Related Experiment Video
Updated: Oct 14, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.0K
Artificial intelligence classification model for macular degeneration images: a robust optimization framework for
Wen-Hsien Ho1,2, Tian-Hsiang Huang3, Po-Yuan Yang4
1Department of Healthcare Administration and Medical Informatics, Kaohsiung Medical University, No. 100, Shin-Chuan 1st Road, Kaohsiung, 807, Taiwan.
BMC Bioinformatics
|November 9, 2021
Summary
This study optimized artificial intelligence (AI) models for interpreting macular degeneration images using uniform design. The resulting ResNet model achieved superior accuracy and a lower false negative rate, aiding in diagnosis.
Area of Science:
- Ophthalmology
- Artificial Intelligence
- Medical Imaging
Background:
- Chronic diseases are increasing in aging populations.
- AI-assisted interpretation of macular degeneration images requires research.
- Hyperparameter optimization is crucial for ResNet models.
Purpose of the Study:
- To optimize hyperparameters of a ResNet model for macular degeneration image classification.
- To improve the robustness and accuracy of AI models in medical diagnostics.
Main Methods:
- Utilized uniform design for systematic hyperparameter optimization of a ResNet model.
- Trained and validated the ResNet model on an open dataset of macular degeneration images.
- Evaluated model performance using accuracy, false negative rate, and signal-to-noise ratio.
Main Results:
- The optimized ResNet model achieved high optimal accuracy (0.9907) and mean accuracy (0.9848).
- The model demonstrated a low mean false negative rate (0.015).
- Performance surpassed previously reported results on the same dataset.
Conclusions:
- Uniform design effectively optimizes ResNet hyperparameters for enhanced stability and performance.
- The optimized ResNet model offers a robust tool for assisting in macular degeneration diagnosis.
- This systematic approach reduces design time and improves diagnostic confidence.