Enhancing flow-mediated dilation analysis by optimizing an open-source software with automated edge detection.

Flow-mediated dilation (FMD) is a common measure of endothelial function and an indicator of vascular health. Automated software methods exist to improve the speed and accuracy of FMD analysis. Compared with commercial software, open-source software offers similar capabilities at a much lower cost while allowing for increased customization specific to users' needs. We introduced modifications to an existing open-source software, to better meet FMD analysis needs. The purpose of this study was to compare the repeatability and reliability of the modified software to the original software and to manual measurements. To assess these outcomes, duplex ultrasound imaging data from the popliteal artery in older adults were analyzed. The average percent FMD for the modified software was 6.98 ± 3.68% and 7.27 ± 3.81% for and respectively, compared with 9.17 ± 4.91% and 10.70 ± 4.47% with manual measurements and 5.07 ± 31.79% with the original software for . The modified software and manual methods demonstrated higher intraobserver intraclass correlation coefficients (ICCs) for repeated measures for baseline diameter, peak diameter, and percent FMD compared with the original software. For percent FMD, the interobserver ICC was 0.593 for manual measurements and 0.723 for the modified software. With the modified method, an average of 97.7 ± 2.4% of FMD videos frames were read, compared with only 17.9 ± 15.0% frames read with the original method when analyzed by the same observer. Overall, this work further establishes open-source software as a robust and viable tool for FMD analysis and demonstrates improved reliability compared with the original software. This study improves edge detection capabilities and implements noise reduction strategies to optimize an existing open-source software's suitability for flow-mediated dilation (FMD) analysis. The modified software improves the precision and reliability of FMD analysis compared with the original software algorithm. We demonstrate that this modified open-source software is a robust tool for FMD analysis.