Related Experiment Video
Updated: Jul 9, 2026
07:53
Three-Dimensional Reconstruction for the Whole Lung with Early Multiple Pulmonary Nodules
Published on: October 13, 2023
A novel ensemble transfer learning approach for lung cancer classification using advance VGGNet16 with wavelet
Manmath Nath Das1, Niranjan Panda1, Rasmita Rautray1
1Dept. of Computer Science and Engineering, ITER(FET), Siksha 'O' Anusandha (Deemed to be) University, Bhubaneswar, India.
Computers in Biology and Medicine
|December 10, 2025
Summary
This study introduces an advanced deep learning model for early lung cancer diagnosis using CT scans. The AI achieves near-perfect accuracy, offering a robust tool for real-time clinical applications.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Lung cancer presents a significant global health challenge, necessitating efficient and accurate diagnostic methods.
- Current diagnostic strategies require enhancement for improved early detection rates and reduced patient burden.
Purpose of the Study:
- To develop and validate an enhanced deep learning framework for early lung cancer diagnosis using computed tomography (CT) scans.
- To improve the sensitivity and accuracy of lung cancer detection, particularly for benign cases, through advanced AI techniques.
Main Methods:
- A VGG-16 deep learning model was fine-tuned using Comprehensive Learning Particle Swarm Optimization (CL-PSO).
- Wavelet Transform Equalization and class-weighted training were employed for preprocessing and to address data imbalance.
- Grad-CAM visualizations were used to interpret model predictions and ensure alignment with radiological standards.
Main Results:
- The model achieved exceptional performance on the IQ-OTH/NCCD dataset, with 99.99% accuracy, 99.98% precision and recall, 99.99% F1-score, and 1.00 AUC-ROC.
- The AI demonstrated robustness against noise, occlusion, and illumination variations, processing images in under 50 ms.
- Interpretability was confirmed through Grad-CAM, correlating predictions with radiological decision points.
Conclusions:
- The proposed deep learning approach offers a highly accurate and efficient method for early lung cancer diagnosis.
- The model's real-time processing capabilities and robustness make it suitable for integration into hospital Picture Archiving and Communication Systems (PACS) and edge healthcare systems.