Multiscale High-Level Feature Fusion for Histopathological Image Classification

Area of Science:

• Digital Pathology
• Computational Biology
• Medical Imaging Analysis

Background:

• Histopathological image classification is crucial for accurate disease diagnosis.
• Current methods may have limitations in extracting comprehensive image features.

Purpose of the Study:

• To propose an advanced method for multiclass histopathological image classification using deep convolutional neural networks.
• To enhance feature representation and classification performance for improved diagnostic accuracy.

Main Methods:

• Developed a deep convolutional neural network (coding network) for high-level feature extraction.
• Fused features from multiple convolutional layers to create multiscale high-level features.
• Utilized sparse autoencoder (SAE) and principal components analysis (PCA) for feature dimensionality reduction.

Main Results:

• The proposed method achieved superior performance in histopathological image classification.
• The enhanced feature representation significantly improved classification accuracy.
• Dimensionality reduction techniques maintained high efficiency without compromising performance.

Conclusions:

• The proposed deep learning approach effectively classifies multiclass histopathological images.
• The method offers improved feature representation and classification accuracy over standard coding networks.
• This technique holds promise for advancing disease diagnosis through automated image analysis.