Qure.ai, has released a new AI-powered technology that correctly identifies bleeds, fractures and other serious abnormalities in brain CT scans.
The healthcare AI startup, Qure.ai, has released a new AI-powered technology that correctly identifies bleeds, fractures and other serious abnormalities in brain CT scans. Qure.ai has also shared the results of an unprecedented clinical validation study that confirms its algorithms’ extremely accurate performance on 21,000 patients and has developed a data-set of close to 500 AI-analyzed head CT scans that are now downloadable.
CT scans of the head are normally the first diagnostic procedures patients undergo after suffering head injuries, or symptoms suggesting a stroke. However, radiologists are not always immediately available to analyze a scan, which is problematic for patients in critical need of prompt care. Swift reading of a CT scan is vital for stroke patients because brain cells die with every minute that goes by.
See also: How do we unlock healthcare IoT innovation?
“Qure.ai’s new head CT scan technology rapidly screens scans in under 10 seconds to detect, localize and quantify abnormalities, as well as assess their severity,” shares Prashant Warier, Co-Founder and CEO, Qure.ai. “This enables patient prioritization and the appropriate clinical intervention.”
AI learned on over 300k CT scans
Qure.ai has honed the new AI’s skills by using a group of 313,318 head CT scans, along with their corresponding clinical reports. From these, 21,095 scans were utilized to validate the AI’s algorithms. Afterward, the AI was clinically validated on 491 CT scans, and results were compared against a panel of three senior radiologists. These experts include Norbert G. Campeau, M.D., a senior neuro-radiologist from the Mayo Clinic’s Department of Radiology. The results of the study show that Qure.ai’s AI is over 95% accurate in discovering abnormalities.
“This is important new technology,” explains Dr. Campeau. “The strong results of the deep learning system support the feasibility for use of automated head CT scan interpretation as an adjunct to medical care. This improves the quality and consistency of radiologic interpretation.”
Along with the study, Qure.ai has created a dataset of 491 AI-interpreted head CT scans, as well as the corresponding interpretations from the radiologists, available to the public for download.
“We are releasing these scans so that others can compare and build on the results we have achieved,” says Warier. “We encourage others to explore the data, review our work, and consider possibilities for the future.”
