RPS 605c - Different views on artificial intelligence (AI) and machine learning (ML)

Purpose:

To explore the attitudes of UK medical students regarding AI, their understanding, and career intention towards radiology. We also examined the state of education relating to AI amongst this cohort.

Methods and materials:

Students from UK medical schools were invited to complete a survey consisting of Likert and dichotomous questions.

Results:

484 responses were received from 19 UK medical schools. 88% of students believed that AI will play an important role in healthcare and 49% reported they were less likely to consider a career in radiology due to AI. 89% of students believed that teaching in AI would be beneficial and 78% agreed that students should receive training in AI as part of their medical degree.

Only 45 students received any teaching on AI; none received such teaching as part of their compulsory curriculum. Students that received teaching in AI were more likely to consider radiology (p=0.01). Despite this, a large proportion of students in the taught group reported a lack of confidence and understanding required for the critical use of healthcare AI-tools.

Conclusion:

UK medical students understand the importance of AI and are keen to engage. Medical school training of AI should be improved. Realistic use-cases and limitations of AI must be presented to students so they will not feel discouraged from pursuing radiology.

Limitations:

Not all UK medical schools were represented in the current study and response rates differed between different institutions, potentially introducing bias. Associated limitations relating to survey-based observational studies would apply. For example, potential misuderstanding or misintrepretation of questions by the respondents.

Ethics committee approval

Minimal risk ethical approval was obtained from the King’s College London research ethics office (KCL-MRA18/19-11127).

Funding:

No funding was received for this work.

Purpose:

To investigate the opinion of radiologists and IT experts on artificial intelligence (AI) and its future impact on radiological work.

Methods and materials:

During a national meeting for AI, eHealth, and IT-infrastructure in 2019, a survey was conducted to obtain the participants’ opinion on AI in radiology. 131 participants (42 radiologists, 89 IT experts/non-radiologists) took part in the survey. A 7-point Likert scale (1: “I disagree at all” to 7: “I totally agree”) was used to assess the views of radiologists, IT experts in hospitals, and healthcare companies.

Results:

All participants agreed that medicine will become more efficient with the use of AI (5.98) and that plausibility checks will be important to be able to understand the decisions of the AI (6.30). All participants stated that the validation of AI algorithms in clinical studies is mandatory (6.32).

Radiologists rated the statement that AI will lead to a change in the working environment of all physicians significantly higher than non-radiologists (6.38 vs. 5.85, p=.04). The need to inform patients about the use of AI (4.12 vs. 5.2, p=.009) was rated significantly lower among radiologists.

Conclusion:

Radiologists and IT experts share various attitudes regarding future use on AI, especially with regard to the improvement in medical care and the need to validate algorithms in clinical studies. The views of radiologists and IT experts differ on some specific issues, such as the impact of AI on physicians’ work and whether patients need to be informed prior to the use of AI.

Limitations:

Not all participants of the meeting participated in the study, so a selection bias is possible.

Ethics committee approval

n/a

Funding:

No funding was received for this work.

Purpose:

To introduce a curricular course for medical students which addresses artificial intelligence (AI) in medical education for the first time.

Methods and materials:

“Medicine in the digital age” represents an interdisciplinary course in medical school that is taught in 5 modules on digital topics such as medical apps, augmented reality, big data, individualised medicine, and AI.

The course was held in 3 consecutive terms. All 32 participants were questioned in focus groups using guideline interviews. The questions of the guideline were first formulated structurally and then in terms of content, from which the 3 interview focuses emerged: fields of applications of AI, Work 4.0, and critical reflection. The interviews were evaluated using qualitative content analysis according to Philipp Mayring.

Results:

47% of the students’ statements could be assigned to Work 4.0, 31% to critical reflection, and 22% to fields of application of AI. Within the category Work 4.0, the topic human-machine interaction accounted for the largest share of statements with 56%. The fears and worries of the students concerning AI played only a minor role with 12% of the statements in this category. The category critical reflection describes a holistic view of medicine in the digital age on the level of social abstraction and self-reflection (e.g. practical implementation, ethics).

Conclusion:

The course conveys the knowledge of digital tools, the acquisition of skills, and the development of an attitude towards AI. The students become aware of the process of change through the increasing digitalisation and recognise the necessity to acquire new competencies for their future.

Limitations:

Currently, the course is offered within the scope of the compulsory elective week. An expansion for all students is the aim.

Ethics committee approval

n/a

Funding:

No funding was received for this work.

Purpose:

To provide a perspective of radiology fellows about the current trends and policy tools for promoting mobility among young radiologists, especially in the EU and USA.

Methods and materials:

Evidence-based medicine requires multicentre collaboration to identify a best practice, standardise it, and share it. Mobility helps to uniform techniques and terminology in different countries, which are crucial to develop widely-shared guidelines.

Results:

While the EU has many talented and skilled researchers, they account for a significantly lower share of the labour force than is the case in the USA.

Given advances in communication technology, a core group of networked researchers may go a long way towards helping a country with modest scientific resources achieve the analogue world-class excellence of the richest countries, in a broader win-win situation. However, these new avenues will require strong leadership and enhanced institutional autonomy. International organisations such as the ESR and RSNA can play a role to involve local centres into global science projects. With the help of these societies, a radiology trainee can easily take advantage of the international training opportunities that are currently offered by public or private grants, enhancing the scientific mobility and the cooperation among research centres of different countries.

Moreover, the cultural-knowledge and the networks developed during mobility can be used by the trainees to advance their career.

Conclusion:

Today, it is not just being a certified radiologist that matters, the place where training is held plays a role when applying for a high-level position. Mobility of trainees is an indispensable prerequisite to facing new challenges, including the application of artificial intelligence to medical imaging, which will require a large multicentre collaboration.

Limitations:

n/a

Ethics committee approval

n/a

Funding:

No funding was received for this work.

Purpose:

Residents' performance quality assurance and improvement is a priority for residency programs in all departments, let alone in radiology. There is a constant need for a general quality control system that covers all modalities in which residents are engaged.

Our radiology department implemented a residents' quality grading system (RQC). The goal of this system is to provide data that will demonstrate residents' progress and areas of strength and weaknesses, hence, tailoring the residency program individually.

Methods and materials:

A computerised information system was developed and deployed. The system enables attending radiologists to grade residents’ reports on a scale of 1-4, according to the following:
1) Complete agreement of the attending radiologist with the resident's report.
2) General agreement between reports, with minor changes.
3) Major changes without clinically significant impact.
4) Major changes with clinical significance that can gravely affect the patient.

The attending radiologist inserts the appropriate grade in the report and the resident is able to see it instantly.

Results:

The RQC system has been implemented successfully in our department, with high compliance rates. Residents have been able to receive on-line feedback, understanding the clinical impact of their reports immediately. Moreover, the department is able to monitor a resident's progress and areas of weaknesses, thus directing his/her residency program accordingly.

Conclusion:

RQC is a new system implemented by our radiology department with a high compliance rate, enabling constant feedback on residents' radiology reports. It continuously provides essential data that will impact residents' progress and quality of radiology reports, thus tailoring the residency program on an individual basis.

Limitations:

Limitations include variability among evaluating radiologists, bias based on previous impression, shift workload, time of shift, and resident seniority.

Ethics committee approval

n/a

Funding:

No funding was received for this work.

Purpose:

Key opinion leaders predicting implementation issues will determine the course of AI (artificial intelligence) in radiology. The perception at large by radiologists and residents of AI remains unexplored. We investigated the existing knowledge and overall attitude towards AI by international radiologists and residents to help facilitate implementation.

Methods and materials:

Between April-July 2019, a multi-language survey was accessible to radiologists and residents containing questions on awareness, knowledge, and attitude towards AI. Relationships of independent variables with an open and proactive attitude towards AI (readiness to use and learn about AI, to collaborate with data scientists, and agreement that radiologists should take the lead) were assessed using multivariable logistic regression.

Results:

The survey was completed by 1,041 respondents from 54 countries (mean age 41 (range 24-74)). Most participants were male (n=670, 65%), radiologists (n=719, 69%), and working in non-academic centres (n=572, 55%) without formal research training (n=727, 70%). Over half had no knowledge of informatics/statistics (n=537, 52%). A minority had profound knowledge of AI (n=168, 16%). Almost half of the participants appeared to have an open and proactive attitude towards AI (n=501, 48%), which was significantly (p<0.05) associated with sex, age, working in an academic centre, having received formal research training, pre-existent knowledge on informatics/statistics, and professional social media use.

Conclusion:

Almost half of the participants showed an open and proactive attitude towards AI, however, a substantial proportion had no knowledge of informatics/statistics. Pre-existent knowledge was independently associated with a positive attitude, indicating a need for additional training to facilitate implementation.

Limitations:

Selection bias is presumably an issue in this survey, therefore the true percentage of the population having an open and proactive attitude is most likely lower.

Ethics committee approval

n/a

Funding:

No funding was received for this work.

Purpose:

Integration of AI education in residency programs is not yet common. As residency programs are already demanding, controversy exists as to what extent it should be incorporated, although it might accelerate the implementation of AI in radiology. We sought to explore the opinion of the radiology community and to identify commonly anticipated hurdles to implementation.

Methods and materials:

Between April-July 2019, a multi-language survey was accessible to radiologists and residents including multiple-choice questions on the integration of AI in residency programs and the anticipated hurdles to implementation of AI in radiology.

Results:

The survey was completed by 1,041 respondents from 54 countries, mean age 41 (range 24-74), mostly male (n=670, 65%), and radiologists (n=719, 69%). A majority (n=819, 79%) indicated AI should be incorporated in residency programs, the remainder indicated maybe (n=182, 18%) or disagreed (n=40, n=4%). A small majority indicated that AI should (n=241, 23%) or maybe should (n=359, 35%) become a radiology subspecialty, while almost half (n=437, 42%) disagreed. Indicated hurdles to implementation were mainly ethical/legal issues (n=630, 61%), limitations in digital infrastructure (n=356, 61%), lack of knowledge (n=584, 56%) of stakeholders (i.e. clinicians, radiology staff, or management), and generalisability of AI algorithms (n=400, 38%).

Conclusion:

A large majority of responders favours incorporation of AI into radiology training. Anticipated hurdles to implementation were mainly ethical and legal issues, and a lack of knowledge and generalisability issues, which is in concordance with key opinion leaders. Based on these results, incorporation of AI training in residency programs seems highly advisable, including ethical and legal aspects as well as methodology to ensure safe and effective use of AI.

Limitations:

Selection bias is present, therefore the opinion of the population will be more moderate.

Ethics committee approval

n/a

Funding:

No funding was received for this work.