Fig. 2

Magnification-based vs micron-based tile extraction. a Comparison between magnification-based and micron-based tile extraction at 20x effective optical magnification. In this example, slide 1 has internal pyramid images stored at 2.5x, 10x, 20x and 40x magnification. Slide 2 has images stored at 2.5x, 10x, and 40x. Magnification-based strategies extract tiles at a matching layer in the image pyramid for a target optical magnification. In this example, Slide 2 is missing the 20x magnification layer, so tiles could not be extracted at 20x magnification. In comparison, with micron-based tile extraction, image tiles would be extracted at the 40x layer for Slide 2 and resized to an effective optical magnification of 20x. b Comparison between magnification-based and micron-based tile extraction at 10x effective optical magnification. In this scenario, the “10x” layers in Slide 1 and Slide 3 have slightly different effective optical magnifications – 10x and 10.5x – due to slide scanner differences. With magnification-based tile extraction, image tiles extracted from Slide 1 and Slide 3 would have slightly different effective optical magnification. With micron-based tile extraction, tiles would be extracted at the 10.5x magnification layer from Slide 3 and resized to match the same effective optical magnification as Slide 1 (10x). This strategy ensures that all image tiles have the same effective optical magnification