Related Experiment Video
Updated: Jun 30, 2026
09:05
Application of Automated Image-guided Patch Clamp for the Study of Neurons in Brain Slices
Published on: July 31, 2017
A Structured Review of Deep Learning Approaches and Image-Preprocessing Techniques for Automated Contact Allergy
Dominyka Stragyte1, Gvidas Mikalauskas2, Katrina Gaidulevic2
1Faculty of Medicine, Medical Academy, Lithuanian University of Health Sciences, LT-44307 Kaunas, Lithuania.
Medical Sciences (Basel, Switzerland)
|June 25, 2026
Summary
Deep learning shows promise for automated patch testing (PT) interpretation in allergic contact dermatitis (ACD). However, further standardized studies are needed for reliable clinical implementation and grading of reactions.
Area of Science:
- Dermatology
- Artificial Intelligence
- Medical Imaging
Background:
- Allergic contact dermatitis (ACD) is a prevalent inflammatory skin condition.
- Patch testing (PT) is the diagnostic gold standard but is labor-intensive and subjective.
- Advancements in AI and digital imaging offer potential for automated PT evaluation.
Purpose of the Study:
- To review the application of deep learning networks (DNNs) for automated patch testing classification.
- To summarize image-preprocessing techniques used in conjunction with DNNs for PT analysis.
Main Methods:
- Literature review of original research from 2020-2025 on deep learning for PT image analysis.
- Assessment of studies based on model architecture, datasets, preprocessing, and performance.
- Included studies utilized various DNN architectures like YOLOv5x, EfficientNetB0, Xception, and custom CNNs.
Main Results:
- Six studies met inclusion criteria, employing diverse deep learning models.
- Reported diagnostic performance showed high accuracy (90%-99.5%), with F1-scores (0.37-0.98) and AUROC (up to 0.94).
- Current models struggle with reliable International Contact Dermatitis Research Group (ICDRG) grading, particularly for severe reactions.
Conclusions:
- Deep learning demonstrates potential for automating PT interpretation.
- Standardized, multicenter studies with detailed protocols are essential for clinical integration.
- Further research is needed to improve ICDRG grading accuracy and ensure model generalizability.
