Nearly automated analysis of coronary Doppler flow velocity from transthoracic ultrasound images: validation with manual tracings

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

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.

Improvement of automated analysis of coronary Doppler echocardiograms

Coronary artery disease is the leading cause of heart disease, and while it can be assessed through transthoracic Doppler echocardiography (TTDE) by observing changes in coronary flow, manual analysis of TTDE is time consuming and subject to bias. In a previous study, a program was created to automatically analyze coronary flow patterns by parsing Doppler videos into a single continuous image, binarizing and separating the image into cardiac cycles, and extracting data values from each of these cycles. The program significantly reduced variability and time to complete TTDE analysis, but some obstacles such as interfering noise and varying video sizes left room to increase the program's accuracy. The goal of this current study was to refine the existing automation algorithm and heuristics by (1) moving the program to a Python environment, (2) increasing the program's ability to handle challenging cases and video variations, and (3) removing unrepresentative cardiac cycles from the final data set. With this improved analysis, examiners can use the automatic program to easily and accurately identify the early signs of serious heart diseases.

Impact of pulsed-wave-Doppler velocity-envelope tracing techniques on classification of complete fetal cardiac cycles

Fetal echocardiography is an operator-dependent examination technique requiring a high level of expertise. Pulsed-wave Doppler (PWD) is often used as a reference for the mechanical activity of the heart, from which several quantitative parameters can be extracted. These aspects suggest the development of software tools that can reliably identify complete and clinically meaningful fetal cardiac cycles that can enable their automatic measurement. Several scientific works have addressed the tracing of the PWD velocity envelope. In this work, we assess the different steps involved in the signal processing chains that enable PWD envelope tracing. We apply a supervised classifier trained on envelopes traced by different signal processing chains for distinguishing complete and measurable PWD heartbeats from incomplete or malformed ones, which makes it possible to determine the impact of each of the different processing steps on the detection accuracy. In this study, we collected 43 images and labeled 174,319 PWD segments from 25 pregnant women volunteers. By considering seven envelope tracing techniques and the 23 different processing steps involved in their implementation, the results of our study reveal that, compared to the steps investigated in most other works, those that achieve binarisation and envelope extraction are significantly more important (p < 0.05). The best approaches among those studied enabled greater than 98% accuracy on our large manually annotated dataset.

Stable Automatic Envelope Estimation for Noisy Doppler Ultrasound

Doppler ultrasound technology is widespread in clinical applications and is principally used for blood flow measurements in the heart, arteries, and veins. A commonly extracted parameter is the maximum velocity envelope. However, current methods of extracting it cannot produce stable envelopes in high noise conditions. This can limit clinical and research applications using the technology. In this article, a new method of automatic envelope estimation is presented. The method can handle challenging signals with high levels of noise and variable envelope shapes. Envelopes are extracted from a Doppler spectrogram image generated directly from the Doppler audio signal, making it less device-dependent than existing image-processing methods. The method's performance is assessed using simulated pulsatile flow, a flow phantom, and in vivo ascending aortic flow measurements and is compared with three state-of-the-art methods. The proposed method is the most accurate in noisy conditions, achieving, on average, for phantom data with signal-to-noise ratios (SNRs) below 10 dB, bias and standard deviation of 0.7% and 3.3% lower than the next-best performing method. In addition, a new method for beat segmentation is proposed. When combined, the two proposed methods exhibited the best performance using in vivo data, producing the least number of incorrectly segmented beats and 8.2% more correctly segmented beats than the next best performing method. The ability of the proposed methods to reliably extract timing indices for cardiac cycles across a range of signal quality is of particular significance for research and monitoring applications.

Automatic detection of complete and measurable cardiac cycles in antenatal pulsed-wave Doppler signals

BACKGROUND AND OBJECTIVE

Pulsed-wave Doppler (PWD) echocardiography is the primary tool for antenatal cardiological diagnosis. Based on it, different measurements and validated reference parameters can be extracted. The automatic detection of complete and measurable cardiac cycles would represent a useful tool for the quality assessment of the PWD trace and the automated analysis of long traces.

METHODS

This work proposes and compares three different algorithms for this purpose, based on the preliminary extraction of the PWD velocity spectrum envelopes: template matching, supervised classification over a reduced set of relevant waveshape features, and supervised classification over the whole waveshape potentially representing a cardiac cycle. A custom dataset comprising 43 fetal cardiac PWD traces (174,319 signal segments) acquired on an apical five-chamber window was developed and used for the assessment of the different algorithms.

RESULTS

The adoption of a supervised classifier trained with the samples representing the upper and lower envelopes of the PWD, with additional features extracted from the image, achieved significantly better results (p < 0.0001) than the other algorithms, with an average accuracy of 98% ± 1% when using an SVM classifier and a leave-one-subject-out cross-validation. Further, the robustness of the results with respect to the classifier model was proved.

CONCLUSIONS

The results reveal excellent detection performance, suggesting that the proposed approach can be adopted for the automatic analysis of long PWD traces or embedded in ultrasound machines as a first step for the extraction of measurements and reference clinical parameters.

Defining Coronary Flow Patterns: Comprehensive Automation of Transthoracic Doppler Coronary Blood Flow

The coronary microcirculation (CM) plays a critical role in the regulation of blood flow and nutrient exchange to support the viability of the heart. In many disease states, the CM becomes structurally and functionally impaired, and transthoracic Doppler echocardiography can be used as a non-invasive surrogate to assess CM disease. Analysis of Doppler echocardiography is prone to user bias and can be laborious, especially if additional parameters are collected. We hypothesized that we could develop a MATLAB algorithm to automatically analyze clinically-relevant and non-traditional parameters from murine PW Doppler coronary flow patterns that would reduce intra- and inter-operator bias, and analysis time. Our results show a significant reduction in intra- and inter-observer variability as well as a 30 fold decrease in analysis time with the automated program vs. manual analysis. Finally, we demonstrated good agreement between automated and manual analysis for clinically-relevant parameters under baseline and hyperemic conditions. Resulting coronary flow velocity reserve calculations were also found to be in good agreement. We present a MATLAB algorithm that is user friendly and robust in defining and measuring Doppler coronary flow pattern parameters for more efficient and potentially more insightful analysis assessed via Doppler echocardiography.

Automated aortic Doppler flow tracing for reproducible research and clinical measurements

In clinical practice, echocardiographers are often unkeen to make the significant time investment to make additional multiple measurements of Doppler velocity. Main hurdle to obtaining multiple measurements is the time required to manually trace a series of Doppler traces. To make it easier to analyze more beats, we present the description of an application system for automated aortic Doppler envelope quantification, compatible with a range of hardware platforms. It analyses long Doppler strips, spanning many heartbeats, and does not require electrocardiogram to separate individual beats. We tested its measurement of velocity-time-integral and peak-velocity against the reference standard defined as the average of three experts who each made three separate measurements. The automated measurements of velocity-time-integral showed strong correspondence (R(2) = 0.94) and good Bland-Altman agreement (SD = 1.39 cm) with the reference consensus expert values, and indeed performed as well as the individual experts ( R(2) = 0.90 to 0.96, SD = 1.05 to 1.53 cm). The same performance was observed for peak-velocities; ( R(2) = 0.98, SD = 3.07 cm/s) and ( R(2) = 0.93 to 0.98, SD = 2.96 to 5.18 cm/s). This automated technology allows > 10 times as many beats to be analyzed compared to the conventional manual approach. This would make clinical and research protocols more precise for the same operator effort.