A New optimized sequential method for lung tumor diagnosis based on deep learning and converged search and rescue algorithm. (July 2021)

• Record Type: Journal Article
• Authors: Tian, Qingji; Wu, Yongtang; Ren, Xiaojun; Razmjooy, Navid
• Is Part Of: Biomedical signal processing and control. Volume 68(2021)
• Publication Date: 2021-07
• Subjects: Lung tumor diagnosis -- Fuzzy C-Ordered means -- Enhanced capsule networks -- Converged search and rescue algorithm (CSAR); Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2021.102761
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: New optimal diagnosis of the lung tumor from CT scan images. Optimized Fuzzy C-Ordered Means (FCOM) for segmentation. The optimization is based on a new converged version of search and rescue algorithm. Classification based on Enhanced Capsule Network (ECN). Abstract: Because the diagnosis of lung cancer and malignancy using imaging techniques such as CT-Scan without the need for sampling reduces the risk of cancer nodules spreading, the development of a computer diagnostic system to process images and lungs and then classify them into two classes of benign and malignant groups plays an important role in the early diagnosis of lung cancer and saving the lives of patients. This study aimed to achieve higher classification accuracy and consequently higher detection accuracy of malignant and benign glands based on deep learning and metaheuristics. In this study, first, the CT scan images of the lung are pre-processed and then the pattern segmented area is achieved by an optimized version of the new fuzzy possibilistic c-ordered mean based on a new version of a metaheuristic, called Converged Search and Rescue (CSAR) algorithm. Then, Enhanced Capsule Networks (ECN) is used for the final diagnosis. To validate the method, it is accomplished to the Lung CT-Diagnosis database and is analyzed based on four indicators including precision, accuracy, recall, and F1-score. The final results of the method are compared with three state-of-the-art methods, including ResNet, KE-CNN, and CNN. The results showed that the suggested method with 96.35 % precision, 96.07 % recall, 96.41 % F1-score, and 96.65 % accuracy has the best results against the compared methods.