Hundred of Thousands of chest X-rays of COVID-19 patients are being acquired in the world. This huge amount of data could be easily analyzed with an Artificial Intelligence tool, which could provide metrics to radiologists to help them with their decisions about how to handle the patients. For instance: is it safe to release them from the hospital?

What it does

At X-COV, we want to help doctors make medical evaluations faster and more objective, giving them relevant information and metrics obtained from the chest radiographs they do. X-rays can be analyzed with Artificial Intelligence (AI) to improve the speed and reliability of medical decisions. We've designed an online tool ( that can be used by physicians to analyze x-rays locally from their browser. Without loss of privacy, allowing great scalability and its massive use. It gives an immediate analysis of the patient's severity level that medical doctors can use to assess their evolution. This AI tool analyzes chest X-rays from dicom, jpg and png formats, and provides a metric of the disease. We also plan to implement a version into the PACS system of a hospital in a pilot program. And it could be also incorporated into the chest X-rays devices.

Who are we

We’re a multidisciplinary team working from Spain and Austria, with comprehensive experience in medical imaging, translational research, and machine learning, including a radiologist with experience in radiomics and AI.

How we built it

X-COV uses a deep-learning model developed based on cases from China, Italy, Spain... We developed it with multiple tools such as Python, Javascript, Tensorflowjs, ONNXjs, Plotly.js

Challenges we ran into

The access to chest X-rays of COVID-19 patients is limited. The only public available data is biased. Without good-quality data no one couldn't do anything real or useful.

Accomplishments that we're proud of

We have been selected as one of the finalist in the Spanish hackathon #VenceAlVirus and the UNESCO hackathon #CodeTheCurve because we’ve created a prototype and trained it until achieving a precision better than 90%. Our capacity of That makes is stand out from other projects" .

We are already collaborating with four Spanish hospitals and through these collaborations we have been able to get access to hundreds of cases. We have the support of several hospitals in Madrid, especially Hospital Clinico San Carlos, which has offered their expertise and datasets, to facilitate a fast implementation.

We’ve started collaborations with companies such as SEDECAL and other research centers in the world (SENAI CIMATEC in Brazil). We’re getting feedback from radiologists and physicians from many different countries such as India, Costa Rica, and Colombia.

What we learned

1 ) It is crucial to develop these tools in close connection with the final user (in our case, radiologists) 2 ) Access to good-quality unbiased data is crucial for any AI tool. 3 ) Software security and data management are essential for online tools that analyze patient data. 4 ) Robust Business Plans must be an integral part of the development of software solutions. 5) Understanding regulation (specially related to COVID-19) is fundamental to provide working solutions.

What's next for X-COV

We will continue working on improving our tool, the webpage and the interaction with radiologists. Our next steps are:

1 ) Continue working on testing and evaluation 2 ) Adapt the tool based on the increasing feedback from users 3 ) Execute a pilot program expected to start in a few weeks. 4 ) Obtain the CE and FDA mark, following the regulation to prove the safety and efficacy of our products. 5 ) Implement within Hospital PACS system.

Built With:

1 ) Javascript 2 ) Python / Jupyter Notebook 3 ) Tensorflow

Check it out:

Built With

+ 6 more
Share this project: