Lung cancer segmentation from CT scan images using modified mayfly optimization and particle swarm optimization algorithm

• Published in: Multimedia tools and applications Vol. 83; no. 2; pp. 3567 - 3584
• Main Authors: Poonkodi, S., Kanchana, M.
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.01.2024; Springer Nature B.V
• Subjects: Algorithms; Artificial neural networks; Computed tomography; Computer Communication Networks; Computer Science; Critical components; Data Structures and Information Theory; Image contrast; Image enhancement; Image quality; Image segmentation; Lung cancer; Lung diseases; Medical imaging; Model accuracy; Multimedia Information Systems; Optimization; Particle swarm optimization; Radiation therapy; Special Purpose and Application-Based Systems; Tumors; Automatic segmentation; Ensemble Deep Convolutional Neural Network; Lung computed tomography; Lung cancer; Modified mayfly optimization
• ISSN: 1380-7501; 1573-7721
• DOI: 10.1007/s11042-023-15688-0

The development of a computer-aided detection system is a critical component of clinical decision-making As the death rate grows, cancer has become a major concern for both men and women. The radiologists need to accurately pinpoint the region of the lung tumor to offer proper radiation therapy for lung cancer patients. Due to low-image quality, higher computational difficulties, and other reasons, the existing lung cancer segmentation methods failed to provide better segmentation accuracy. To overcome these challenges, we proposed a novel approach for lung tumor segmentation. Initially, the input CT scan image contrast level is increased using histogram equalization (HE) during pre-processing. The adaptive bilateral filter (ABF) provides enhanced CT scan images for de-noising. Next to pre-processing, we introduced an ensemble deep convolutional neural network (EDNN) based on Modified mayfly optimization and modified particle swarm optimization (M 2 PSO) algorithm for the segmentation of lung cancer from the pre-processed CT images. The proposed model accurately segments the lung disease tumor without manual supervision and the need for fully annotated data. Finally, the measures like dice similarity score (DSS), precision, sensitivity, dice loss, and generalized dice loss analyze the performance of the proposed model. Based on the experimental investigations, the proposed EDCNN- M 2 PSO algorithm demonstrated superior performance in terms of lung tumor segmentation than other existing techniques. The proposed model has average accuracy, sensitivity, and precision scores of 97%, 98%, and 98%, respectively. The proposed model's DSS value is 98.6%, which is relatively higher than the existing approaches.