Federated Computer Vision for Cervical Cytology Under Uneven Laboratory Annotation Budgets: An Incentive-Compatible Aggregation Evaluation of Healthcare Screening Equity

Federated learning promises to train cervical cytology classifiers across laboratories without centralising images, but the aggregation rule that combines local models determines whether laboratories with small annotation budgets contribute useful signal or are systematically outvoted. the present analysis formalises federated aggregation as a mechanism-design problem in which each laboratory's annotation budget is private information and the central server must choose weights that are both incentive-compatible and equity-preserving. We derive a budget-proportional aggregation rule and prove that it satisfies dominant-strategy incentive compatibility, individual rationality, and a screening-equity bound that limits the disparity in per-laboratory sensitivity. The operating characteristics of the rule are characterised under a declared variance structure with intraclass correlation 0.64, between-cluster standard deviation 6.0 and residual standard deviation 4.5; power was estimated across 2000 replications per design cell for cluster counts of 8, 16, 24 and 32 and effect sizes of 3, 5 and 8 percentage points.54 clusters are required to detect a 5-percentage-point effect at 80% power, and that conventional equal-weight aggregation inflates the minimum detectable effect for low-budget laboratories by a factor that the proposed rule removes. The contribution is a formal aggregation criterion that makes annotation-budget heterogeneity explicit in federated cytology screening and a design-power characterisation that laboratories can use to plan participation.