Imagine a future in which healthcare is of the best quality, inexpensive and available to everyone. That's the power of open-source medical AI. That's what we are building.
The story behind how EasyMedAi started
The COVID-19 pandemic thrust healthcare into uncharted territory. Hospitals were overwhelmed, and doctors faced a novel disease they barely understood.
Artificial intelligence seemed poised to help. Research teams worldwide raced to develop predictive tools, hoping to arm frontline clinicians with insights from the data flowing out of early hotspots, eventually rolling out more than 400 AI models. Some were used in hospitals, despite not being properly tested. But despite the best of intentions, the reality fell short. None of these tools were fit for clinical use, as studies from reputable journals later revealed 1, 2. Flawed data, biased algorithms, lack of validation—the pandemic exposed the immaturity of medical AI.
Iacopo Ciampa, the CEO of EasyMedAi, saw both a cautionary tale and a call to action. He believed this failure stemmed from a fundamental flaw in the way medical AI is developed: a lack of collaboration and transparency. Too many researchers worked in silos, on inappropriate datasets, without input from medical experts. The result was a multitude of models that were biased, fragile, and ultimately unsuitable for real-world deployment.
We founded EasyMedAi to chart a different path—one based on an open source community (or open science, more broadly). This means:
Open data: Researchers make their datasets publicly available on the platform, allowing others to scrutinize the data for bias or errors, and to use the data for their own research. This is crucial, as the quality and representativeness of training data directly impacts the performance and fairness of the resulting models.
Open-source code: AI models and algorithms are shared openly on the platform. This allows other researchers to examine the code, suggest improvements, and adapt the models for their own use. It also enables reproducibility, as others can run the same code on the same data to verify results.
Open-access publishing: Research papers are made freely available on the platform, rather than locked behind paywalls. This makes it easier for researchers, particularly those in resource-limited settings, to access the latest findings and build upon them.
Open collaboration: Researchers actively engage with the wider scientific community, including clinicians and patient advocates, throughout the research process. This helps ensure that the AI tools being developed are not only technically sound but also clinically relevant and acceptable to patients.
These approaches not only deliver the best results for patients and for scientific progress 3, but also favor economic development. For example, an EU study found that an investment of €1 billion in open-source software brought a positive impact of between €65 and €95 billion on the European economy alone 4.
EasyMedAi's journey began at the height of the pandemic, when our team came together around the shared vision indicated at the top of this page. Through all these efforts, our mission remains constant: to facilitate the development of medical AI that is trustworthy, impactful, and accessible, at scale.
As the pandemic wanes, the hard lessons it taught about developing medical AI remain. In the same way, the challenges facing healthcare as a whole persist. We believe that only by working in a community can we realize the full potential of AI in medicine. Come build it with us.
References
1. https://www.nature.com/articles/s42256-021-00307-0
2. https://www.bmj.com/content/369/bmj.m1328
3. https://www.ncbi.nlm.nih.gov/books/NBK525412/
4. https://op.europa.eu/en/publication-detail/-/publication/29effe73-2c2c-11ec-bd8e-01aa75ed71a1/language-en
