Related Experiment Video
Updated: Aug 13, 2026
08:05
Lung CT Segmentation to Identify Consolidations and Ground Glass Areas for Quantitative Assesment of SARS-CoV Pneumonia
Published on: December 19, 2020
CNN-RNN framework for lung cancer classification using CT imaging and GAN-based augmentation
Bodicherla Siva Sankar1, D Natarajasivan1, M Purushotham Reddy2
1Department of Computer Science and Engineering, Faculty of Engineering and Technology, Annamalai University, Chidambaram, Tamil Nadu, India.
Frontiers in Artificial Intelligence
|August 12, 2026
Summary
This study introduces a novel AI framework combining Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs) for accurate lung cancer classification from CT scans. The AI model achieved high accuracy, aiding early detection in challenging medical imaging scenarios.
Area of Science:
- Artificial Intelligence in Medical Imaging
- Deep Learning for Cancer Diagnosis
- Computational Pathology
Background:
- Lung cancer is a leading cause of cancer mortality globally, necessitating improved early detection methods.
- Accurate classification of lung cancer from CT imaging is hindered by limited data, class imbalance, and interpretability issues in current deep learning models.
- Existing deep learning systems often lack transparency, making it difficult to understand their diagnostic reasoning.
Purpose of the Study:
- To develop an explainable, Generative AI (GenAI)-driven framework for lung cancer classification using CT imaging.
- To address challenges of limited datasets and class imbalance through Generative Adversarial Network (GAN)-based augmentation.
- To enhance classification performance and interpretability for improved clinical decision support.
Main Methods:
- A hybrid CNN-RNN framework was proposed, integrating CNNs for spatial feature extraction and RNNs (specifically LSTMs) for sequential analysis.
- GAN-based augmentation was employed to enhance the representation of minority classes within the limited dataset.
- Attention-guided feature fusion, Grad-CAM, and SHAP visualization techniques were utilized for improved performance and interpretability.
Main Results:
- The proposed framework achieved high performance metrics on the IQ-OTH/NCCD dataset: 92.84% accuracy, 91.76% precision, 90.42% recall, and 91.08% F1-score.
- A ROC-AUC value of 0.94 was obtained, indicating strong discriminative ability.
- Visualization methods successfully highlighted critical lesion regions, enhancing the model's explainability.
Conclusions:
- The developed GenAI-driven CNN-RNN framework demonstrates significant potential for accurate and explainable lung cancer classification from CT images.
- The framework effectively addresses data limitations and class imbalance issues common in medical imaging datasets.
- This approach offers a promising tool to support early lung cancer detection, particularly under constrained data conditions.
