Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2023 May 3;4(4):291-301.
doi: 10.1093/ehjdh/ztad030. eCollection 2023 Aug.

Development of artificial intelligence tools for invasive Doppler-based coronary microvascular assessment

Henry Seligman  1   2 , Sapna B Patel  1 , Anissa Alloula  1 , James P Howard  1   3 , Christopher M Cook  4   5 , Yousif Ahmad  6 , Guus A de Waard  7 , Mauro Echavarría Pinto  8 , Tim P van de Hoef  7 , Haseeb Rahman  9 , Mihir A Kelshiker  1   3 , Christopher A Rajkumar  1   3 , Michael Foley  1   3 , Alexandra N Nowbar  1 , Samay Mehta  1 , Mathieu Toulemonde  10 , Meng-Xing Tang  10 , Rasha Al-Lamee  1   3 , Sayan Sen  1   3 , Graham Cole  1   3 , Sukhjinder Nijjer  1   3 , Javier Escaned  11 , Niels Van Royen  12 , Darrel P Francis  1   3 , Matthew J Shun-Shin  1   3 , Ricardo Petraco  1   3
Affiliations

Development of artificial intelligence tools for invasive Doppler-based coronary microvascular assessment

Henry Seligman et al. Eur Heart J Digit Health. .

Abstract

Aims: Coronary flow reserve (CFR) assessment has proven clinical utility, but Doppler-based methods are sensitive to noise and operator bias, limiting their clinical applicability. The objective of the study is to expand the adoption of invasive Doppler CFR, through the development of artificial intelligence (AI) algorithms to automatically quantify coronary Doppler quality and track flow velocity.

Methods and results: A neural network was trained on images extracted from coronary Doppler flow recordings to score signal quality and derive values for coronary flow velocity and CFR. The outputs were independently validated against expert consensus. Artificial intelligence successfully quantified Doppler signal quality, with high agreement with expert consensus (Spearman's rho: 0.94), and within individual experts. Artificial intelligence automatically tracked flow velocity with superior numerical agreement against experts, when compared with the current console algorithm [AI flow vs. expert flow bias -1.68 cm/s, 95% confidence interval (CI) -2.13 to -1.23 cm/s, P < 0.001 with limits of agreement (LOA) -4.03 to 0.68 cm/s; console flow vs. expert flow bias -2.63 cm/s, 95% CI -3.74 to -1.52, P < 0.001, 95% LOA -8.45 to -3.19 cm/s]. Artificial intelligence yielded more precise CFR values [median absolute difference (MAD) against expert CFR: 4.0% for AI and 7.4% for console]. Artificial intelligence tracked lower-quality Doppler signals with lower variability (MAD against expert CFR 8.3% for AI and 16.7% for console).

Conclusion: An AI-based system, trained by experts and independently validated, could assign a quality score to Doppler traces and derive coronary flow velocity and CFR. By making Doppler CFR more automated, precise, and operator-independent, AI could expand the clinical applicability of coronary microvascular assessment.

Keywords: Angina; Artificial intelligence; Coronary; Diagnosis; Doppler; Invasive.

PubMed Disclaimer

Conflict of interest statement

Conflict of interest: C.M.C. has received grants or research support from Edwards Lifesciences, honoraria or consultation fees from Boston Scientific, Philips, Viz.ai, and Medtronic and is a shareholder of Cerebria.ai and Viz.ai. S.N. has received honoraria for speaking on Coronary Physiology from Phillips. H.S. has received educational funding from Amgen. R.A.-L. has received speaking fees from Philips Volcano. R.P. has acted as consultant to Philips. T.P.v.d.H. has served as a speaker at educational events organized by Philips Volcano, St Jude Medical (now Abbott), and Boston Scientific. S.S. reports speaking fees for Philips, Pfizer, and Astra Zenenca and an Educational Grant from Medtronic. N.V.R. reports grants from Philips, Abbott, and Biotronik, and personal fees from Medtronic. All other authors have no conflicts to disclose.

Figures

Graphical Abstract
Graphical Abstract
Figure 1
Figure 1
Examples of Doppler recordings from the coronary arteries during microvascular assessment. The current console algorithm is used to trace the flow velocity from Doppler spectral signals. The digitized flow is used to derive flow reserve (coronary flow reserve) values and guide clinical decisions. In the upper left panel, the Doppler flow signal is weak and so the line tracks only noise. In the upper right panel, the flow signal is stronger and so the line tracks it well. In the lower left panel, the Doppler flow signal is strong and well formed, but the flow tracking line is measuring only noise due to the parameter set by the operator. In the right lower panel, the settings are optimized and the line is following the Doppler waveform more precisely.
Figure 2
Figure 2
Summary of study design and methodology. AI, artificial intelligence; CFR, coronary flow reserve.
Figure 3
Figure 3
Correlation between quality scores generated by artificial intelligence and individual human experts is shown in top panel. Each dot represents a single score for a single image from nine experts and artificial intelligence. A Bland–Altman plot (lower panel) shows there is no systematic bias between artificial intelligence and expert consensus and a high numerical agreement. AI, artificial intelligence.
Figure 4
Figure 4
Examples of Doppler signals (extracted from coronary arteries during microvascular assessment) are shown in increasing order of quality from top left to bottom right. Quality scores generated by artificial intelligence and expert consensus are displayed (0 being the lowest possible quality and 1 the highest possible quality). A good agreement between scores can be visually appreciated.
Figure 5
Figure 5
Numerical relationships between artificial intelligence–derived, console-derived, and expert consensus flow and flow reserve values. Bland–Altman plots show that, relative to expert consensus flow and coronary flow reserve, artificial intelligence showed better agreement and less bias than the console algorithm. AI, artificial intelligence.
Figure 6
Figure 6
The same raw Doppler recording is shown as seen in the console with tracking by the console algorithm and artificial intelligence. AI, artificial intelligence; CFR, coronary flow reserve.
Figure 7
Figure 7
The mean percentage error (with spline regression of the 95% confidence interval) between each of the tracking methods (console, artificial intelligence, and individual experts) is shown across the spectrum of Doppler signal qualities (judged using artificial intelligence scores). AI, artificial intelligence.
Figure 8
Figure 8
Examples of flow tracking methods acting over Doppler envelops of different qualities. Higher individual variability among experts and increased noise from console-derived flow can be visually appreciated for lower quality Doppler signals. Artificial intelligence–derived flow is less sensitive to Doppler flow quality. AI, artificial intelligence.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

References

    1. Schumann CL, Mathew RC, Dean J-HL, Yang Y, Balfour PC, Shaw PW, et al. . Functional and economic impact of INOCA and influence of coronary microvascular dysfunction. JACC Cardiovasc Imaging 2021;14:1369–1379. - PMC - PubMed
    1. Mileva N, Nagumo S, Mizukami T, Sonck J, Berry C, Gallinoro E, et al. . Prevalence of coronary microvascular disease and coronary vasospasm in patients with nonobstructive coronary artery disease: systematic review and meta-analysis. J Am Heart Assoc 2022;11:e023207. - PMC - PubMed
    1. Kelshiker MA, Seligman H, Howard JP, Rahman H, Foley M, Nowbar AN, et al. . Coronary flow reserve and cardiovascular outcomes: a systematic review and meta-analysis. Eur Heart J 2021;43:1582–1593. - PMC - PubMed
    1. Shimokawa H, Suda A, Takahashi J, Berry C, Camici PG, Crea F, et al. . Clinical characteristics and prognosis of patients with microvascular angina: an international and prospective cohort study by the Coronary Vasomotor Disorders International Study (COVADIS) group. Eur Heart J 2021;42:4592–4600. - PMC - PubMed
    1. Heggie R, Briggs A, Stanley B, Good R, Rocchiccioli P, McEntegart M, et al. . Stratified medicine using invasive coronary function testing in angina: a cost-effectiveness analysis of the British Heart Foundation CorMicA trial. Int J Cardiol 2021;337:44–51. - PubMed

LinkOut - more resources