What do patients know about the use of artificial intelligence (AI) in radiology? While they may not understand much about the technology, they do have some thoughts on the subject, according to a qualitative study published online on 14 March in the Journal of the American College of Radiology.
In an attempt to ascertain the views of patients on radiology AI, a team of researchers led by Marieke Haan, PhD, of the University of Groningen in the Netherlands interviewed 20 patients after they received outpatient CT scans of the chest and abdomen. In addition to finding that patients generally lacked knowledge on radiology and AI, the researchers found that patient responses focused on six important themes: proof of technology, procedural knowledge, competence, efficiency, personal interaction, and accountability.
"The six identified domains of patients' perspective on the use of AI in radiology could provide a framework for patient education, and for future quantitative research to investigate and match patients' expectations with the development and implementation of AI systems in radiology practice," the authors wrote.
In semi-structured interviews conducted on 10 random days during July and August 2018, the researchers asked the patients for their views on six topics: the radiology department in general, AI in general, the combination of AI and radiology, the evaluation of scans by a radiologist versus a computer, receiving results from a medical doctor versus a computer, and the focus of the scan (i.e., whether the computer should only answer the questions of the referring physicians or also search for incidental findings).
The researchers found that the patients' views on radiology were diverse and often incorrect, failing to understand the differences in roles and responsibilities among different radiology department staff members. In terms of AI and radiology, the researchers found that answers centered on six main themes or domains:
- Proof of technology: Patients indicated a need for high-quality studies that prove the value and reproducibility of AI systems.
- Procedural knowledge: Patients have a lack of knowledge about radiology departments and how AI would specifically be used. As a result, clear communication between the patient, the referring physician, and the radiologist is needed on how to implement diagnostically validated AI systems in radiology.
- Competence: Patients were rather skeptical about the skills of AI systems and regarded a second reading by a radiologist to be useful.
- Efficiency: Patients believed the speed of AI systems, which may decrease waiting times, would be an advantage.
- Personal interaction: Patients unambiguously indicated a need for human interaction when receiving scan results.
- Accountability: While some patients wonder who can be held responsible for errors that computers make, others believed that radiologists will always be accountable.
"The results of this exploratory study suggest that patients' level of knowledge of AI and radiology may be rather limited, although further research is warranted," the authors wrote.
