Automated MRI Protocolling and Scheduling: A Multi-Institutional Survey and Results

The purpose of this study was to obtain feedback from both requisition-ordering clinicians and radiologists on novel practices pertaining to electronic entry for medical imaging in a Canadian province. Particular focus was put on whether or not AI-support tools can help facilitate these changes.

While surveys in the past reflected on general undertones of clinicians regarding “fear of replacement,” or of attitude towards AI, there is relatively little amount of research depicting physician attitudes for the use of AI tools in the case of priority listing, helping with imaging wait-times, and for more local solutions pertaining to the geographical area of the Canadian province that this study pertains to. ¹

The first aspect of the survey was to conduct a questionnaire for image-ordering clinicians regarding their views on current practices for image requisition, including their views on the utility of electronic support tools in their practice. The second aspect of the survey was to broadly ask radiologists in a Canadian province on their attitudes surrounding AI-based support tools. A web-based survey called Qualtrics was used. One survey was sent to requisite ordering clinicians (n = 61), with questions pertaining to interest of primary care practitioners regarding AI utility, suggestions for improvement of MRI wait times, and suggestions for usage of electronic forms. The second survey was sent to radiologists (n = 39).

Radiologists cited reducing wait times, increasing workplace efficiency, and assisting in formulating priority (P-value) assessments as some of the many capacities that an AI prioritization initiative can help them in their practice (Figure 1).

Figure 1.

What factors explain the cause of long MRI wait times in a particular Canadian province, according to radiologists.

Additionally, 95% of clinicians would support the use of an accurate AI tool that prioritizes MRI requisitions according to the Canadian Association of Radiology (CAR) Guidelines (Table 1).

Table 1.

Key Responses from Requisition-Ordering Clinicians.

Question	Yes	No
1. If made available, would you use an electronically-generated imaging referral form, made available through your EMR or an online portal, to replace the current paper-based requisitions currently in use?	46	1
2. Do you hesitate to order an MRI because of wait times? In other words, do you refer to a different modality (US, CT, etc.) instead to decrease the wait time for your patient?	31	17
5. The MRI central intake office has a mandate to support MRI referrals from clinicians across the Lower mainland, and aims to increase the equity of access to MRI scanners, and to decrease inequity in waitlist times between the main MRI sites. Based on your current experience with referral of patients for an MRI, are you satisfied with the current MRI central intake system for medical imaging distribution processes?	20	23
8. If an evidence-based decision support tool were to be made available and/or recommended by the Lower mainland medical imaging office, would you consider using the tool for order/modality suggestion(s)?	30	3
9. Would you be open to the collection and aggregation of electronically received requisition/order data to support other regional or local quality improvement initiatives?	39	2
10. Lower mainland medical imaging completed a pilot project that developed a simple artificial intelligence tool that generates a priority value and determines the need for intravenous contrast from the MRI referral. Initial testing has shown improved workflow and waitlist equity potential. In your subjective opinion, would you support the use of an accurate artificial intelligence tool that prioritizes MRI requisitions according to the canadian association of radiology (CAR) guidelines (https://car.ca/patient-care/practice-guidelines/)? Note: This tool is intended to complement radiologist workflow, not replace it. Radiologists at each site will still be required to review all referrals and can override the priority and contrast prediction by the AI software	40	2
11. Do you support an attempt/trial period of Lower mainland medical imaging to use artificial intelligence solutions to triage MRI requisitions and smooth wait times for patients across a particular canadian province?	41	1

As evidenced by Figure 2, over 48% of clinicians are currently unaware of any evidence-based support tools that can help in their decision-making while ordering imaging studies. Additionally, while 11% of the surveyed clinicians are aware of them, they do not use them regularly in their practice. In turn, many indicate (almost 70%) that they would indeed be open to using evidence-based support tools if made available or if recommended by a local government-backed agency.

Figure 2.

Clinician awareness of evidence-based support tools for ordering of imaging studies.

Strengths of this survey include a relatively large number of participants, which ensures that there is enough data and suggestions from both image-ordering clinicians and radiologists. Through both the surveys, perspective of both sides – clinicians and radiologists alike was gained – which will further the discussion regarding steps and attitudes surrounding AI and non-AI automation to imaging.

Limitations of the study include the fact that the surveys were restricted to physicians in a particular Canadian province. As such, resources such as MRIs, and other local problems may be different for various localities – which in turn mean that the same AI modeling tools may not work in other places. Additional limitations to this study include the fact that this was only a survey, and not a quantitative assessment of the best AI systems in this space.

The results of the surveys showed that many local radiologists support various evidence-based support tools that increase process efficiency, decrease waitlist times, and potentially improve clinical outcomes (Table 1). While much needs to be done to pave the way for systemic technological change within the particular Canadian province that this study was based off of, this preliminary survey depicts that electronic entry for medical imaging should indeed be employed, and that AI should be facilitating this change. ²