To systematically evaluate the evolution of peer-reviewed evidence for CE-certified radiological AI products between 2020 and 2023, emphasizing advancements in demonstrating clinical relevance and patient-centered outcomes.
Key Findings:
Only 36 out of 100 CE-certified radiological AI products had supporting evidence published in peer-reviewed literature by 2020.
The analysis identified a total of 173 CE-certified radiological AI products from 90 vendors as of March 2023.
There was a steady increase in new CE-certified products from 2010 to 2020, followed by a decline in new introductions in 2021 and 2022.
Interpretation:
The limited peer-reviewed evidence for many AI products raises concerns about their clinical readiness and highlights the need for comprehensive studies on broader outcomes beyond technical performance.
Limitations:
Data on clinical evidence for AI products is not publicly available, limiting comprehensive evaluation and potentially affecting the study's conclusions.
The study excluded certain products and types of software, which may affect the generalizability of findings.
Conclusion:
The study highlights the need for more robust clinical evidence to support the adoption of AI in radiology and suggests specific strategies, such as increased funding for independent studies, to enhance its clinical impact.
