Related Experiment Video
Updated: Jul 16, 2025
09:05
Application of Automated Image-guided Patch Clamp for the Study of Neurons in Brain Slices
Published on: July 31, 2017
11.6K
Demonstration of Convolutional Neural Networks to Determine Patch Test Reactivity
Adarsh Ravishankar1,2, Nicholas Heller3, Paul L Bigliardi2
1From the Department of Medicine, University of Minnesota, Minneapolis, Minnesota, USA.
Dermatitis : Contact, Atopic, Occupational, Drug
|September 12, 2023
Summary
Convolutional neural networks (CNNs) can accurately detect reactions in patch tests, improving diagnostic consistency for dermatologists. This AI model shows high performance, aiding in the interpretation of delayed-type hypersensitivity reactions.
Area of Science:
- Dermatology
- Artificial Intelligence
- Medical Imaging
Background:
- Patch testing is crucial for diagnosing allergic contact dermatitis.
- Interpreting patch test results can be subjective, leading to variability.
- Convolutional neural networks (CNNs) offer potential for objective image analysis.
Purpose of the Study:
- To evaluate a CNN's ability to differentiate between positive and negative patch test reactions.
- To establish a proof-of-concept for AI-assisted patch test interpretation.
- To assess the performance metrics of the CNN model.
Main Methods:
- Retrospective analysis of 13,622 patch test images (March 2020-March 2021).
- Training a CNN as a binary classifier for reaction/non-reaction.
- Performance evaluation using summary statistics and ROC curves.
Main Results:
- The CNN achieved an Area Under the Curve (AUC) of 0.940.
- Overall accuracy was 90.1%, with 86.0% sensitivity and 90.2% specificity.
- The model demonstrated high discriminative performance.
Conclusions:
- CNNs can effectively identify delayed-type hypersensitivity reactions in patch tests.
- AI holds promise for reducing subjectivity in patch test interpretation.
- Further prospective studies are needed to confirm generalizability.
