Related Experiment Video
Updated: Jan 15, 2026
09:37
Combining Reflectance Confocal Microscopy with Optical Coherence Tomography for Noninvasive Diagnosis of Skin Cancers via Image Acquisition
Published on: August 18, 2022
2.8K
Automated Detection of Benign and Malignant Skin Lesions from Reflectance Confocal Microscopy Images Using Deep
Jesutofunmi A Omiye1,2, Babar K Rao3,4, Shazli Razi5
1Department of Dermatology, Stanford University, Stanford, California, USA.
JID Innovations : Skin Science From Molecules to Population Health
|October 13, 2025
Summary
Artificial intelligence (AI) models can now interpret reflectance confocal microscopy (RCM) images of skin lesions, automating diagnosis. This AI approach shows high accuracy, potentially reducing the need for invasive skin biopsies.
Area of Science:
- Dermatology
- Medical Imaging
- Artificial Intelligence
Background:
- Reflectance confocal microscopy (RCM) is a noninvasive tool for skin lesion diagnosis.
- Its underutilization stems from the need for specialized interpretation skills.
- Artificial intelligence (AI) offers a solution for automating RCM image analysis.
Purpose of the Study:
- To develop and evaluate deep learning models for automating RCM image analysis.
- To improve the diagnostic accuracy of RCM for skin lesions.
- To assess AI's potential in distinguishing benign from malignant skin lesions.
Main Methods:
- Developed two deep learning models: ResNet-18 for skin layer detection and ResNet-34 with a gated recurrent unit for lesion classification.
- Models were pre-trained on 3rd generation RCM images and fine-tuned on 4th generation data.
- Utilized a cohort of 845 patients with 1147 lesions across 4391 RCM images.
Main Results:
- The layer detection model achieved areas under the curve (AUC) of 0.70 (dermis), 0.71 (epidermis), and 0.57 (dermoepidermal junction).
- The lesion classification model distinguished malignant from benign lesions with an AUC of 0.80 and 0.91 specificity.
- The AI models demonstrated diagnostic accuracy comparable to expert dermatologists.
Conclusions:
- AI-powered analysis of RCM images can effectively differentiate benign from malignant skin lesions.
- This technology has the potential to enhance RCM interpretation, reduce unnecessary biopsies, and guide future research.
- AI integration can significantly improve point-of-care skin lesion diagnosis.
