Medical image classification using synergic deep learning.

Area of Science:

• Computer-aided diagnosis
• Medical image analysis
• Deep learning

Background:

• Medical image classification is crucial for diagnosis, retrieval, and mining.
• Deep learning excels over traditional methods but faces challenges due to intra-class variation and inter-class similarity in medical images.
• Diversity in imaging modalities and clinical pathologies complicates accurate classification.

Purpose of the Study:

• To propose a synergic deep learning (SDL) model to improve medical image classification.
• To address the limitations of existing deep learning models in handling complex medical image variations.
• To enable mutual learning between multiple deep convolutional neural networks (DCNNs).

Main Methods:

• Developed a synergic deep learning (SDL) model utilizing multiple DCNNs simultaneously.
• Implemented a synergic network where concatenated representations from pairs of DCNNs predict class similarity.
• Enabled end-to-end training supervised by both DCNN classification errors and synergic errors from DCNN pairs.

Main Results:

• The SDL model demonstrated state-of-the-art performance on multiple medical image classification benchmarks.
• Experimental validation was conducted on ImageCLEF-2015, ImageCLEF-2016, ISIC-2016, and ISIC-2017 datasets.
• The synergic error mechanism effectively updated the model, improving classification accuracy.

Conclusions:

• The proposed SDL model effectively addresses challenges in medical image classification.
• Mutual learning between DCNNs significantly enhances classification performance.
• The SDL approach represents a significant advancement in computer-aided diagnosis and medical image analysis.