Applying Gartner’s Hype Cycle to Artificial Intelligence Research in Radiology: An Empirical Study

Purpose:
Gartner, a research and advisory firm, introduced the concept of the hype cycle in 1995. It considers hype at various stages and is essentially a "graphic representation of the maturity, adoption, and business application of specific technologies." The hype cycle demonstrates the over-enthusiasm and the hype typically accompanying technological innovation and the disappointment that follows. Artificial intelligence (AI) is thought to have the potential to change the scope of medical imaging. We aim to evaluate research attention and productivity on artificial intelligence research in radiology based on Gartner's Hype Cycle theorem.

Materials and Methods:
A Pubmed search strategy was created to identify all possible AI research in radiology and was adopted in the Altmetric Explorer as the Pubmed Query. The source of publications was filtered to ‘Radiology, Nuclear Medicine & Medical Imaging’ journals in the Web of Science Journal Citation Reports list. The number of publications over time was used to assess the productivity aspect of the hype cycle. The number of mentions from various sources was extracted using Altmetric Explorer. For each research output, an Altmetric Attention Score, which indicates how much attention it has acquired, was calculated based on the default weightings for different sources and was used to assess the hype aspect of the cycle.

Results:
This study included a total of 4298 articles from 188 journals. From 2013 to 2021, the total number of publications has increased steadily. The sum of Altmetric Attention Scores per year increased steadily from 2013 to 2020 before declining in 2021. Considering the slower but steadier increase in the number of publications and the declining attention, it could be said that Gartner’s hype cycle applies to AI research in radiology.

Conclusion:
These findings show that we are coming over the top of the hype graph since our graph closely resembled Gartner’s Hype Cycle. There are things we can do on the downward slope of the hype graph to reverse the trend. Even though the hype seems to be declining, we may be reaching a steadier slope of productivity. Increasing or steadier productivity on the AI research is more likely to ensure the clinical utilization of AI applications in radiology, rather than hype around it.

Burak Berksu Ozkara M.D.

The University of Texas MD Anderson Cancer Center

Appearances