One year into the pandemic: + 7,000 studies and a pledge to continue the fight with AI

Tags: Artificial IntelligenceCovid-19imaging biomarkers
One year pandemic


From 23 to +7,000 studies in 12 months: how Quibim’s Imaging Covid-19 platform became one of the strongest, freely available AI tools for classification and segmentation of Covid-19 pneumonia.

A year ago, Quibim pulled a thread that would help unravel the mystery of Covid-19 pneumonia.


The epidemic seemed confined to Asia. But signs that the virus was spreading rapidly to other parts of the world were already visible on medical images.

Radiologists were uploading images of patients presenting with a very special form of pneumonia on social media,” Quibim CEO and Co-founder Angel Alberich-Bayarri recalls. “We wondered how we could best help them diagnose this highly suspicious infection and decided to build and test a deep learning module with what we had at hand.”

This was the beginning of our journey to help medical teams detect and control Covid-19 from computed tomography (CT) images of the lungs. We asked clinicians to upload their images to our newly created Imaging Covid-19 platform to help us generate a large database and train a model that would make a difference in clinical practice and alleviate overwhelmed health services.

Quibim Covid-19 quantitative report

The response has been incredible. A few months after our call, 3,000 cases from all over the world had already been stored in our database. In just one year, more than 7,000, cases have been uploaded and 3,041 analyses launched in our platform, which is freely available here.

Providing the basis for the Imaging Covid-19 AI initiative

Thanks to everyone’s input, we’ve developed one of the best deep learning tools to automatically detect, classify and segment Covid-19 on CT. Our tool has served as a framework for a European project to enhance CT in Covid-19 diagnosis using artificial intelligence (AI).

The Imaging Covid-19 AI initiative is coordinated by the European Society of Medical Imaging Informatics (EuSoMII), Robovision and Quibim (1). The project is led by Erik Ranschaert, visiting professor of radiology at Ghent University in Belgium, and Laurens Topff, radiologist at the NKI.

The initiative has received the participation of many hospitals and radiology centers around the world and even the collaboration of the Radiological Society of North America (RSNA) (2).

Quibim’s classification and segmentation models have been praised for their performance and the module will soon be deployed in 26 hospitals.

The tool is of added value in Covid-19 diagnosis and severity assessment, as segmentation enables to quantify the volume of ground glass opacities or consolidations inside the lungs. This information could support therapy decision-making in the near future.

We are close to a solution that will significantly improve patient triage and disease evaluation,” Alberich-Bayarri said. “CT is used in disease control and we can reasonably expect that, coupled with our module, it will help make therapeutic decisions very soon.

We are extremely proud of these achievements that highlight the value of imaging biomarkers in advancing knowledge of Covid-19 disease and take us closer to our goal of improving human health. Our small contribution has had a domino effect, with now more possibilities to properly tend to patients and save more lives.

Many thanks for helping us achieve this landmark!