Ask any radiologist, and they will tell you medical image data is on the rise; diagnostic imaging procedures are expected to increase 10-12% annually in Europe alone¹. Combined with well-documented radiologist shortages like those in the UK², it’s no secret that doctors (and patients) are feeling the crunch. Physicians are forced to choose between longer working hours or decreased time spent evaluating images. In addition, the way radiologists search for case-relevant information has not changed in over a decade. It’s a frustrating and time-consuming process in which they have to describe complex visual structures in their own words and search through multiple books or reference websites for relevant reference cases or consult with other physicians.
contextflow revolutionizes this information search process with the help of deep learning; their Vienna-based firm has developed a 3D image-based search engine for radiologists. Their solution searches for regions of interest within images (currently 18+ patterns in lung CTs). When a suspicious pattern is detected, they immediately find visually similar cases, descriptions and relevant reference information, affording doctors easy access to information currently hidden in large medical image data and their associated reports, all within seconds. No more leafing through books, guessing the right keyword or consulting various text-based reference search engines. Furthermore, the software can be configured to allow for prioritization of cases where patterns of interest are detected.
Although contextflow is currently focused on lung CT scans for their market entrance, their search engine offers a general solution which can be extended to other anatomies and modalities, setting them apart. Another point of distinction, contextflow’s aim is not to make a diagnosis for you; rather, they use AI to assist you in determining differential diagnosis for difficult cases. This makes their solution particularly valuable in teaching scenarios. Their next challenges: analyzing patient trajectories and improving the user interface to make it completely intuitive.
contextflow is a spin-off from the Medical University of Vienna and European research project KHRESMOI and was founded by a team of AI and engineering experts in July 2016. With their four partner hospitals, they have developed a knowledge sharing platform that already has more than 10,000 anonymized Lung CT scans (>3mio images) and anonymized reports that can be used as a search database as well as for training their algorithms. With more than 250 peer reviewed publications, the contextflow team of 13 has a strong scientific background. They are supported by two incubators, INiTS and i2c TU Wien, guided by experienced sales experts and develop their technology in close collaboration with radiologists from their partner hospitals. contextflow received the Most Promising Startup award by the BCS Search Industry, the Austrian Digital Innovative Solution Award in 2017 and was chosen as one of 19 startups out of over 700 submissions to participate in the Philips Healthworks Accelerator in 2018.
Currently, contextflow is working on obtaining certification as a CE Class 1 medical device and expect their first sales later this year. They have received public grants from multiple Austrian institutions like the AWS, FFG and WAW, as well as H2020, and closed their first seed investment round from IST Cube and APEX Ventures in the spring of 2018. They will extend their knowledge sharing database with additional proof of concept partners and PACS in 2019, launching their next funding round in July 2019.
Together with their radiology partners, contextflow’s goal is to develop the next generation of deep learning-driven tools for radiologists to save time, increase convenience and decrease uncertainty. In doing so, they ensure that doctors can provide patients access to the highest quality of care, exactly when they need it.
To find out more about contextflow and the technology they are developing, be sure to visit them at the Artificial Intelligence Exhibition (AIX) at ECR 2019!
¹ Medical Image Sharing And Management Drives Collaborative Care: Overcoming Fragmentation To Create Unity(Rep.). (n.d.). Retrieved https://www.emc.com/collateral/analyst-reports/4_fs_wp_medical_image_sharing_021012_mc_print.pdf
² Clinical Radiology UK Workforce Census 2017 Report(Rep.). (2018, September). Retrieved https://www.rcr.ac.uk/system/files/publication/field_publication_files/bfcr185_cr_census_2017.pdf
* Contents provided by contextflow and adapted by the ESR office