Class-wise dataset mixing boosts breast MRI tumor classification accuracy

Researchers have developed a method called Class-Wise Dataset Mixing to improve the generalization of deep learning models for breast MRI tumor classification. When models like EfficientNet-B3 and WaveViT-Small were trained on data from the Duke Breast Cancer MRI and fastMRI datasets, they performed poorly on an independent multi-center cohort (MAMA-MIA) due to dataset-origin bias. By mixing samples from both datasets within each class during training, the accuracy and F1 scores on the MAMA-MIA cohort significantly improved, demonstrating the effectiveness of this technique in overcoming domain shift. AI

IMPACT This research could lead to more reliable AI diagnostic tools for medical imaging by improving cross-dataset generalization.

RANK_REASON The cluster contains a research paper detailing a new methodology for improving AI model generalization. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• arXiv
• Duke Breast Cancer MRI
• EfficientNet-B3
• fastMRI
• Hugging Face
• MAMA-MIA
• WaveViT-Small

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →