Tested using Python 3.6.8, numpy 1.16.4, scipy 1.2.1, pillow 6.0.0, keras 2.2.4, tensorflow 1.12.0 Tested using 20x EFI Images from Olympus Slide Scanner as outlined in paper

This program uses a two-step method to detect synaptic boutons on histological images The first step, a proposal algorithm, quickly scans through the image and detects potential boutons by looking for groups of pixels that are darker than the surrounding tissue

The second step, a verification algorithm, scans each of these potential boutons and verifies them via a pretrained convolutional neural network.

Command line arguments allow the user to select how the program will search for input images --folder: Intended for a single histological section. Required format of this folder is as follows Folder Image001.tif <-- RGB or Greyscale microscopy image. Should be 345 nm/pixel Image002.tif <-- Image001, Image002,... will be stitched together horizontally Mask.jpg <-- Optional Mask with same resolution as Image.jpg. Boutons will only be found in white areas --directory: Intended for an entire brain with many histological sections. Required format Directory Section_01 Image.jpg Mask.jpg Section_02 Image.jpg Mask.jpg ... --order: Intended for multiple brains, each with many histological sections. If the brains are organized as so Order.txt Brain_01 Section_01 Image.jpg Mask.jpg ... Brain_02 Section_01 Image.jpg Mask.jpg ... ... Then Order.txt should read as so: Brain_01 Brain_02 ...

Command line arguments allow the user to select three different types of output. These output files will be placed in the folder where the Image.jpg file was located --png: Program will output a png file with the same dimensions as the image file, with red 5x5 boutons placed on it --svg: Program will output a svg file with bouton symbols placed on it --csv: Program will output a csv file with the x,y positions of each bouton