[arXiv]score: 0.46
ARMA-C3 Graph Contrastive Framework Classifies Neuroimaging with Limited Labels
May 26, 2026
ARMA-C3 combines ARMA graph convolutions with contrastive learning and graph-cut regularization for semi-supervised node classification, validated on ADNI, NIFD, and three medical imaging benchmarks.
cs.CV
HOW THIS AFFECTS YOU
●
researcherGraph-cut regularization combined with contrastive learning offers a structured approach to label-scarce classification on inter-sample relationship graphs.
●
healthValidated on Alzheimer's and frontotemporal dementia datasets, this framework could reduce labeled data requirements for neuroimaging classification tasks.