Synergistic Deep Learning Fusion for Precision Lung Cancer Staging

P, Sinthia; Albert, Anitha Juliette; M, Malathi; G, Gurumoorthy; S, Rajalakshmi

doi:10.31557/APJCP.2026.27.6.2335

Synergistic Deep Learning Fusion for Precision Lung Cancer Staging

Document Type : Research Articles

Authors

¹ Department of Biomedical Engineering, Saveetha Engineering College, Chennai, Tamil Nadu, India.

² Department of ECE, Loyola-ICAM College of Engineering and Technology, Chennai, India.

³ Department of ECE, Rajalakshmi Institute of Technology, Chennai, India.

⁴ Department of Medical Electronics, Saveetha Engineering College, Chennai, India.

⁵ Department of CSE, Sri Venkateswara College of Engineering,Chennai, India.

10.31557/APJCP.2026.27.6.2335

Abstract

Objective: To develop and evaluate an automated deep learning–based lung cancer staging system using computed tomography (CT) scan images. Methods: CT scan images were obtained from publicly available datasets (LIDC-IDRI/TCIA) comprising 1,018 patient scans. The dataset consisted of three subsets, which were: training (70 percent of total), validation (15 percent), and testing (15 percent). Lung region segmentation, anisotropic filtering, and data augmentation were used as preprocessing. To classify lung cancer stages, a customized CNN network based on multi-scale feature extraction and softmax-enabled probabilistic output was trained. Statistical confidence intervals, F1-score, ROC-AUC, recall, accuracy, and precision were used to test the performance of the model. Results: Using an area under the curve (AUC) of 0.98 (Stage I), 0.96 (Stage II), 0.95 (Stage III) and 0.97 (Stage IV) the proposed model indicates a total classification of 93.0 (95% CI: 91.2-94.8). Statistical analysis revealed a significant improvement compared to baseline CNN models (p < 0.05). Compared with state-of-the-art techniques, quantitative comparisons showed either equivalent performance or slightly higher performance, particularly in separating between early-stage (I-II) and advanced-stage (III-IV) disease. Conclusion: The findings demonstrate that the suggested CNN-based architecture can effectively and precisely classify the stage of lung cancer based on CT images, which assists in automated clinical decision-making and enhances the early detection process.

Keywords