Influenza and Norovirus Incidence and Optical Flow activity video material for 2008-2018

These videos are part of our work on the analysis of Noro and Influenza patterns and propagation in central europe/germany. Incidence intensity and location plots are shown on the left. Optical flow direction and intensity of the registered cases is shown on the right. For the decade of 2008-2018. The second video shows the season 2009-2010.

Excerpt from the material and methods regarding these videos: Incidence plots were generated with a jupyter notebook based script (with Matplotlib 3.1.1 in Python 3.750). From these images stacks were created with Multi Stack Montage Plugin for ImageJ FIJI. Image analysis, optic flow analysis and preparation of figures was done in ImageJ52 with the with the Gaussian MSE plugin in the distribution FIJI. Optic Flow (Parameters: 8 px Neighboorhood, Sigma 4) on histogram stretched XY incidence plots of Noro and Influenza were compiled into an AVI movie and singular frames. This limits the resolution to the one presented here, as each pixel represents data like a heatmap does. Optical flow is a technique to detect moving objects in videos or image sequences (for example a series of microscopic images). In this case it is used to detect movement of infection peaks through a map. In this case we utilize optic flow to detect the direction and intensity of possible shifts of case numbers of Noro and Infuenza based on a intensity plot (done with Python MatplotLib). A plot for every week results in 570 frames that are analyzed as a stack in ImageJ Fiji with the Gaussian Window MSE Optic Flow plugin. This plugin analyzes the stack of images by taking pairs of images and comparing the pixels within a Gaussian kernel. This results in frames that show a black silhouette (the search radii for each detected datapoint) of the german municipalities on a white background (no data). The intensity of the shift is indicated by the brightness of the color traces. The vector is indicated by the color of the trace according to the RGB colorwheel.
Since historical data (0-235 cases per week per municipality) falls within a 0-255 range no scaling was necessary to stay within 8 bit range of the imaging analysis used. Contrast was enhanced (histogram range was stretched up to 255) in the print figures to ensure that everything is visible when printing out on paper. Red no-data areas were replaced by white for print-purposes. No data was dropped in the process, no over-stretching of the histogram occurred. Country boarders were added as overlay after processing to enable easier orientation.

Epidemical Data provided by Robert-Koch Institut Survstat 2019. Coordinate data was downloaded from BKG.De 2019.

Funding by BMBF.