JMIR Publications

ABSTRACT

Background:

Artificial intelligence (AI) applications come with high promises for medical imaging diagnosis, increasing outcome accuracy and accelerating the diagnosis process. However, despite the expected benefit of AI applications, widespread adoption of the technology is processing slower than expected due to various obstacles that hinder the adoption of AI applications. Besides technological, organizational, and regulatory obstacles, physicians, as technology users, play a central role in adopting AI applications.

Objective:

This study aims to guide how to enable physicians to make an informed adoption decision regarding AI applications by identifying and discussing possible measures to address the barriers to adoption from physicians' perspectives.

Methods:

We employed a two-step qualitative research approach. First, we conducted a structured literature review by screening 865 articles from PubMed and ScienceDirect to identify potential enabling measures. Second, we interviewed 14 experts to evaluate the literature-based measures and enriched them.

Results:

By analysing the literature and interview transcripts we revealed eleven measures (1) Educate Physicians, (2) Prepare Future Physicians, (3) Practical Train Physicians, (4) Integrate Physicians in Technology Development, (5) Provide Transparency, (6) Show Medical Value, (7) Show Business Value, (8) Establish Central Expert Panels, (9) Establish Cross-disciplinary Teams, (10) Provide Marketplace for AI Applications, and (11) Provide Implementation Guidelines. While measures (1) – (9) can be categorised as Enabling Adoption Decision Measures, measures (10) – (11) can be summarised as Supporting Adoption Measures. Enabling Adoption Decision Measures include measures aimed at enabling physicians to make informed decisions about the appropriate adoption of AI applications in medical diagnosis. Supporting Adoption Measures, on the other hand, comprise measures to prospect that the physician receives support in the adoption process following a positive adoption decision.

Conclusions:

This study provides a comprehensive overview of measures to enable physicians to make an informed adoption decision on AI applications in medical imaging diagnosis. We contribute to the adoption research stream by pointing out a comprehensive overview of specific measures and how these measures can address the known barriers to the adoption of AI applications in medical diagnosis. Thereby, we are the first to give specific recommendations on how to realise the potential of AI applications in medical imaging diagnosis from a user perspective.