Related Experiment Video
Updated: Jul 27, 2025
07:15
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
6.9K
An Adaptive Early Stopping Technique for DenseNet169-Based Knee Osteoarthritis Detection Model.
Bander Ali Saleh Al-Rimy1, Faisal Saeed2, Mohammed Al-Sarem3
1Department of Computer Science, Faculty of Computing, Universiti Teknologi Malaysia, Johor Bahru 81310, Malaysia.
Diagnostics (Basel, Switzerland)
|June 10, 2023
Summary
This study introduces an adaptive early stopping technique with gradual cross-entropy loss for improved knee osteoarthritis detection using DenseNet169 on X-ray images, enhancing diagnostic accuracy.
Area of Science:
- Health Informatics
- Medical Imaging Analysis
- Artificial Intelligence in Healthcare
Background:
- Knee osteoarthritis (OA) diagnosis is crucial for managing this prevalent condition.
- Accurate detection of knee OA from X-ray images remains a significant challenge in health informatics.
- Deep learning models show promise but require optimization to prevent overfitting and maximize diagnostic performance.
Purpose of the Study:
- To evaluate the efficacy of the DenseNet169 deep learning architecture for knee OA detection using X-ray images.
- To propose and integrate an adaptive early stopping technique with gradual cross-entropy (GCE) loss estimation for optimizing the training process.
- To enhance the accuracy and reliability of knee OA detection models.
Main Methods:
- Utilized the DenseNet169 convolutional neural network architecture for analyzing knee X-ray images.
- Developed an adaptive early stopping mechanism based on validation accuracy thresholds to prevent model overfitting.
- Integrated a gradual cross-entropy (GCE) loss estimation technique into the training epochs.
- Combined adaptive early stopping and GCE within the DenseNet169 model for knee OA detection.
Main Results:
- The proposed model, incorporating adaptive early stopping and GCE, demonstrated superior performance compared to existing methods.
- Achieved higher accuracy, precision, and recall in detecting knee osteoarthritis.
- The adaptive early stopping and GCE techniques effectively optimized the DenseNet169 model's training, reducing overfitting and improving generalization.
Conclusions:
- The integration of adaptive early stopping with gradual cross-entropy loss significantly enhances the performance of DenseNet169 for knee OA detection.
- This approach offers a robust method for accurate and efficient diagnosis of knee osteoarthritis from X-ray images.
- The findings suggest a promising direction for improving computer-aided diagnosis systems in musculoskeletal imaging.
