LUNG LOBE SEGMENTATION AND LUNG CANCER DETECTION WITH HYBRID OPTIMIZATION-ENABLED DEEP LEARNING USING CT IMAGES

LUNG LOBE SEGMENTATION AND LUNG CANCER DETECTION WITH HYBRID OPTIMIZATION-ENABLED DEEP LEARNING USING CT IMAGES SAJEEV RAM ARUMUGAM Department of Artificial Intelligence and Data Science, Sri Krishna College of Engineering and Technology, Coimbatore 641008, Tamilnadu, India https://orcid.org/0000-0002-4151-0539 BALAKRISHNA RAVICHANDRAN Department of Computer Science and Engineering, Vels Institute of Science Technology and Advanced Studies (VISTAS), Chennai 600117, Tamilnadu, India https://orcid.org/0000-0002-0372-2725 DIWAN BASKARAN Department of Computer Science & Engineering, St. Joseph’s College of Engineering, Chennai 600119, Tamilnadu, India https://orcid.org/0000-0001-5449-5289 RAJESH ANNAMALAI Department of Computer Science and Engineering, Vels Institute of Science Technology and Advanced Studies (VISTAS), Chennai 600117, Tamilnadu, India https://orcid.org/0000-0001-5435-0629

One of the deadly diseases and a leading cause of death worldwide is lung cancer. Compound tomography (CT) is commonly used to identify tumors and it classifies the phase of cancer in the human body. Detection of cancer disease in the lung at an early stage is quite difficult and is essential to increase the survival patient’s rate. In this research paper, a deep learning (DL)-based optimization approach is developed for the detection of lung cancer and lung lobe segmentation using CT scan images. Initially, an adaptive wiener filter is used to pre-process the input images and the segmentation process is done by the pyramid scene parsing network (PSPNet) classifier which is effectually trained using the developed honey badger golden search optimization algorithm (HBGSO). Grid-based scheme is used to identify lung nodules and then the features are extracted. Finally, lung cancer detection is done by the Shepard convolutional neural networks (ShCNN) classifier and is trained using the proposed fractional HBGSO (FHBGSO) algorithm. The FHBGSO-based ShCNN outperforms with the highest accuracy of 93.4%, f-measure of 91.2% and precision of 89.8%. Thus, lung cancer is detected in the earlier stages by the devised scheme.
02 03 2025 08 2025 2450047 10.1142/S0219519424500477 10.1142/S0219519424500477 https://www.worldscientific.com/doi/10.1142/S0219519424500477 https://www.worldscientific.com/doi/pdf/10.1142/S0219519424500477 10.1016/j.bbe.2019.11.004 10.1109/ACCESS.2020.2992645 10.1002/ima.22769 10.1007/s00521-020-04842-6 10.1109/TII.2020.2972918 10.1109/CCAA.2015.7148560 10.1109/ICICICT54557.2022.9917585 J Healthc Eng Shamim S 6566982 2022 1 10.4018/IJSIR.287544 10.1016/j.patrec.2019.11.014 10.1109/TMI.2019.2947595 10.1016/j.procs.2017.12.016 10.1109/ICAC353642.2021.9697158 10.1002/ima.22515 10.3390/electronics9101634 10.1109/CVPR.2017.660 10.1016/j.matcom.2021.08.013 10.1109/ACCESS.2022.3162853 10.1142/S0218001422520218 10.1109/TIP.2010.2042645 10.1109/TIP.2010.2041397 Int Res J Eng Technol Praveena KS 991 4 8 2017 10.1038/s42256-019-0110-8 Int J Comput Sci Issues Zulpe N 354 9 3 2012 Adv Comput Eng Bhaladhare PR 396529 2014 1 2014 28th Int Conf Neural Inf Process Syst Ren JS 901 1 2015 Int J Ind Eng Comput Rao R 19 7 1 2016 10.1016/j.cie.2021.107250 CMES Comput Model Eng Sci Xiao Y 815 131 2 2022