Can AI Solve the Radiology Workflow Problem?
Welcome to our December newsletter on neuroradiology and AI where we have collected some of the recent news on whether AI can dealing with increasing radiology workflow. We hope you find it helpful. Please feel free to share it with others.
If you would like to be informed when the next Cercare Bulletin is out, please sign up here.
CT causes on-call radiology workloads to jump almost 300% over 15 years. A Dutch study in Insights into Imaging showed an almost 300% increase in radiology workloads in the past 15 years, due mostly to increases in CT usage. Authors warned that hospitals needed to hire more radiologists and technicians, to avoid the rising radiologist burnout rates, which can harm patient safety. They saw no potential future decrease or stabilization of workload during on-call hours. Their study analyzed imaging performed at the authors’ institution over 15 years, during on-call hours. Other options for managing the increased workflow, authors said, was to use AI, outsource some work to teleradiologists, or use physician assistants to manage routine imaging tasks.
AI’s beneficial effect on turnaround time in interpreting head CTs with intracranial hemorrhage. This study in Radiology: Artificial Intelligence showed how a Texas hospital reduced wait times and turnaround times, by implementing an AI tool to help reprioritize CT work lists for radiologists. The algorithm, studied from 2017-2019, flagged abnormal, non-contrast exams with intracranial hemorrhages. By reprioritizing the worklists using AI, wait times dropped to about 12 minutes per study, from 15.75 minutes.
ACR is making AI technology more accessible to radiologists. The American College of Radiology (ACR) is expanding ACR Connect, its communication services platform that has APIs to allow seamless data exchange with vendor products at local facilities. It interfaces with the National Radiology Data Registry and the Qualified Clinical Data Registry, used for the physician quality reporting system. In expanding ACR Connect, they will allow for federated learning. They are currently testing their first version with seven sites, and more on the way. Their goals are for ACR Connect to be user-friendly and secure, allowing imaging sites of all size and sophistication levels, to participate.
Preparing senior radiology residents with a focused data science AI pathway. In November, we shared Dartmouth-Hitchcock Medical Center’s approach to teaching their radiology residents about AI. Brigham and Women’s Hospital (BWH) at Harvard Medical School is taking a different route. Their 4th year radiology residents can elect to take the Data Science Pathway program, which introduces AI with a flexible schedule of research, educational and experiential activities, in collaboration with the Massachusetts General Hospital and BWH Center for Clinical Data Science. The initiative is explained in Radiology: Artificial Intelligence. The residents participated in didactic sessions, but were also integrated full time into AI projects, including model design, application development, data curation, quality control and clinical testing. The work led to 12 abstracts accepted into national meetings.
What to consider when purchasing AI. Not all radiology departments have a good grasp on what to look for or think about when purchasing and implementing an AI program. University of California San Francisco’s department of radiology and biomedical engineering published a checklist for radiology departments seeking to purchase AI. It includes ensuring that the AI program addresses a well-defined problem in the practice, the ease and use of workflow integration, monitoring the programs for patient safety and unintended bias, and the cost implications on IT infrastructure. Radiology: Artificial Intelligence also published a piece on how to consider purchase decisions, with a market overview of current software offerings based on type, subspecialty and modality.
As AI becomes more accepted, the radiology community is looking at better ways to integrate it into the workflow, get it into the residency curriculum, and determine the factors needed to make purchase and integration decisions. The radiology workload is increasing, as more studies are ordered and referring physicians want results back quickly. It’s not easy to train radiologists, and the training pipeline is long. That means radiology departments and medical centers need to find other ways to get quality results back in a time-effective manner. AI can help with both goals. As more residency programs and medical schools expose and train physicians in AI development and concepts, the field will gain acceptance and interest. Medical societies making imaging databases available and promoting data acquisition abilities will also continue to move the field forward.
Would you like to stay up-to-date with the latest news in AI and neuroradiology? Stay informed with our Cercare Bulletin.
Read other news