Related Experiment Video
Updated: Sep 16, 2025
09:37
Combining Reflectance Confocal Microscopy with Optical Coherence Tomography for Noninvasive Diagnosis of Skin Cancers via Image Acquisition
Published on: August 18, 2022
2.5K
Explainable deep learning approaches for high precision early melanoma detection using dermoscopic images.
Md Abdullah All Mahmud1, Sadia Afrin2, M F Mridha3
1Department of Computer Science, American International University-Bangladesh, Dhaka, Bangladesh.
Scientific Reports
|July 8, 2025
Summary
This study developed an advanced deep learning model for early skin melanoma detection from dermoscopic images. The system achieved high accuracy, improving diagnostic reliability and aiding medical experts in early cancer identification.
Area of Science:
- Dermatology
- Artificial Intelligence
- Medical Imaging
Background:
- Early skin melanoma detection is challenging due to image variability.
- Existing automated diagnostic systems lack reliability and explainability across diverse conditions.
Purpose of the Study:
- To develop a robust Automated Diagnostic System for early skin cancer detection using dermoscopic images.
- To enhance the reliability and explainability of AI-driven dermatological diagnostics.
Main Methods:
- A novel deep learning model incorporating Global Average Pooling, Batch Normalization, Dropout, and dense layers with ReLU and Swish activations was proposed.
- Explainable AI techniques, including Gradient-weighted Class Activation Mapping and Saliency Maps, were employed for model interpretability.
Main Results:
- The proposed model achieved high accuracies of 95.23% and 96.48% on two distinct datasets.
- The system demonstrated robust performance and reliability across various metrics, validated by explainable AI insights.
Conclusions:
- The developed model significantly enhances early skin cancer diagnostics, offering a reliable tool for medical experts.
- This research advances the acceptance and application of deep learning in healthcare for improved clinical outcomes.
