Melanoma Classification Using a Novel Deep Convolutional Neural Network with Dermoscopic Images.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Dermatology

Background:

• Melanoma detection from dermoscopic images is complex.
• Deep learning offers potential solutions for automated analysis.
• Existing methods may lack efficiency or simplicity.

Purpose of the Study:

• To develop a lightweight and efficient deep convolutional neural network (DCNN) for automated melanoma classification.
• To accurately distinguish between malignant and benign melanoma.
• To provide an advanced yet less complex framework for melanoma diagnosis.

Main Methods:

• A novel deep convolutional neural network (DCNN) architecture was designed.
• The DCNN incorporates multiple layers for feature extraction.
• The model was trained and validated using dermoscopic images from ISIC datasets (2016, 2017, 2020).

Main Results:

• The proposed DCNN achieved high classification accuracies: 81.41% (ISIC 2016), 88.23% (ISIC 2017), and 90.42% (ISIC 2020).
• Performance metrics included accuracy, precision, recall, specificity, and F1-score.
• The DCNN demonstrated superior performance compared to state-of-the-art networks.

Conclusions:

• The developed DCNN offers an efficient and less complex approach for automated melanoma detection.
• This automated system can expedite the diagnostic process.
• The findings support the potential of AI in improving melanoma identification and patient outcomes.