Abstract
After the first radiology lecture in medical school this year, one student’s question stood out, ‘Is radiology going to be replaced by Artificial Intelligence (AI)?’ Anyone with prior experience working with AI would know that this is unlikely. However, despite being reassured by the radiologist that AI is used to aid physicians rather than replace them, many students still had questions about AI and its role in the future of healthcare. Currently, there is no formal AI education in most medical schools to clarify the evolving role of AI in patient care and to address students’ misconceptions. This experience highlighted a knowledge gap that must be addressed in order to prepare medical students for a future where they will inevitably be working with AI.
Despite the interdisciplinary nature of AI, radiologists may be the best physicians to address this knowledge gap at the medical student level. As a dynamic and technology-dependent field, radiology is often first to adapt to new technologies. We have seen radiology shift multiple times, such as with the introduction of digital radiology and PACS, each time gaining the capacity to extract more information to better help patients. Consequently, radiologists take on the role of championing change by demonstrating the need for new technologies and collaborating with multidisciplinary teams to highlight the benefits of incorporating these innovations into their workflow. 1 Thus, radiologists are the most likely clinicians to be early adopters of AI and also the best situated to share their knowledge and experience with learners.
AI is a branch of computer science that focuses on solving problems and making decisions given a specific set of data. The most prominent form of AI in radiology is Machine Learning (ML) and the most prominent type of ML is supervised learning. 2 Supervised learning is used to train algorithms in order to make a prediction. In a clinical setting, the ML systems focus on improving the accessibility and quality of care patients receive, while reducing the burden on radiologists. For instance, AI systems protect patients by ensuring that the vertebral segments are counted correctly on cross-sectional imaging or that all areas of the breast are examined on a mammogram. 2 Over the next decades in radiology, AI will expand to increase our accuracy, enhance our clinical decision-making algorithms, and aid in advanced post-processing of imaging data.
One of the greatest misconceptions of AI – that it will replace radiologists – remains a source of anxiety for medical students considering a career in radiology and a reason for radiologists’ hesitancy to embrace AI. Looking at the trend of past industrial revolutions and the evolution of technology within the workforce, we can observe how the nature of jobs change and how they are flexible with the implementation of updated systems.
2
For example, with the advancement of medical imaging technology modalities in the third industrial revolution, various roles, such as patient advocacy, resource allocation and economic gatekeeping, were added to image gathering and interpretation for radiologists.
3
AI would thus be another pivotal change that would complement radiologists by changing the type of administrative and clinical work they do,
Another misconception that medical students and radiologists may have towards working with AI is that they will need to learn new skills, such as coding, in order to use AI as a radiologist. This is
The best way build this foundation of AI knowledge is to invest in high-quality, clinically-focused AI education for both undergraduate and graduate medical education programmes. 4 This requires early introduction of these topics, ideally in the first year of medical school, to dispel misconceptions and provide students with enough experiential learning time to become familiar with using AI. There are currently no standardized guidelines, objectives or competencies outlined by the Medical Council of Canada or the Royal College of Physicians and Surgeons of Canada for teaching AI to undergraduate or postgraduate medical learners, despite the movement to more competency-based training. This highlights a gap in our educational programmes and an avenue of future instructional design and research.
Within healthcare, the public is worried that AI will inhibit the physician-patient relationship and hinder the ‘humanity’ in their interactions. 5 This is likely also concern medical students, who may already worry that radiology offers limited patient interaction. In reality, AI has the potential to increase the accuracy and efficiency of care, leaving more time for physicians to spend with patients and preserving the empathetic care patients desire. The question raised by the concerned first year medical student about AI replacing radiologists is echoed all around us. It is not just students – but also our colleagues and patients – who worry for the future of radiology. This is an opportunity to arm ourselves with knowledge of AI to dispel any AI misconceptions for patients and learners, foster trust between them and the future state of healthcare and encourage learners to pursue a career in radiology.
