In a new survey, 83 percent of radiology residents agreed that artificial intelligence/machine learning (AI/ML) should be part of their curriculum but approximately 24 percent of residents said there are currently no AI/ML educational offerings in their residency program.
Over 80 percent of radiology residents said artificial intelligence (AI) and machine learning (ML) should be included in the curriculum for radiology residency programs and 76 percent emphasized having a continuous course on these subjects that spans all four years of residency, according to a new survey.
The survey results, published recently in Academic Radiology, were based on responses from 209 radiology residents representing 21 of 35 radiology residency programs in the Radiology Residency Education Research Alliance (RRERA). Ninety-eight respondents (47 percent) were in residency programs with more than 40 residents. The survey respondents included 79 first-year residents (38 percent), 48 second-year residents (23 percent), 42 fourth-year residents (20 percent) and 41 third-year residents (20 percent).
According to the survey, 75 respondents (36 percent) strongly agreed and 100 respondents (48 percent) agreed that AI and ML should be incorporated in the curriculum for radiology residency. Twenty-six respondents were neutral on the inclusion of AI and ML in their residency training.
When asked their preferred method of education on AI and ML, 141 respondents (67 percent) preferred hands-on and laboratory training, 127 residents (61 percent) opted for a lecture series and 86 respondents (41 percent) indicated preferences for in-house/institutional courses and online videos (whether they were available via YouTube or from professional societies). In regard to the goals of AI and ML education, 171 respondents (82 percent) wanted to be able to troubleshoot AI tools and assess whether they are working in practice. Twenty-five residents(12 percent) wanted to know how to utilize an AI tool in practice without a need to learn troubleshooting or evaluation of whether an AI system is working as intended.
“To the best of our knowledge, no prior study has surveyed radiology trainees regarding their preference for the depth of AI education or the expected outcome of such a curriculum,” wrote study co-author Lars J. Grimm, M.D., MHS, an associate professor of radiology at Duke University, and colleagues.
(Editor’s note: For related content, see “Assessing the Value Proposition of AI in Radiology,” “Recognizing and Addressing Biases with AI and Radiologists” and “Can AI Improve the Consistency of Breast Density Assessment by Radiologists?”)
Fifty-four survey respondents (24 percent) said their residency programs offered no resources for AI and ML as part of the curriculum. According to the study authors, 168 respondents (80 percent) said they had not been involved in AI or ML research and 107 residents (51 percent) had not used an AI or ML tool for work or research.
Researchers also noted a prevailing lack of awareness about currently available AI resources. Seventy-eight percent (163 residents) were not aware of AI-related videos offered by the Radiological Society of North America (RSNA), 86 percent (180 respondents) were not aware of the American College of Radiology’s (ACR) Data Science Institute, and 88 percent (183 residents) were unaware of an AI Journal club hosted by the Resident and Fellow Section of the ACR.
“These data demonstrate the need for improving awareness of existing AI education resources through better marketing/education as well as (identifying) opportunities for developing courses/curricula according to the preferred modes of learning reported by the residents,” noted Grimm and colleagues.
In regard to study limitations, the study authors acknowledged a possible selection bias with residents who are interested in AI being more likely to complete the survey. While 67 percent of respondents (141) indicated that hands-on training in AI/ML would be a preferred method of training, the researchers pointed out that only 10 percent (19) had experience with hands-on training on these topics. The study authors also noted that 27 percent of residents and 60 percent of radiology residency programs responded to the survey.
AI Mammography Platform Shows Promising Results for Detecting Subclinical Breast Cancer
October 3rd 2024Mean artificial intelligence (AI) scoring for breasts developing cancer was double that of contralateral breasts at initial biennial screening and was 16 times higher at the third biennial screening, according to a study involving over 116,00 women with no prior history of breast cancer.
The Reading Room: Artificial Intelligence: What RSNA 2020 Offered, and What 2021 Could Bring
December 5th 2020Nina Kottler, M.D., chief medical officer of AI at Radiology Partners, discusses, during RSNA 2020, what new developments the annual meeting provided about these technologies, sessions to access, and what to expect in the coming year.
Can Radiomics and Autoencoders Enhance Real-Time Ultrasound Detection of Breast Cancer?
September 10th 2024Developed with breast ultrasound data from nearly 1,200 women, a model with mixed radiomic and autoencoder features had a 90 percent AUC for diagnosing breast cancer, according to new research.