Artificial Intelligence in Medicine: The Pros and Cons of AI in Radiology

I am an enthusiastic supporter of integrating Artificial Intelligence (AI) into prostate MRI and other clinical imaging modalities. However, where I see only gleaming promise, others see dark shadows. For instance, Liu, et al (2023) remark, “One of the primary worries by medical students is that AI will cause the roles of a radiologist to become automated and thus obsolete.”^[i] Perhaps this pessimism is overestimated. According to one survey, the majority of medical students would like more classroom information about AI but they’re not getting it: “… 91.5% responded that they thought AI education would be useful for their future, while 91.2% reported their schools offered no AI education resources.”^[ii]

I came across an interesting viewpoint essay in JAMA. The author, radiologist Saurabh Jha, is an Associate Professor of Radiology at the Hospital of the University of Pennsylvania/Penn Medicine. To judge from previous commentaries he has published, he recognizes the human side of radiologists facing many professional uncertainties posed by the accelerating pace of AI technologies.

In his Oct. 6, 2023 essay, “Algorithms at the Gate—Radiology’s AI Adoption Dilemma,” Dr. Jha identifies the following considerations which represent the pros and cons of AI in radiology^[iii]:

1. The business side of AI – software vendors emphasize increased productivity, allowing busy radiologists to efficiently interpret more scans thanks to AI tools that rapidly “red flag” suspicious areas. While this has the potential to ease time constraints on overworked radiologists, Jha counters with “efficiency gains might extend the dominance of a smaller labor force: fewer radiologists working more efficiently but just as intensely, and similarly predisposed to burnout.”
2. The diagnostic side of AI – on one hand, sophisticated identification/detection algorithms are a blessing because of how swiftly the AI system spots variances from what would normally be expected. On the other hand, however, such increased sensitivity has a down side in terms over false positives. Jha cites the example of mammograms for which AI was supposed to never miss cancer, but in fact many abnormalities that were picked up were not actual malignancies. He writes, “Widespread adoption of AI may change how radiologists interpret images. Inexperienced radiologists in particular may be unable to ignore AI, succumbing to an avalanche of false-positive results.”
3. The definitional side of AI – Not every patient who undergoes, say, a multiparametric MRI of the prostate will turn out to actually have cancer. Many will simply have normal results. Jha points out that a radiologist must know “normal” in order to define “abnormal,” yet even healthy individuals will vary in anatomy, tissue structure and shape, and conditions like prostatitis that aren’t life threatening but may be flagged by AI. Many incidental findings end up being red herrings that consume more reading/interpreting time, inadvertently increasing inefficiencies.

After weighing the considerations, Jha turns to a positive vision of how AI may contribute to the ever evolving role of radiologists. Virtually all physicians, whatever their specialty, now rely on accurate imaging to speed diagnosis and design treatment plans. For instance, countless lives have been spared thanks to AI identification of a blood vessel about to burst, even while the patient is still in the scanner.

Jha imagines a future expanded role for radiologists. They may play a bigger part in clinical management by activating the next step in a patient’s treatment, based on imaging results. AI might afford them an important place in information management. They may also take on coordination of programs such as lung cancer screening, planning and overseeing population-based prevention while AI does the work of detecting nodules. Jha suggests, “To harness AI’s full potential, radiologists must relinquish some of their work to algorithms and reimagine the human nonautomatable components of their work.”

Jha sums it all up with a rosy conclusion: “AI, which is still young, is already being used in the radiology value chain, from image reconstruction to report generation. Eventually, AI will allow radiologists to perform at the top of their license.

NOTE: This content is solely for purposes of information and does not substitute for diagnostic or medical advice. Talk to your doctor if you are experiencing pelvic pain, or have any other health concerns or questions of a personal medical nature.

[i] Liu DS, Abu-Shaban K, Halabi SS, Cook TS. Changes in Radiology Due to Artificial Intelligence That Can Attract Medical Students to the Specialty. JMIR Med Educ. 2023 Mar 20;9:e43415.
[ii] Hathaway, Quincy A., Jeffery P. Hogg, and Dhairya A. Lakhani. “Need for medical student education in emerging technologies and artificial intelligence: fostering enthusiasm, rather than flight, from specialties most affected by emerging technologies.” Academic Radiology 30, no. 8 (2023): 1770-1771.
[iii] Jha S. Algorithms at the Gate—Radiology’s AI Adoption Dilemma. JAMA. Published online October 06, 2023.