Advancing Health Care Informatics Through You Only Look Once v10 for Lung Nodule and Cancer Detection in Medical Imaging Systems
Document Type
Conference Proceeding
Source of Publication
2026 International Conference on Cybersecurity Digital Forensics and AI Applications Iccsdfai 2026
Publication Date
6-1-2026
Abstract
Early detection of lung nodules plays a pivotal role in reducing lung cancer mortality; however, recent advancements in real-time object detection models have struggled to balance detection accuracy with computational efficiency in high-resolution CT scans. While architectures such as YOLOv8 and EfficientDet have demonstrated promising performance, their limitations in capturing fine-grained contextual dependencies and handling class imbalance remain significant challenges. This study proposes a novel hybrid attention-based YOLOv10 framework, termed Optimized-YOLO, designed to enhance both localization and classification of lung nodules in complex thoracic CT images. The model integrates spatial-channel hybrid attention modules to improve feature representation and employs the Sharpness-Aware Minimization (SAM) optimizer to enhance generalization and robustness against noisy annotations. Training is conducted on the LIDC-IDRI dataset using bounding-box annotations with advanced data augmentation strategies. Experimental results demonstrate that the proposed model achieves a precision of 89.7%, recall of 90.8%, and a balanced sensitivity-specificity trade-off, outperforming Faster R-CNN, EfficientDet, and recent YOLO variants. Additionally, the model achieves real-time inference speed without compromising diagnostic reliability. The findings indicate that integrating hybrid attention with modern optimization techniques significantly improves detection performance.
DOI Link
ISBN
[9798319520487]
Publisher
IEEE
First Page
797
Last Page
802
Disciplines
Computer Sciences
Keywords
Class imbalance handling, Hybrid attention mechanism, Lung nodule detection, Real-time CT analysis, Sharpness-Aware Minimization (SAM), YOLOv10
Scopus ID
Recommended Citation
Tubishat, Mohammad; Abugabah, Ahed; and Al-Smadi, Ahmad Mohammad, "Advancing Health Care Informatics Through You Only Look Once v10 for Lung Nodule and Cancer Detection in Medical Imaging Systems" (2026). All Works. 8446.
https://zuscholars.zu.ac.ae/works/8446
Indexed in Scopus
yes
Open Access
no