By Mark Watts
I attended the European Society of Radiology’s virtual 2020 European Congress of Radiology (ECR). It was striking to see the 100 exhibits with imaging artificial intelligence being offered.
There were 113 posters listed under the research title artificial intelligence (AI). Examples are:
AI has been described as the use of computers to simulate the human’s intelligence, a characteristic intellectual process of human beings. Image recognition, image classification, segmentation of tissues and organs and localization of various pathologies are the main applications of machine learning in radiology.
The implementation of AI in the field of medical imaging is continuously increasing, and there is a consensus that radiology is the most upfront branch of medicine regarding the use of these technologies. The number of publications related to AI in medical imaging, strengthen this belief, as there has been a dramatic increase in the papers regarding AI the past two years .
However, some radiologists worldwide, seem to see AI as a threat to their profession, instead of welcoming this new technology. This belief is based on the fears that with the wide use of AI in radiology, there will not be any need for trained radiologists in the future. Many anxieties exist among radiologists, especially junior trainees, as some predictions see disciplines such as radiology being replaced by AI in the future. Radiologists interpret large amounts of data daily, however, an overloaded workflow in conjunction with thousands of images can make radiologists more vulnerable to reporting errors. AI is a helpful tool for radiologists, as it can make the workflow faster and better. Besides, it can offer the advantage of quantitative information, such as Radiomics signatures. Therefore, as the European Society of Radiology (ESR) outlines, radiologists, and especially trainees, must not be afraid of AI applications. The lives of radiologists will change in the light of AI, however, AI will not be able to replace radiologists, as they are the only ones who can solve complex clinical problems. Education regarding AI must be implemented in the radiologists’ educational programs, to help radiologists feel safe and confident about the use of AI in medical imaging.
A specifically designed questionnaire with 12 closed questions was distributed to all the radiologists on a large Greek island with imaging centers operating within the private and public sectors.
Out of 22 active radiologists working on the island, 22 valid responses were received. The response rate was 100%.
The participants were asked to define when they think AI will be implemented in clinical practice in Greece. Most of the responders (63.6%) answered that AI will need between 5 and 10 years to be widely used in Greece. Similarly, 27.3% of the responders reported that it will take more than 10 years for AI to be in the daily life of Greek radiologists, while only 9.1% think that AI will have been implemented in clinical practice in less than 5 years. Within the literature, it is reported that high-income countries have already integrated AI in their health care systems. However, there is no widespread use of AI in Greece within the discipline of medical imaging. Therefore, systematic national strategy plans must be established to promote the integration of AI into health care in Greece.
Regarding the level of knowledge, the participants have related to AI applications, exactly half of them reported being only a little informed about AI. In addition, the vast majority (95.5%) reported that they have not received any education regarding AI. However, almost all the responders (90.9%) reported that AI must be integrated into the educational programs of both radiologists and technologists in the future. This is in line with the recommendations of the ESR, as it strongly recommends that an AI-based module must be included in the educational curriculum of radiologists. This will be an effective way of training young radiologists how to benefit from AI applications and integrate them into clinical practice.
The results about any past or present use of AI in clinical practice strengthen the belief that AI has not yet been widely integrated into medical imaging within the country. Specifically, most of the responders (63.6%) reported that they have not ever used AI during their career, while only 2 (9.1%) of the responders reported a serious experience regarding AI in clinical practice. Similarly, the scenery has not changed, as only 5 (22.7%) of the radiologists reported that they are currently using AI applications in clinical practice, while the vast majority (77.3%) reported not using such applications.
Despite the generally low level of knowledge regarding AI, a relatively positive perspective of AI was noted. Specifically, exactly half of the responders reported seeing the implementation of AI in radiology rather positively. Similarly, 9 (40.9%) of them reported positive thoughts about AI in radiology, while only 2 (9.1%) of them reported a neutral attitude towards AI. None of the responders reported having negative or rather negative perspectives of AI in radiology.
Most of the responders (68.2%) think that AI cannot substitute radiologists in the future and that their profession is not at risk from the implementation of AI in clinical practice. Similarly, 5 (22.7%) of them think that there is a possibility that AI will be able to replace radiologists in the future.
Most of the responders (81.8%) believe that AI is a useful tool for radiologists, while 2 (9.1%) of them see AI as a moderate tool for them. Regarding the implementation of AI in clinical practice in relation to the various imaging modalities, most of the responders (59.1%) believe that AI must be integrated into all the imaging modalities. This is justified within the literature, as AI-based algorithms can be used in computed tomography (CT), magnetic resonance imaging (MRI) and mammography to facilitate with image interpretation, characterization of lesions, diagnostic predictions, as well as for making radiation treatment planning faster and more accurate. Therefore, AI should be integrated in all imaging modalities to improve diagnosis, treatment and outcomes.
In summary, the USA and RSNA can take away these key points:
- This study noted a relative lack of knowledge and training regarding AI applications among Greek radiologists.
- Currently, there is a lack of experience in using AI in clinical practice.
- The majority of the responders seem to have positive perceptions of AI-based applications and most of them consider AI as a helpful tool for radiologists.
- There is a consensus that AI will be widely implemented in Greek radiological departments within 5-10 years.
- AI must be integrated into radiological education, and Greece must invest in AI-based algorithms to improve diagnosis, treatment and workflow.
This is a limited study, but I think there are pearls of wisdom offered here. I look forward to creating an improved future in medical imaging with AI. Plan, train and use this tool.
Mark Watts has over 20 years as an imaging professional with vast expertise in imaging informatics and IT issues. He has served in many roles in both hospitals and industry as a health care vice president, imaging director, and IT consultant. His knowledge and experience in the convergence of IT and imaging has made him a sought after author, speaker and consultant. He has authored a textbook on informatics and was a pioneer in the adoption and development of PACS and VNA technologies.