In the spirit of reproducibility, our group decided to get data from the local supermarket. For this post we will focus on the anchovies we obtained. The anchovies are the dehydrated ones found in packets, and they are usually added as flavoring to soups and other dishes.

The first sample of the anchovy we examined using the foldscope was a cross section of the eye that contains the crystalline lens and aqueous humor. Here we observe the loss of structure of the sample, especially of the lens which can be seen as the dark spots near the top of the image, and the anchovy appears to have become discolored due to the dehydration and preservation process.

The inclusion of surrounding anchovy tissue in the sample makes it difficult to determine exactly which part of the eye is being shown, but the lens and aqueous humor are visible. The images were then post processed by adding a scale bar (100 um). This was done by imaging 1mm length lines using the foldscope and calibrating the images we took of the samples to the images with the 1mm lines. We then added the scale bars to the corner of the foldscope image using a matlab script. Contrast was additionally enhanced through post processing using a neural network. This was done to evaluate what we believe to be a bad example of post processing. As can be seen in the second contrast enhanced image, the style transfer based post processing may suffice as an art piece, but it does not add any meaningful improvements to the image.

Below is the contrast enhanced image for the fish eye sample:

We then repeated this process for a sample taken from the anchovy gill flap. The image was acquired under the same conditions and with the same methodology for generating the scale bar. We also applied contrast enhancement through a style transfer neural network to provide an example of bad post processing.

Below is the gill flap sample with scale bars:

Below is the style transfer network post processed sample: