Automated color Doppler ultrasound analysis of bull reproductive tissues using a machine learning-based image processing algorithm

Color Doppler ultrasound is effective for studying tissue perfusion of various organs, but current analysis methods are subjective and time-consuming. This study aims to develop and validate an algorithm for analyzing color Doppler images of the bull's testis and pampiniform plexus. For the study, we selected 2304 color Doppler images (1152 for both the testicular parenchyma and the pampiniform plexus) that were analyzed by a conventional method (CON Group), by pixel separation and counting using Adobe Fireworks® and ImageJ®, or by an algorithm developed in Python version 3.10 (ALGO Group) that can be set to analyze up to 35 variables simultaneously. The processing speed for the ALGO Group was 270 images/0.14 sec. The coefficients of determination (R²) for the variables analyzed by the conventional method and the algorithm were found to be considerably high (0.84–0.97, p < 0.001 for testicular parenchyma images; 0.97–0.99, p < 0.001 for pampiniform plexus). The high correlations indicate that the algorithm produces results consistent with the conventional method, demonstrating its reliability. The Pearson correlation coefficients between the conventional analyses and the algorithm were significant (0.92–0.98, p < 0.001 for testicular parenchyma images; 0.98–0.99, p < 0.001 for pampiniform plexus). In addition, Bland-Altman concordance analyses showed that most points fell within the 95 % confidence interval for both techniques in the organs evaluated. Given the strong correlations demonstrated, the reduced processing time, and the reliability of the results, it can be concluded that this algorithmic approach can effectively replace conventional methods for assessing vascular function.