Improvement of automated analysis of coronary Doppler echocardiograms

Abstract

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.

Figure 1

A diagram of the conceptual blocks and logic flow of the new Python version of the software.

Figure 2

Images representing the steps taken by the Python algorithm for Doppler video analysis. Panel (A) displays a binarized image of the Doppler region, panel (B) shows the generated output image with critical values labeled with colored points, and panel (C) is an example of a vessel diameter measured during Color Mode analysis.

Figure 3

Figure showing the change in average peak velocity values from MATLAB to Python program analysis of each baseline Doppler video. Videos affected by top noise are indicated with red points, accurate analysis is indicated with green points, videos with fainter peaks indicated with yellow, and inaccurate ECG peak identification shown with blue points. The average change in values excluding those top noise videos is represented by the gray line.

Figure 4

Images displaying the removal of top noise from analysis in the updated Python program. Panel (A) displays analysis of representative baseline (above) and hyperemic (below) videos where top noise was included in the Doppler envelope. Panel (B) shows the same videos processed by the Python program, where top noise has been discarded from the analyzed pattern. In these images, green points indicate the beginning of the diastolic phase, yellow indicates peak velocity, pink indicates decay velocity, red indicates peak diastolic deceleration, and blue points indicate the end of the cycle.

Figure 5

Example of corrected QRS-complex peak identification in the ECG region where the original analysis skipped several peaks. The MATLAB program (above) identifies 9 peaks, indicated by red circles, while the Python program (below) identifies all 15 peaks, indicated by white vertical bars.

Figure 6

Example of corrected QRS-complex peak identification in the ECG region where the original analysis added several incorrect peaks. The MATLAB program (above) identifies 7 additional peaks, indicated by red circles, while the Python program (below) identifies only the correct 15 peaks, indicated by white vertical bars.

Figure 7

Figures displaying partially and completely captured fainter cycles in the Doppler region. Panel (A) above shows cycles 2 and 4 are not fully captured by the MATLAB program, but they are captured and analyzed by the Python program in Panel (B) below. As in previous images, green points indicate the beginning of the diastolic phase, yellow indicates peak velocity, pink indicates decay velocity, red indicates peak diastolic deceleration, and blue points indicate the end of the cycle.

Figure 8

Image generated from the Python program showing unrepresentative cycles in the Doppler region. Cycles 5, 6, 11, and 12 are significantly lower than the surrounding peaks and comparison of these peak values to the average peak velocity leads to rejection from the final dataset after analysis is complete.