American
International
University-
Bangladesh

Students & Supervisors

Abstract

Melanoma, a dangerous form of skin cancer, is rapidly spreading and poses a serious threat to global health. Early and accurate detection is essential for saving lives, yet conventional diagnostic approaches often fall short in precision. This research presents a robust detection system utilizing ResNet34, a deep learning model known for its efficiency in image classification tasks. The study emphasizes effective preprocessing and data augmentation to improve image quality and diversity. ResNet34 was employed for feature extraction and classification, achieving an im-pressive accuracy of 98%, with a precision of 99% for malignant and 97% for benign cases. Recall scores reached 97% and 99% for malignant and benign cases respectively, resulting in a balanced F1-score of 0.98. ROC analysis further confirmed the model’s reliability in distinguishing melanoma from benign lesions. The classification report supports its effectiveness and potential for real-world deployment. This work highlights the powerful role of deep learning in enhancing medical diagnostics and offers a practical solution for early melanoma detection, contributing to global skin cancer prevention efforts.

Keywords

Publication Details

• DOI: https://doi.org/10.1007/978-3-031-98161-6_8
• Type of Publication: Conference 
• Conference Name: International Conference on Computational Intelligence in Engineering Science (ICCIES 2025)
• Date of Conference: 23/07/2025 - 23/07/2025
• Venue: Ton Duc Thang University, Ho Chi Minh City, Vietnam. 19 Nguyen Huu Tho Street, Tan Hung Ward, Ho Chi Minh City, Vietnam.
• Organizer: Springer, Wroclaw University of Science and Technology