Analysis of PCIS smooth muscle contractions
Statistical analysis was conducted using GraphPad Prism 6 (GraphPad Software, San Diego, CA, USA). Responses between different stimulations were compared using the Mann-Whitney U test with a P value <.05 considered statistically significant.
A novel video analysis software was developed internally (available at: https://github.com/celalp/video_parser). This software measures muscle contraction using individual pixel movement per frame (Figure 2). Videos were analyzed using a semi-supervised approach with variables kept consistent between the unstimulated control and stimulated sample. Videos were processed using OpenCV (Open Source Computer Vision Library) and scikit-image Python packages.28,29 Pixels that changed from frame x-1 to frame x were calculated using a Gaussian mixture-based background/foreground segmentation algorithm.30,31 A background subtraction algorithm was chosen for this analysis as videos consisted of one large central object with a mostly static background, meaning that any observed movement was captured as foreground. The selection of the specific algorithm (MOG2 in OpenCV) was influenced by the large variability of exposure levels (e.g., glare) between different slices. A dynamic Gaussian Mixture Method allowed for the same background removal method for all experiments. Measurements were validated and confirmed by an independent assessor. Overall pixel movement per frame of the video was plotted for each stimulation and total Area Under the Curve (AUC) was calculated and compared for each contraction response using GraphPad Prism 6.