How is it prepared? A thin layer of rabbit retina is sliced using a micro tome. The biological sample is then pigmented and prepared to be scanned by an microscope. The image is scanned under a 5000x magnification. One pixel in the digital image represents 2.18 nm at this magnification level. Images in the experiment are scanned using an electron microscope. At this magnification the neuron cells are visible, some organelles could also be seen. Part of the down sampled image is shown below.
The synapse: The synapse is a region where the neurons talk to each other. These are distinctly visible as dark and kind of blobby regions along the cell membranes. These are 3D structures. The shape of the synapse seen in the image depends on the position (Orientation & location) of the synaptic region with respect to the slice. The size of The synapses may vary from 100 nm - 1 micron. Thus in the ori A sample synapses image is shown below.
Thus in raw pixels a synapse may span from 45.87 (100 nm/ 2.18 nm) to 458.7 (1000 nm/ 2.18 nm) pixels in the original image.
How is the image used in the experiments? The aim of the experiments is to identify the synapses in these images. The image that I am currently working on is Layer1_0_0_card.tif. This image is 16720 x 16750 in size. It is a 8-bit gray scale uncompressed image. For the learning I am using a down sampled image. The size of the image (Layer1_0_0_card_resize_p25.tif) is 4180 x 4188. I am doing this to decrease the number of learning learning points. This makes the algorithm run at least 16 times faster.
Posts
No comments:
Post a Comment