Transfer learning with fuzzy decision support for multi-class lung disease classification: performance analysis of
Nand Lal Yadav1, Sudhakar Singh2, Rajesh Kumar1
1Department of Electronics and Communication, University of Allahabad, Prayagraj, India.
Scientific Reports
|October 8, 2025
Summary
This study introduces a novel method for classifying lung diseases using transfer learning and fuzzy logic, achieving high accuracy in detecting conditions like COVID-19 and pneumonia from X-ray images.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Computer-Aided Diagnosis
Background:
- Accurate lung disease classification from medical images is crucial for timely diagnosis and treatment.
- Traditional computer-aided diagnosis systems face challenges in handling the complexity and uncertainty of lung pathologies.
Purpose of the Study:
- To develop and evaluate a novel approach for multi-class lung disease classification by integrating transfer learning with fuzzy decision support systems.
- To enhance the accuracy and interpretability of automated lung disease diagnosis using chest X-ray images.
Main Methods:
- Employed transfer learning with pre-trained Convolutional Neural Network (CNN) architectures (VGG16, VGG19, ResNet50) adapted for lung disease classification.
- Integrated a fuzzy logic decision layer to manage uncertainty in medical image classification.
- Utilized a k-symbol Lerch transcendent function for image enhancement, improving contrast and feature visibility by 23.4% and 18.7%, respectively.
- Evaluated the models on a dataset of 8,409 chest X-ray images across six disease classes.
Main Results:
- The ResNet50-based model with fuzzy integration achieved the highest accuracy (98.7%), sensitivity (98.4%), and specificity (98.8%).
- The proposed approach demonstrated significant improvement in classifying uncertain cases (8.4% enhancement) where traditional CNN confidence was low.
- Fuzzy inference provided transparent reasoning with an average of 8.4 rules per decision, enhancing clinical interpretability while maintaining real-time processing (0.23s per image).
Conclusions:
- The integration of transfer learning and fuzzy logic offers a robust and interpretable solution for automated multi-class lung disease classification.
- This methodology significantly improves diagnostic accuracy and uncertainty handling in computer-aided diagnosis systems for lung pathologies.
- The developed system shows promise for clinical application, aiding radiologists in faster and more reliable lung disease detection.