Fig. 4: pH analysis using k-means clustering for in vivo wound bed image segmentation.

The wound bed pH clusters are tracked over a 4-day experiment. a This figure illustrates the sequence for visualizing the segmented wound. The first step is to load the raw fluorescent image, where the pixel intensity in the red color channel is directly proportional to pH. To this image, a brightness threshold is applied to eliminate the dark background. Next, the clustering features for classification are assembled, including the red, green, and blue color channels, along with the x and y pixel positions. These features are input to the k-means clustering algorithm with k = 5. Cluster 0 contains the background pixels and is excluded from the remainder of the figure. Finally, each cluster is colored based on its mean pH. b These plots demonstrate the dynamics of the wound clusters. The pH cluster means and number of elements per cluster, or how many pixels are in each cluster, are shown as a function of the experimental day. c This figure displays the clustered wound beds based on the sequence depicted in (a). The three wounds are shown over the 4-day experiment, segmented into four clusters. The color bars indicate pH and are shown individually to highlight variations within the wound bed. d These plots show the magnitude of pH change throughout the experiment for each cluster. The subplots visualize the magnitude of pH change between consecutive days for each cluster. The information is also represented as a boxplot, where the x-axis is the cluster index and the y-axis is the magnitude of pH change. The variation and mean in the magnitude of pH change are higher in cluster four, which is generally at the wound center.