Contrast sensitivity, the ability to distinguish the foreground from the background is the foundation of human pattern vision. For this reason, contrast sensitivity has been considered as a major barometer of human visual function. By capitalizing on deep learning techniques and retinal imaging data, we showed that human contrast sensitivity can be reliably predicted from the retinal structural data. Importantly, the activation maps from the deep network elucidate the exact retinal layers closely linked to the contrast sensitivity, i.e., the ganglion cell and its associated layers. Our findings help to improve our understanding of retinal mechanisms underlying human spatial vision.
Read more here.
