This week at the Society for Imaging Informatics in Medicine’s 2022 meeting, a group of directors and representatives of RAD-AID detailed their efforts to implement AI in low- and middle-income countries.
RAD-AID started with the mission of increasing and improving radiology in low-resource populations. Since the charitable organization was founded in 2008, it has solidified its presence in 40 countries and 91 hospitals thanks to the help of 15,000 volunteers across the world.
Ameena Elahi, MPA, RT(R) and informatics operations director for RAD-AID detailed the three-pronged approach the organization uses when implementing AI applications in low resource areas. The “teach, try, use” method focuses first on educating personnel at each site about imaging informatics and the utility of artificial intelligence.
The “teach” portion of the approach centers on clinical radiology education and training radiologists, technologists, nurses, informaticists, etc.
The “try” portion focuses on infrastructure and the actual implementation of AI. This includes installing the necessary hardware for each modality, selecting IT personnel, setting up servers (donated by a roster of sponsors) and networks, training on PACS, RIS and EMR software, as well as provisioning Cloud resources. This takes about two weeks for volunteers to set up onsite, with remote support offered thereafter.
The final step—use—consists of slowly phasing in AI applications by encouraging radiologists to use them in specific cases and eventually the population after making patients and the public aware of its use in their care.
These efforts have resulted in the implementation of numerous AI applications—Koios, used for breast ultrasound, Qure.AI for chest radiographs and Densitas, which is used in mammography to assess breast density and correct mammographic positioning.
Steve Surrat, MD, co-director of RAD-AID Guyana detailed how the use of one of these clinical AI applications has improved the skill level of technologists at Georgetown Public Hospital Corporation in Guyana. Densitas AI was introduced there in October of 2021. Through its use, and without any formal onsite training for 6 out of 7 technologists, mammographic positioning error rates dropped from 20% to just 5%.
“This software remains very promising to us as a way to not only maintain good quality mammographic images, but also to assist in training technologists where resources are limited,” Surrat said.
In addition to Guyana, RAD-AID has been able to implement these applications in places like Peru, Nigeria and Kenya, just to name a few. As far as the future goes, Surrat stated that there is a need to determine whether there is bias between the populations used to train AI software and the populations it serves in low-resource countries, in addition to improving throughput and efficiency.
For more information on SIIM 2022, click here.
For more information on RAD-AID, click here.
Related AI and informatics content:
How to prevent cherry picking on radiology worklists
Managing incidental radiologic findings: ACR-led initiative proposes several recommendations
In addition to her background in journalism, Hannah also has patient-facing experience in clinical settings, having spent more than 12 years working as a registered rad tech. She joined Innovate Healthcare in 2021 and has since put her unique expertise to use in her editorial role with Health Imaging.