A contrastive learning framework with adaptive feature fusion for brain tumor classification

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Biology

Background:

• Accurate brain tumor classification from MRI is crucial but challenging due to tumor variability.
• Existing deep learning models struggle to capture subtle pathological features for reliable classification.

Purpose of the Study:

• To develop a novel contrastive learning framework with adaptive feature fusion (AFF-CL) for improved brain tumor classification.
• To enhance the discriminative power of deep learning models for fine-grained pathological feature recognition.

Main Methods:

• Proposed a contrastive learning framework (AFF-CL) utilizing a dynamic label queue for multiple positive pair construction.
• Incorporated a local input stream and an adaptive feature fusion (AFF) module to integrate multi-scale contextual information.
• Trained and evaluated the framework on a public figshare dataset for brain tumor classification.

Main Results:

• The AFF-CL framework achieved state-of-the-art performance on brain tumor classification.
• Demonstrated significant improvements over existing methods in accuracy and discriminative power.
• The adaptive feature fusion effectively integrated localized and global representations for enhanced classification.

Conclusions:

• The proposed AFF-CL framework offers a promising approach for precise computer-aided diagnosis of brain tumors.
• The method's ability to handle intra-class heterogeneity and inter-class differences enhances diagnostic accuracy.
• This technique has strong potential for clinical application in neuro-oncology and treatment planning.