Pigment network detection and classification in dermoscopic images using directional imaging algorithms and convolutional neural networks

2025-01-01

Biomedical Signal Processing and Control

AI Summary

This work proposes a two-stage method to automatically identify regular and irregular pigment networks (PN) in dermoscopic images for aiding early melanoma diagnosis. In the first stage, PN regions are precisely localized by integrating directional imaging, principal component analysis (PCA), and contrast enhancement. The second stage employs a lightweight convolutional neural network (CNN) to classify typical versus atypical PN patterns. Evaluated on a small dataset of 200 images, the approach achieves 90% accuracy, 90% sensitivity, and 89% specificity, with a 100% PN detection rate—significantly outperforming existing techniques under limited data conditions. By effectively combining classical image processing with deep learning, the method enhances both the automation and reliability of melanoma screening.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Data Mining & Knowledge Management: Anomaly/Outlier Detection

Application Category

• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Problem

• pigment network
• melanoma diagnosis
• dermoscopic images
• atypical vs typical classification

Innovation

• Directional Imaging Algorithm
• Convolutional Neural Network (CNN)
• Pigment Network Classification
• Dermoscopic Image Analysis
• PCA-based Preprocessing