Related Experiment Video
Updated: Jun 2, 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
A model combining deep learning and ensemble learning for melanoma recognition via dermoscopy.
1Dermatology, Heilongjiang Zhongyiyaodaxue Fushu Dier Yiyuan, Harbin, Heilongjiang, China.
Frontiers in Oncology
|June 1, 2026
Summary
This study developed a deep learning-ensemble model to improve melanoma diagnosis from skin lesion images. The model achieved high accuracy, offering potential for clinical use in differentiating malignant melanoma.
Area of Science:
- Dermatology
- Artificial Intelligence
- Medical Imaging
Background:
- Accurate melanoma identification is crucial for patient prognosis.
- Traditional dermatoscopic diagnosis is subjective and inconsistent.
- Developing objective diagnostic tools is essential.
Purpose of the Study:
- To create a deep learning-ensemble model for enhanced dermatoscopic differential diagnosis.
- To improve the accuracy of distinguishing malignant melanoma from other malignant skin lesions.
Main Methods:
- Utilized ISIC-2024 and HAM10000 dermatoscopic datasets.
- Employed nine convolutional neural network models for initial classification.
- Integrated an XGBoost ensemble model using raw scores as feature vectors.
Main Results:
- All nine models significantly differentiated between lesion types (p<0.001).
- Individual model AUCs ranged from 0.921-0.967.
- The XGBoost ensemble model achieved an AUC of 0.988 on the test dataset.
Conclusions:
- Deep learning and ensemble strategies enhance diagnostic accuracy.
- The model offers technical support for clinical diagnosis.
- Potential for clinical translation exists for this diagnostic approach.
