Federated Bayesian Surveillance of Mechanical Thrombectomy Adverse Events: A Population Risk Layer for Surgical Digital Twins
A federated Bayesian protocol privately estimates thrombectomy adverse-event rates from FDA MAUDE reports.
The paper proposes federated Bayesian surveillance of mechanical thrombectomy adverse events as a privacy-preserving belief layer for surgical digital twins. Each of eight manufacturer sites shares Rényi-differentially-private updates of Gamma-Poisson posteriors rather than raw reports. On 8,617 FDA MAUDE reports for thrombus-retrieval catheters, including 6,491 classified cases, the protocol scores -5.78 per test event versus -26.58 for differentially private FedAvg at epsilon about 2.09. Because MAUDE lacks procedure denominators, the outputs are relative rate orderings rather than absolute risks.
- Evaluated on 8,617 FDA MAUDE reports for product code NRY.
- 6,491 reports were classified into five complication classes across eight sites.
- At epsilon about 2.09, the Poisson score is -5.78 versus -26.58 for DP-FedAvg.
- Missing procedure denominators mean outputs are relative rates, not absolute risks.
From the original article
1 min read at arxiv.orgLearned surgical simulators and world models can roll out plausible procedural futures, but they carry no grounded estimate of how often interventional devices actually harm patients. We propose treating population-scale adverse-event surveillance as a distinct belief layer of the surgical digital twin, and we evaluate a federated Bayesian protocol for learning it under formal privacy guarantees.
By Damini Rijhwani · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.