A survey conducted during a 2019 meeting for AI, eHealth, and IT infrastructure, aimed to assess participants’ attitudes on artificial intelligence (AI) and its future impact on radiological work.
Investigators utilised the Kruskal-Wallis test to identify differences between a group of 123 radiologists, IT specialists, and industry representatives.
Participants were questioned on 28 topics related to AI usage in medicine:
- 93% of respondents agreed that plausibility checks are important to understand the decisions of the AI;
- 91% felt that validation of AI algorithms is mandatory, and
- 86% believe that AI improves efficiency in medicine.
Surprisingly, only 25% of the queried respondents had confidence in the AI results, and a mere 17% felt that the use of AI will further humanise medicine.
The discrepancy between the high expectations for the future role of AI in medicine and the low confidence in the field of AI results was similar across all three groups polled. Study authors note that the clear desire of respondents for plausibility checks signals that future randomised controlled studies into AI decision-making are necessitated.
Source: Academic Radiology