Integrative Image Processing Framework for Enhanced Detection of Leprosy-Associated Chronic Wounds.

Area of Science:

• Medical Imaging
• Computer Vision
• Dermatology

Background:

• Automated chronic wound analysis from smartphone images faces challenges like poor illumination and indistinct lesion margins.
• Accurate segmentation and feature extraction are crucial for reliable diagnostic support.

Purpose of the Study:

• To introduce the CO-WinF framework for robust automated leprosy chronic wound analysis.
• To enhance image quality, segmentation accuracy, and feature fusion for improved diagnostic capabilities.

Main Methods:

• The CO-WinF framework integrates three modules: AINCE for contrast remapping and texture restoration, MEPS for multi-resolution encoding and edge-aware segmentation using gradient-driven attention, and HSTFC for attention-weighted fusion of deep and handcrafted features.
• A gradient-boosted ensemble classifier optimized for class imbalance is used for final classification.

Main Results:

• CO-WinF achieved high performance on the CO2Wounds-V2 dataset, with 94.98% precision, 95.78% recall, 93.10% F1-score, and 87.09% IoU.
• The framework demonstrated superior results compared to existing state-of-the-art methods.

Conclusions:

• CO-WinF provides a computationally efficient and robust pipeline for smartphone-based wound analysis.
• The framework offers interpretable diagnostic support, establishing a new benchmark for resource-constrained clinical environments.