Risk stratification for early-stage NSCLC progression: a federated learning framework with large-small model synergy.

The improvements were highly significant (p < 0.001) across nearly all head-to-head comparisons in Centers B and D.

also in 132,142 other papers

Although the direct differences in AUC did not reach statistical significance in DeLong tests with some federated learning methods at certain centers (e.g., Centers A and C), FedCPI still showed clear improvement trends or significant gains in key evaluation metrics (particularly NRI and IDI), demonstrating the effectiveness of its methodological design.

also in 111,027 other papers