Automated Vessel Diameter Measurement From Intravital Microscopy
Intravital video microscopy is widely used for observing mammalian vasculature in vivo. In microcirculation research, the effect of various stimuli on vasculature is often studied using intravital microscopy. The vessel diameter is a commonly reported indicator for such an effect and needs to be accurately measured. Image sequences acquired using intravital microscopy span several minutes in length, and the current practice of manual measurement using electronic calipers or image shearing is time-consuming and prone to measurement error. Automation of vessel measurement would provide an alternative that is faster and more reliable. The goal of this work was to develop and evaluate an algorithm for automatically measuring the diameter of a vessel from intravital video microscopy data. The proposed method tracks the vessel diameter throughout the entire image sequence once the diameter is marked in the ﬁrst video frame. Two seed points, indicating the vessel diameter, are placed on the vessel walls and tracked throughout the entire image sequence using feature tracking algorithm. The algorithm parameters were optimized using intravital microscopy image sequences. The ground truth was established manually for each case, and the optimal parameters were found by minimizing deviation from the ground truth. The accuracy of the method was validated using both synthetic and real intravital image sequences. On synthetic dataset, The automated measurements deviated from the ground truth by an average of 0.0 pixels, while the manual measurement had the average mean squared error of 1.74 pixels. When the ac- curacy was further evaluated on ﬂuorescence-confocal and non-confocal transmission microscopy image sequences, it was found that the automated method can measure the diameter accurately based on expert visual assessment. Furthermore, the repeatability of the automated measurements was evaluated based on Bland-Altman analysis and compared to that of the manual measurements. The 95% limits of agreement were found to be [-1.36 µm, 1.52 µm] for the automated method and [-3.08 µm, 2.17 µm] for the manual measurements. The automated method resulted in narrower limits of agreement, indicating that it has a better repeatability than human raters. The presented algorithm performs well in terms of measurement accuracy and reproducibility. The automated vessel measurement, with the validated performance, will be highly useful for many biological studies that require vessel diameter measurements over time.
dissertation or thesis