XVISION  Artificial Intelligence for Smarter Healthcare

By project ACSELL

Published on 06 July 2022

Romania

About this good practice

• By automatically distinguishing between normal x-rays and those with anomalies, XVision provides doctors with a simple way to triage or prioritise patients. Having learned how the most common 17 chest radiological findings look like, XVision’s AI models can spot their presence in radioraphs. The AI model shows where anomalies are on an image, it generates the bone suppressed and subtracted versions of a radiograph and it calculates the CTR index based on the image.
• The algorithm development process is based on a tight iteration loop between trying out new ideas and then quickly testing them both statistically and in practice, by gathering feedback from radiologists that use XVision. In order to construct a dataset, the company gather anonymized images and have their own internal radiology team annotate them. Having an internal radiology team and not outsourcing the process helps them keep and increase the quality of their datasets. Their AI development team is constantly reading the newest papers and the most recent discoveries in the field and trying to apply them in order to test what works best. Along with this, they have their proprietary innovations that helps them keeping the pace and outperforming our competition.
• They are currently working with distributors of x-ray equipment to sell XVision to public hospitals.They are currently in Romania, Hungary, Poland, Slovakia and Croatia.

Resources needed

O 4 people (3 full-time, one part-time) at coding, 5 radiologists are labeling the data needed to train their models; 2 business people asking our clients for feedback about the algorithms, process just as important as the actual coding.
O 300K EUR investment was used to implement this practice

Evidence of success

They increased the number of hospitals they were working in from 2 to 30 and the number of x-rays analysed by doctors in XVision from 50 a week to 5500 a week. XVision is a well-regarded assistant to most of the radiologists, and a very strong healthcare startup in the CEE. Since starting XVision they deploying in over 60 hospitals in 5 different markets. This technology has great potential when it comes to radiology and it may be integrated as an add-on to the medical imaging devices.

Potential for learning or transfer

The cycle of iterating the initial product on customer feedback is the basis of creating any successful scalable business. Feedback coming from real users can be invaluable in assessing the solution you are developing and in understanding if you are moving in the right direction or not. XVision is collaborating with a large and varied team of radiologist that features both resident doctors and senior radiologists. The iteration & scale-up process that is used by XVISION for growth is potentially interesting for other regions to learn from.

XVISION is on the list of MIT Innovators Under 35 Europe for this year.