Artificial intelligence has become a significant force in medicine, and its use in radiology is no exception. AI’s potential to revolutionise diagnostic imaging is at the heart of a passionate debate within the medical community: will it make radiologists redundant, or can radiologists continue staying abreast of AI developments and remain essential experts in the future of healthcare?
While there are some compelling cases for AI implementations to replace radiologists, it is more likely that they will supplement the role of radiologists rather than replace them completely. In this article, we discuss why AI is unlikely to substitute radiologists entirely and how AI applications in radiology may influence the future of this profession.
Radiologist is a multifaceted job
While current AI systems can perform narrow image recognition tasks, they can’t handle an entire diagnostic workflow. Radiologists’ work entails performing a comprehensive diagnostic analysis with data from multiple sources and drawing on their medical expertise to make decisions about patient care.
Essentially, radiologists are more than just interpreters of images. They connect the findings from imaging analysis to other patient data and test results, discuss treatment plans with patients, and consult with their colleagues.
Interestingly, radiologists of the future will continue to use their expertise and will also be responsible for setting up the technical parameters of imaging examinations that AI performs.
AI’s accuracy is still a problem
A recent study published in The British Medical Journal compared the performance of a commercially available FDA-approved AI tool with radiologists who had passed the Fellowship of the Royal College of Radiologists (FRCR) exam. To pass the FRCR examination, candidates must analyse and interpret 30 radiographs in 35 minutes and correctly report at least 90 per cent of them.
When uninterpretable images were excluded from the analysis, the AI candidate achieved an average overall accuracy of 79.5 per cent, passing two out of 10 mock FRCR exams. In comparison, the average radiologist’s accuracy was 84.8 per cent, with four out of 10 mock exams passed. These results indicate that AI tools still need further development and refinement to replace radiologists.
Radiological datasets are hard to access
It is also important to consider that the rate of AI advancement in radiology may be slower and less linear than in other industries like autonomous vehicles or retail. For example, while the latest AI art generation tools like DALL-E 2 can produce jaw-dropping results, they have been trained on billions of labeled images publicly available on the internet.
In the medical field, such open-access datasets are scarce and often suffer from limited size, coverage, quality, and standardisation. Radiological datasets are also often guarded by privacy regulations and owned by vendors, hospitals, and other institutions.
What will radiologists of the future look like?
The integration of AI into radiology is likely to be like other automation technologies disrupting different industries. For example, automated checkouts have not replaced cashiers; instead, the technology allowed them to focus on more complex tasks that require customer service and interpersonal skills. Similarly, pilots are still highly sought after worldwide despite autopilot being mandatory on most commercial flights.
Although AI poses no immediate threat to radiologists, their roles will likely evolve as technology advances. The question is not whether AI will replace radiologists but whether the radiologists who can embrace and learn from technological advances will replace those who don’t.
The answer is yes: radiologists who can leverage AI to improve the accuracy and efficiency of their workflow will be better positioned to provide more accurate, efficient diagnoses that lead to better patient outcomes. In addition, there may be a shift in job roles, with many radiologists taking on more managerial and supervisory responsibilities as their primary roles.
A long road ahead
The integration of AI into radiology has the potential to revolutionise medical imaging and diagnosis. Radiologists who embrace and use this technology to better understand diseases and improve patient outcomes will have a significant advantage in the field.
However, while AI accuracy is still a challenge, and access to radiological datasets remains limited, its development is likely to be slower than that of some other industries. Consequently, for the time being, radiologists are still an essential part of healthcare and will remain so in the foreseeable future.
AI applications in radiology
Inga Shugalo is a US-based Healthcare Industry Analyst.
This article appears in Omnia Health magazine. Read the full issue online today.