This is really fascinating to me. I'm amazed they're able to image the cells of the eye with sufficient resolution and speed to achieve this. From the paper, "a

This is really fascinating to me. I'm amazed they're able to image the cells of the eye with sufficient resolution and speed to achieve this. From the paper, "and targeting 10^5 visible-wavelength laser microdoses per second to each cone cell.". If I understand correctly, they first use one type of spectroscopy (AO-OCT) to image the eye and build a map classifying the type of cells, and then use AO-SLO to find the positions of cells in real time. I assume that AO-OCT can't image at a sufficient rate for the second part (or they would just use one type?) so they need to first build this classification map, and then use it to match the position of cells to their type (e g., by overlaying the positions of cells with the classifications and making them line up).