circadian-lesion

Necessary software:

• Fiji https://imagej.net/Fiji + FFT bandpass filter https://imagej.nih.gov/ij/plugins/fft-filter.html
• OpenCV-3.1.0 and OpenCV_contrib-3.1.0: https://github.com/NIB-SI/circadian-lesion/blob/main/Installation_OpenCV3.1.0

Algorithms

Preprocessing with Fiji

bandpass adjustment of uneven illumination (all pictures are much brighter in the middle part)

• Fiji macro for FFT bandpass filter: filter_large=40, filter_small=3, suppress=None, tolerance=5, autoscale, saturate
• example:

Running .cpp files

• create test-folder at your HOME directory and copy respective .cpp file

• rename .cpp file e. g. Rywal 7-18 dpi each 6 hours (point zero 2017_09_01 at 07.00.00).cpp to Rywal7-18.cpp (remove spaces)

• create folders results/filtered_lesions within the test-folder

• copy input images to test-folder

• create CMakeLists.txt

• run cmake .

• run make

• ./Rywal7-18 'input_folder'

• example: https://github.com/NIB-SI/circadian-lesion/tree/main/example_leaves

Output

Terminal:

1. row file number

following rows: timepoint | Time after inoculation [min] | lesion area red marked timepoints indicate night images

paramters of last image (possibility to insert these paramters to a .cpp file with other settings e. g. Part1/Part2/Part3 to receive optimal results, it's also possible to insert artificial lesion parameters here) -lesion indices -lesion areas -mass points -radii -centers

Documentation of algorithms/methods

int main()

extraction of experiment information

• folder name

• file names

• extraction of timepoints + calculation of time after inoculation

  vector measure_brightness(): measurement of brightness

  • to discriminate day and night images

  Mat normalization() and Mat adjusted_brightness(): adjustment of brightness

  • to create even distribution of intensities

  Mat findLesions1() and Mat findLesions2(): find potential lesions

  • convert RGB images to HSV
  • extract range of hue values (different range between day and night images)
  • apply morphological operations (dilation, erosion) to remove noise

find contours

• apply contours (vector of points) to potential lesions
• exclude contours near to the image border
• set contour area size cut-off
• extract geometrical features for all contours: mass center, minimum enclosing circle, radius

processing of contours (to extract real lesions)

• compare all candidates with each other via area of intersecting circles (minimum enclosing circles) --> possibility to insert manual determined lesions (necessary parameters: mass point (x, y), lesions area)
• if candidate is inside of another, add area to bigger one
• calculate all distances between mass points to all given lesions from the previous picture
• find most suitable candidate via smallest distance: all lesions from the previous picture should have a follower evaluation of new candidates:
• if distance is too near to proofed lesions: add to proofed lesion
• if area size is too small: just ignore divide same candidates:
• for two or more lesion from the previous picture only one candidate was found (lesions grow together)
• adjust area size with the help of previous picture areas calculate growth of lesion areas at night pictures
• set area size of lesions from first night picture to areas of last day image
• add the difference between areas of night images to calculated one --> save parameters of last evaluated lesion (mass point (x, y), lesions area): → manual adjustment possible to continue with better results draw proofed lesions and areas

save results

• images with lesion numbers and areas
• timepoints, time after inoculation, lesion numbers and areas to text file