Authors:
B. Rajalakshmi, B. Manjunatha, Thirumalraj Karthikeyan, S. Venkatasubramanian, Rahul Panakkal
Addresses:
Department of Computer Science and Engineering, New Horizon College of Engineering, Bengaluru, Karnataka, India. Department of Mechanical Engineering, New Horizon College of Engineering, Bengaluru, Karnataka, India. Department of Artificial Intelligence, Quest Technologies, Trichy Research Labs, Tiruchirappalli, Tamil Nadu, India. Department of Computer Science and Business Systems, Saranathan College of Engineering, Panjapur, Trichy, Tamil Nadu, India. Department of Computer Science and Engineering, University of Illinois Urbana‑Champaign, Champaign, Illinois, United States of America.
Abstract:
To enhance the outcomes of early diagnosis and treatment, research on the detection of bone cancer is necessary. The purpose of this research is to present a method for identifying bone cancer that uses sophisticated image preprocessing techniques, ResNet-based feature extraction, and hyperparameter tuning with the Black Eagle Optimiser (BEO). Utilising its deep architecture and residual learning methods, the ResNet model is utilised to extract discriminative features from medical imaging data. This is accomplished by capitalising on the model's capabilities. The retrieved features are then further refined and optimised using the BEO method, resulting in improved overall model performance and generalisation. In addition, it presents a novel Attention Mechanism-Squeeze-and-Excitation-Convolutional Neural Network (AM-SE-CNN) model for classification. This model blends attention mechanisms and channel-wise feature recalibration to enhance the model's discriminative power and interpretability. The proposed method offers improved performance in bone cancer identification when compared to cutting-edge approaches. This was accomplished through intensive experimentation on benchmark datasets. The proposed model achieved highly impressive performance metrics, including 99.62% accuracy, 99.47% precision, 99.35% recall, 99.21% specificity, and an F1 score of 99.17%, demonstrating its efficiency in identifying bone cancer.
Keywords: Bone Cancer; Residual Network; Black Eagle Optimiser; Squeeze and Excitation; Convolutional Neural Network; Diagnosis and Treatment; Image Preprocessing; Cancer Detection; Medical Imaging.
Received: 10/10/2024, Revised: 27/01/2025, Accepted: 09/04/2025, Published: 07/09/2025
DOI: 10.64091/ATIHL.2025.000172
AVE Trends in Intelligent Health Letters, 2025 Vol. 2 No. 3 , Pages: 133-149