medical images to accurately identify and isolate lung tissue and nodules from CT scan images. The proposed method for lung CT image segmentation and identification of lung tumors consists of two phases: Spatial Statistics and Connected Component based lung lobe Segmentation (SCS) and Thresholded Bounding box-based lung nodule Extraction (TBE). The phase 1 SCS, uses symmetric division of lung CT image histogram, binarization and complementing process to segment the lung lobe region and phase 2 TBE, uses threshold-based bounding box method to identify the lung nodules. The proposed method is named as SCS-TBE. This SCS-TBE was evaluated on the LIDC dataset. The algorithm's performance was analyzed using Dice Coefficient (DC) and Jaccard Coefficient (JC), with an average DC of 0.93 and an average JC of 0.91, indicating better accuracy segmentation which helps in the identification of lung tumors in CT scan images. The SCT-TBE has outperformed other methods, as evaluated in terms of DC and JC values. The DC and JC values of SCS-TBE are 0.93 and 0.91 respectively. [ABSTRACT FROM AUTHOR]