The emperor has few clothes: a realistic appraisal of current AI in radiology

Huisman, Merel; van Ginneken, Bram; Harvey, Hugh

doi:10.1007/s00330-024-10664-0

Eur Radiol

2024

DOI: 10.1007/s00330-024-10664-0

|View full text |Cite

The emperor has few clothes: a realistic appraisal of current AI in radiology

Merel Huisman,

Bram van Ginneken,

Hugh Harvey

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...

Citation Types

Supporting

Mentioning

Contrasting

Year Published

2024

Publication Types

Select...

Article4

Relationship

Self Cite0

Independent4

Authors

Journals

Cited by 4 publications

(2 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Despite AI's potency and disruptive potential, widespread adoption of AI-based CAD remains constrained for different reasons. This was also the case for 'traditional' CAD systems, where it is essential to manage expectations and challenges [5,6].Presently, we encounter first-generation AI-based CAD products in the market, and it is unsurprising that their real-world performance sometimes lags behind the lofty claims made during their introduction. Generalizability…”

mentioning

confidence: 99%

See 1 more Smart Citation

Managing expectations and challenges of AI in radiology

Meijer

2024

Eur Radiol

View full text Add to dashboard Cite

The artificial intelligence (AI) revolution is both impressive and exhilarating, as AI is already making profound changes in healthcare and radiology. Its potential advantages are evident, enhancing radiological processes across the board, from image acquisition and reporting to Computer-Aided Diagnosis (CAD) and treatment decision-making [1,2]. However, it is essential to recognize that AI solutions are tools, not magic. They come with limitations and pitfalls, including overfitting, model drift, and automation bias [3,4]. End-users must be well-acquainted with these aspects. An important point to stress is that AI is not a purpose in itself, but rather a means to enhance radiological workflows and to benefit end-users and patient outcomes. Ultimately, in clinical practice, it is the combination of domain expertise and compassionate human care that truly matters.AI is an umbrella term encompassing various technologies and applications. To address misconceptions, it is crucial to be specific and to distinguish between different AI use cases in radiology. These range from basic image acquisition and image reconstruction to complex decision support tools. In clinical practice, one area of contention centers around CAD functionality-nowadays a rather old-fashioned term we should reintroduce. Despite AI's potency and disruptive potential, widespread adoption of AI-based CAD remains constrained for different reasons. This was also the case for 'traditional' CAD systems, where it is essential to manage expectations and challenges [5,6].Presently, we encounter first-generation AI-based CAD products in the market, and it is unsurprising that their real-world performance sometimes lags behind the lofty claims made during their introduction. Generalizability

show abstract

mentioning

confidence: 99%

mentioning

confidence: 99%

Managing expectations and challenges of AI in radiology

Meijer

2024

Eur Radiol

View full text Add to dashboard Cite

show abstract

Automated assessment of brain MRIs in multiple sclerosis patients significantly reduces reading time

Sieber,

Rusche,

Yang

et al. 2024

Neuroradiology

View full text Add to dashboard Cite

Introduction Assessment of multiple sclerosis (MS) lesions on magnetic resonance imaging (MRI) is tedious, time-consuming, and error-prone. We evaluate whether assessment of new, expanding, and contrast-enhancing MS lesions can be done more time-efficiently by radiologists with assistance of artificial intelligence (AI). Methods Baseline and three follow-up (FU) MRIs of thirty-five consecutive patients diagnosed with MS were assessed by a radiologist manually, and with assistance of an AI-tool. Results were discussed with a consultant neuroradiologist and time metrics were evaluated. Results The mean reading time for the resident radiologist was 9.05 min (95CI: 6.85–11:25). With AI-assistance, the reading time was reduced by 2.83 min (95CI: 3.28–2.41, p < 0.001). The reading decreased steadily from baseline to FU3 for the resident radiologist (9.85 min baseline, 9.21 FU1, 8.64 FU2 and 8.44 FU3, p < 0.001). Assistance of AI further remarkably decreased reading times during follow-ups (3.29 min FU1, 3.92 FU2, 3.79 FU3, p < 0.001) but not at baseline (0.26 min, p = 0.96). The baseline reading time of the resident radiologist was 5.04 min (p < 0.001), with each lesion adding 0.14 min (p < 0.001). There was a substantial decrease in the baseline reading time from 5.04 min to 1.59 min (p = 0.23) with AI-assistance. Discussion of the reading results of the resident with the neuroradiology consultant (as usual in clinical routine) was exemplary done for FU-3 MRIs and added another 3 min (CI:2.27–3.76) to the reading time without AI-assistance. Conclusion We found that AI-assisted reading of MRIs of patients with MS may be faster than evaluating these MRIs without AI-assistance.

show abstract

From Wile E. Coyote to Nobel laureate: reflections on Geoffrey Hinton’s impact on radiology and AI

Kelly

2024

Eur Radiol

View full text Add to dashboard Cite

scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.

Contact Info

customersupport@researchsolutions.com

10624 S. Eastern Ave., Ste. A-614

Henderson, NV 89052, USA

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.

Blog Terms and Conditions API Terms Privacy Policy Contact Cookie Preferences Do Not Sell or Share My Personal Information

Made with 💙 for researchers

Part of the Research Solutions Family.

The emperor has few clothes: a realistic appraisal of current AI in radiology

Cited by 4 publications

References 8 publications

Managing expectations and challenges of AI in radiology

Managing expectations and challenges of AI in radiology

Automated assessment of brain MRIs in multiple sclerosis patients significantly reduces reading time

From Wile E. Coyote to Nobel laureate: reflections on Geoffrey Hinton’s impact on radiology and AI

Contact Info

Product

Resources

About