Early detection of lung infection in CT images using clustering algorithms controllers: Fuzzy c-means, Gaussian mixture model, and k-means-based feature extraction

Recently, imagery of the chest has become the key clinical procedure for diagnosing and predicting chest infection in the lungs. Computed tomography (CT) images of the chest were thus considered in this study as a screening strategy for early-stage detection of chest infections and other abnormalities in the human lung. Raw computed tomography is difficult to interpret, leading to a need to develop computer algorithm diagnostic (CAD) approaches to improve the detection of abnormalities in the resulting CT images. The data samples used in this paper were obtained from Al-Hussein Teaching Hospital and the Radiology Department of Imam Al Hujja Charity Hospital in Iraq, Kerbala. The number of images in the assembled dataset was 150 across two different class types, normal and with confirmed lung disease. The Fuzzy C-Means Clustering (FCMC), K-Means Clustering (KMC), and Gaussian Mixed Model Clustering (GMMC) techniques were then applied to the chest CT images to test the detection and classification of the normal and infected scans. The features of the CT images in this paper were then filtered to remove clusters identified as belonging to normal areas to develop a full algorithm for the identification of abnormal areas by mans of anatomic segmentation of chest infection Region of Interest (ROI), which is presented as part of this work. The experimental results show that the proposed segmentation techniques offered clustering accuracy of 93.12 %, 90.23%, and 91.41% for FCMC, KMC, and GMMC, respectively. The performance metrics thus show that the FCMC algorithm outperforms the KMC and GMMC algorithms, as well as being fully autonomous, and having the capability to isolate abnormal infected regions in the lung tissue where such anomalies exist accurately to the benefit of radiologists, a function few other computational algorithms can offer.