• Brandminds 2020

COVID-19 innovations (#5) – Romanian Startup uses AI to analyse chest X-rays for signs of coronavirus

This is the fifth article in our series dedicated to innovations designed to fight COVID-19. Read COVID-19 innovations #1, COVID-19 innovations #2, COVID-19 innovations #3 and COVID-19 innovations #4.

Chest X-ray is one of the most used methods to screen and diagnose COVID-19 infections. Performing chest X-rays on suspected or confirmed patients infected with coronavirus is part of the protocol followed by healthcare professionals.

Time is of the essence when fighting this pandemic. Because the results of the PCR test may take several hours, the chest X-ray is a discriminating element. If the chest X-ray film is normal, the patient is sent home to wait for the results of the etiological test. But if the film shows pathological findings, the patient is admitted to the hospital for observation.

What does a COVID-19 infection of the lungs looks like?

The typical radiological finding related to a COVID-19 infection shows ground-glass patterned areas on the lungs.

These areas affect the lower lobes in particular and especially the posterior segments of the lungs.

On the right: coronavirus-free lungs. On the left: lungs affected by coronavirus (source)

Can AI help radiologists screen and diagnose COVID-19 more efficiently?

The top three skills that clinical radiologists need to perform at the highest level are an analytical mind, an eye for detail and good observational abilities.

Radiologists are among the healthcare frontliners fighting the pandemic. With the rise in coronavirus cases worldwide, their workload has increased exponentially.

Can AI help radiologists in their work to screen and diagnose COVID-19?

Romanian startup XVision says it can.

XVision – The AI-based application that can analyze chest X-rays for signs of coronavirus

XVision is an application that uses artificial intelligence algorithms to help radiologists with their everyday tasks.

The application is based on a proprietary pathology classification algorithm which can identify 17 of the most common chest abnormalities on x-rays.

The XVision app (source: xvision.app)

The algorithm was developed by working closely with XVision’s team of highly skilled radiologists and provides doctors with quick access to interpretations and additional information such as pathology localisation, bone suppressed and subtracted images and automatically CTR calculation. All this allows radiologists to work faster, with more precision and certainty in their investigations.

The algorithm has been trained on a dataset of more than 1 million X-rays and has a general AUC score of over 90%. (AUC stands for Area Under the Curve and is an efficient, sorting-based algorithm providing an aggregate measure of performance.) The team of radiologists at XVision ensures that the algorithm is continuously improving.

The XVision application provides radiologists with a tool they can use to triage patients and medical specialists with support in investigating radiological findings.

The application is easy-to-integrate into existing hospital infrastructure and fits perfectly into the radiologist’s usual workflow.

XVision’s Mission 

XVision was co-founded in 2018 by Andrei Tenescu, Bogdan Bercean, Cristian Avramescu, Marius Mihail Benta and Stefan Iarca with the mission to make healthcare smarter with the help of artificial intelligence.

The core company consists of a multidisciplinary team that has skills in machine learning, radiology, business development and regulatory affairs.

The team’s advisors and investors are experienced individuals that further complete their knowledge and work alongside them to provide a solution that is helpful in clinical, real-world scenarios.

XVision has raised a total of €100K in funding over 1 round from GapMinder VC, a venture capital fund, targeting IT software and services startups in Romania and Central Eastern Europe. GapMinder’s latest investment was $7M in TypingDNA.

Raul Popa (CEO of TypingDNA): Choosing the right co-founders is critical

Radiologists and other specialists have free access to the XVision application

The limited number of healthcare workers, the high-risk of contagion and the large volume of required X-rays investigations are but a few of the problems that make this crisis challenging to manage efficiently.

The XVision application is a useful tool in the fight against the COVID-19 pandemic. This is why the XVision team has decided to offer free access to their application to frontline doctors.

In the context of the aggressive SARS-COV-2 spread, XVision stays true to its goal of providing AI services for smarter healthcare. Our existing algorithms are free for use and we have adapted them to help medical workers in managing the current crisis. XVision can analyse X-rays for signs of COVID-19, detecting consolidation and ground-glass opacity patterns and associating them with their position in order to generate a score that indicates the possibility of SARS-COV-2 infection.


How does the XVision algorithm work?

Image source: xvision.app

The XVision application generates an interpretation of the X-ray on the spot based on the AI-based algorithm providing radiologists with a likelihood score.

This score describes the degree to which the algorithm believes the patterns on the patient’s X-ray are similar to those of COVID-19 confirmed cases.

The likelihood-of-COVID-19 score has been developed by the XVision team of radiologists and is based on the scores and standards used now worldwide. The score has 6 levels, starting with level 1 to level 6. The higher the score, the higher the likelihood of the patient being infected with COVID-19.

The first level indicates a normal X-ray free of infectious abnormalities, while level 6 indicates the patient has been confirmed positive by the PCR test. In practice, the algorithm will never reach level 6, 5 being the maximum level, a score which indicates the presence of imaging patterns specific to the COVID-19.

Benefits of using the XVision app

The main benefit of this application is the reduced time required to diagnose each patient as well as the increased precision of X-ray interpretation, especially during night shifts or heavy workloads caused by a pandemic like coronavirus. The AI processing works in the background and doctors are presented with the algorithm’s results.

Both healthcare workers and patients are protected against the risk of contagion by avoiding contact with suspected infected persons inside the medical unit.

The hospital saves on disinfectants because the space where X-rays are carried out is limited and therefore easier to manage.

The score provided by the XVision algorithm allows the medical staff to act swiftly and triage patients based on their needs.

Subscribe to our blog to make sure you read the next article in our series dedicated to COVID-19 innovations!

Join the Conversation

We’d love to hear what you have to say.

Get in touch with us on our LinkedIn GroupFacebook Group or Twitter.

If you liked the article, you might also like the followings: