A novel lung extraction approach for LDCT images using discrete wavelet transform with adaptive thresholding and Fuzzy C-means clustering enhanced by genetic algorithm

Purpose: Lung cancer is the second most common type of cancer prevalent in men worldwide. The early diagnosis of lung cancer can reduce cancer-related deaths considerably and increase the survival rate for a few years. In recent years, computer-aided detection (CADe) and computer-aided diagnosis (CADx) systems emerged as promising techniques for radiologists for the early diagnosis of lung cancer. An efficient pre-processing technique and accurate segmentation of the lung parenchyma in medical images will reduce false positives, which in turn can considerably improve the specificity of classification of the lung nodules as benign or malignant. Methods: A novel framework for preprocessing lung images and segmentation of the region of interest is proposed in this study. The noise removal in low-dose computed tomography (LDCT) images is performed by discrete wavelet transform with adaptive thresholding (DWTWAT). The segmentation of the lung region is performed by genetic algorithm enhanced K-means clustering (GAK-means) and genetic algorithm enhanced Fuzzy c-means clustering (GAFCM) in LDCT images. The segmentation is followed by lung reconstruction to preserve the juxta-pleural and Pleural tail nodules attached to the lung boundary. Results: The proposed methods were individually evaluated with the commonly used metrics of sensitivity, specificity, and accuracy. The novel noise removal technique of DWTWAT and segmentation with GAFCM has achieved a sensitivity, specificity, and accuracy of 99.54%, 99.99%, and 99.54%, respectively. The noise removal technique of DWTWAT for preprocessing and segmentation with GAK-means has achieved sensitivity, specificity, and accuracy of 99.47%, 99.77%, and 99.26%. Conclusion: The proposed techniques of noise removal and segmentation is a novel combination that showed improved results compared to the existing state-of-the-art method.