New XtraLight-MedMamba model classifies precancerous polyps with high accuracy

Researchers have developed XtraLight-MedMamba, a novel deep learning framework designed for classifying neoplastic tubular adenomas from whole-slide images. This architecture combines a ConvNeXt feature extractor with vision Mamba blocks to efficiently analyze local textures within a global context. The model incorporates a Spatial and Channel Attention Bridge module for enhanced multiscale feature extraction and a Fixed Non-Negative Orthogonal Classifier for parameter reduction and improved generalization. XtraLight-MedMamba achieved high accuracy and F1-scores with a significantly lower parameter count compared to existing transformer and Mamba models, making it suitable for resource-constrained environments. AI

IMPACT This lightweight model could enable more accessible and accurate polyp classification in colonoscopies, potentially improving early detection of colorectal cancer.

RANK_REASON The cluster describes a new deep learning model presented in an arXiv paper for a specific classification task. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• alphaXiv
• Aqsa Sultana
• arXiv
• CatalyzeX
• ConvNeXt
• DagsHub
• Fixed Non-Negative Orthogonal Classifier
• Hugging Face
• Spatial and Channel Attention Bridge
• XtraLight-MedMamba

• paper
• model release
• product

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