# Benchmarking Medical AI

Factored contributed to MedPerf, enabling federated medical AI across 50+ institutions worldwide.

Written by:

Data Science Center of Excellence

1
min read

## Key Takeaways:

Streamlined data preprocessing pipelines.

Real-time dashboards for experiment monitoring.

Accelerated evaluation timelines from months to hours.

As key contributors to **MedPerf**, a groundbreaking platform for federated learning and evaluation in healthcare, Factored played a pivotal role in shaping the future of medical AI.

### Accelerating Federated Learning Through Advanced Data Pipelines

Factored’s expertise was central to the development of **data preparation pipelines** that streamlined preprocessing and annotation across **30+ hospitals worldwide**. These pipelines tackled one of the most significant bottlenecks in medical AI: preparing diverse and decentralized datasets. By ensuring consistency, adaptability, and reproducibility, Factored enabled seamless collaboration between institutions.

Factored Data Architecture

### Enabling Real-Time Insights with Custom Dashboards

Our team also integrated user-friendly **data inspection tools** and **real-time dashboards**, empowering researchers to monitor and troubleshoot federated AI experiments efficiently. These innovations accelerated data preparation workflows, reduced errors, and enhanced decision-making during critical stages of the **MedPerf project**.

### Shaping the Future of Medical AI

MedPerf, powered by **Factored’s** **Alejandro Aristizábal**, a co-author of the report, is transforming how AI models are developed and evaluated in healthcare, setting new standards for privacy-preserving, scalable, and effective AI benchmarking. **Our collaboration on the Federated Tumor Segmentation (FeTS) studies involved over 50 institutions across six continents**, where MedPerf enabled benchmarking of **41 AI models in record time**, reducing evaluation timelines from months to mere hours.

To see the full paper click [here](https://www.nature.com/articles/s42256-023-00652-2/figures/1)