Related Experiment Video
Updated: May 13, 2025
Author Spotlight: Advancing 3D Modeling for Enhanced Diagnosis and Treatment of Pulmonary Nodules in Early-Stage Lung Cancer
Published on: October 13, 2023
Explainable AI for lung cancer detection via a custom CNN on CT images
Mohamed Hammad1,2, Mohammed ElAffendi3, Ahmed A Abd El-Latif3,4
1EIAS Data Science Lab, College of Computer and Information Sciences, Center of Excellence in Quantum and Intelligent Computing, Prince Sultan University, Riyadh, 11586, Saudi Arabia. mhammad@psu.edu.sa.
This study introduces a custom AI model using convolutional neural networks (CNNs) and explainable AI (XAI) for accurate lung cancer subtype classification from CT scans. The AI achieved 93.06% accuracy, improving early detection and patient outcomes.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Lung cancer is a leading cause of global cancer deaths, with late-stage detection limiting treatment efficacy and survival rates.
- Traditional CT image analysis for lung cancer is subjective, time-consuming, and prone to errors, hindering early diagnosis.
- Accurate classification of lung cancer subtypes is crucial for effective treatment planning and improved patient prognosis.
Purpose of the Study:
- To develop and validate a custom convolutional neural network (CNN) model for reliable classification of lung cancer subtypes (squamous cell carcinoma, large cell carcinoma, adenocarcinoma) using CT images.
- To integrate explainable AI (XAI) techniques, specifically gradient-weighted class activation mapping (Grad-CAM), to enhance model transparency and clinical interpretability.
- To demonstrate the potential of an AI-driven approach to improve the accuracy and efficiency of early lung cancer detection.
Main Methods:
- Development of a custom convolutional neural network (CNN) architecture for lung cancer subtype classification.
- Application of gradient-weighted class activation mapping (Grad-CAM) for visualizing and interpreting model predictions.
- Training and validation of the model on a comprehensive dataset of computed tomography (CT) images.
- Evaluation of model performance using accuracy, precision, recall, and F1-scores.
Main Results:
- The custom CNN model achieved an overall accuracy of 93.06% in classifying lung cancer subtypes.
- The model demonstrated robust performance across all cancer types, with strong precision, recall, and F1-scores.
- Grad-CAM visualizations provided clinically relevant interpretability, aiding in understanding the model's diagnostic reasoning.
Conclusions:
- The proposed AI model, combining CNNs and XAI, offers a highly accurate and interpretable solution for lung cancer subtype classification from CT images.
- This approach has the potential to significantly improve early lung cancer detection rates and enhance patient survival by providing reliable and transparent diagnostic support.
- The integration of explainable AI features ensures clinical validation and facilitates the adoption of AI tools in routine oncological diagnostics.
