Innovators bring AI into imaging skills development

By Jerry Zeidenberg

October 30, 2019

Two Ontario hospital organizations – encompassing six sites – will soon deploy artificial intelligence to help with continuous learning and peer review in their imaging departments. By automatically detecting the types of cases being read by radiologists at St. Joseph’s Healthcare Hamilton and Hamilton Health Sciences, the system will deliver the latest journal findings, as well as personal pattern recognition and error avoidance, direct to their desktops.

While radiologists at all Canadian hospitals are experts in their field, with years of education and experience, our understanding of diseases and illnesses is rapidly expanding and new insights are constantly appearing. To ensure that they’re aware of the latest research and best practices, many radiologists conduct journal and web searches while they’re reading cases at the hospital, or at night, from home.

“Our radiologists and physicians spend a lot of time reading and searching for literature,” said Shairoz Kherani, who until recently was Director of Diagnostic Services at HHS. (She has since moved to Halton Health Care, in nearby Oakville, Ont., where she is Director of Diagnostic Services and Laboratory.) “Finding the right information can be a daunting process. Now it will be readily available.”

“There are hundreds of new findings every day,” said Ian Maynard, CEO of Real Time Medical, of Mississauga, Ont., the company that’s providing the AI-powered solution, called AICloudQA™.

“Radiologists can spend two or more hours a day searching independent medical data sources,” said Maynard. “Our solution saves radiologists a significant amount of time and effort by searching multiple data sources simultaneously, relative to the case at hand. We’re like a Google search on steroids for relevant medical data, helping radiologists apply the latest findings to their patient care”.

Indeed, Real Time Medical is collaborating with Google Cloud and Sightline Innovation to deliver its AI-fueled solutions. The project is also supported by the National Research Council of Canada’s Industrial Research Assistance Program (NRC IRAP), resulting in a collaboration between these organizations and the hospitals using the solution.

Not only does the automated searching save time and contribute to better medical outcomes for patients, but it helps reduce radiologist “burnout”, a serious issue today as radiologists feel overloaded by the demands placed on them, Maynard said.

St. Joseph’s Healthcare Hamilton and Hamilton Health Sciences will introduce AICloudQA for peer learning and skills development across their sites by the end of this year. The hospitals will probably start with one site, or one physician group across all sites, and then steadily roll out the solution.

The context-sensitive provision of journal articles and other sources of medical information is expected to be of great help to the radiologists, nuclear medicine physicians, cardiologists and other clinicians who use the system.

There are 70 to 80 radiologists and medical imaging experts at Hamilton Health Sciences and St. Joseph’s Healthcare Hamilton who will be the prime users of AICloudQA.

Real Time Medical’s Ian Maynard said the importance of timely and accurate information cannot be underestimated. As they’re reading cases, radiologists want the latest literature and personal pattern recognition notifications of what to be on the lookout for.

“What they don’t want our patients and their families coming back to them later, asking why they didn’t know about the latest finding from Cleveland Clinic for example,” said Maynard.

Dr. Karen Finlay, radiologist and Interim Chief of Radiology at Hamilton Health Sciences, agreed that radiologists are currently taking “a lot of time for research”.

“If a radiologist steps off a case for five to 10 minutes to go to Google Scholar, that can really add up over the course of a day,” she said. Additionally, for those familiar with the impact of interruptions on the efficiency of the diagnostic process, that time impact can be significantly magnified to the detriment of diagnostic efficiency, which collectively impacts system-wide efficiency.

The feed from AICloudQA, by contrast, is instantaneous, meaning the radiologist doesn’t have to stop what they are doing.

Notably, the Real Time Medical system also uses AI to scan the readings done by radiologists, and to provide feedback on areas where they might want to focus on or look more closely in future. “It’s like the blind spot warning system in your car, only it’s anonymously helping you avoid possible gaps in your own reading patterns,” said Maynard.

“This is very valuable,” said Kherani. “The system can do intelligent sampling and note where a radiologist may want to improve. It can even spot patterns, time of day and other conditions when they may be more vulnerable.”

Dr. Finlay observed that AICloudQA will also transform the process of peer learning at Hamilton Health Sciences and St. Joseph’s Healthcare Hamilton.

It will do this, in one way, by increasing the pool of radiologists participating. One of the limitations of current peer review methods is that there’s often a limited number of potential reviewers, especially when a sub-specialty is involved – such as breast or neuro-imaging.

Real Time Medical’s cloud-based solution offers the potential to connect with other hospitals across the province and the country, creating a critical mass of peers with a cross-section of experiences in each sub-speciality. This will enable a level of peer learning and best practice sharing that’s simply not possible with site-based systems. Increasing the number of radiologists in the peer learning pool also helps with the issue of anonymity. With site based solutions, it’s sometimes possible to guess the identity of the radiologist or clinician being assisted, as physicians are often familiar with the reporting styles of their peers.

Like all physicians – and people in general – radiologists don’t like to be judged. By making the system more anonymous, the Real Time Medical system makes peer learning more objective, valid and hence palatable for participants. This part of what is being called a “just culture” approach, that physicians are calling for in such solutions.

AICloudQA embraces the “just culture” principles that physicians want and deserve.

It is not punitive, and the information is not shared. Instead, it’s sent privately to the participating radiologist or clinicians, who can use it for self-improvement.

At Hamilton Health Sciences and St. Joseph’s Healthcare Hamilton, the peer reviewing will be prospective – that is, it’s done before the results are reported to the referring physician.

Of course, there are only so many cases that can be reviewed before the process becomes counter-productive. The need for continuous learning must be balanced with the extra burden that’s placed on reviewers.

“The trick is to make it a rich and rewarding learning experience, but not burdensome,” said Dr. Finlay.

Hamilton Health Sciences and St. Joseph’s Healthcare Hamilton currently aim to review 2 percent of the cases, which is in keeping with other Canadian programs.

Kherani noted there are other potential benefits to the AICloudQA platform. It has a workload balancing function, where it uses its intelligence to feed cases to the appropriate radiologist – based on availability and expertise.

That not only offers the organization advantages with workflow and wait times, but it also benefits patients, as they obtain the most expert radiologist available.

She said the system can eventually support different types of physicians involved in imaging, such as cardiologists, and not only radiologists. “It’s a multi-ology solution.” Dr. Finlay noted the system also supports critical results reporting – so that urgent findings are quickly sent to referring doctors. It can also be tweaked to include notification of unexpected findings – flagging colleagues about problems that were unanticipated, but should be addressed.

Link to the original source.

Recommended Posts
0
Artificial Intelligence for radiologyquality assurance in diagnostic radiology