Do not understand how native Magnification is calculated

[elizaan]

https://github.com/girder/large_image/blob/724dc6781a954c9bb7586be4ad4228addfd7b2b7/sources/tiff/large_image_source_tiff/__init__.py#L590C5-L605C10

1. I do not understand these lines:
   pixelInfo = self._tiffDirectories[-1].pixelInfo
   mm_x = pixelInfo.get('mm_x')
   mm_y = pixelInfo.get('mm_y')

    what is pixelInfo, how they are calculated and how they are mapped to mm_x, mm_y.
   

2. Is mm_x, mm_y means length of one pixel?

[manthey]

Yes, mm_x and mm_y are the size in millimeters of one pixel. Each file format specifies resolution in different ways. The metadata from a tile source tries to harmonize this. For example:

source = large_image.open(<some file>)
pixel_size = source.metadata['mm_x'], source.metadata['mm_y']   # these could be None if unknown

In hindsight, I wish we had used either meters (which would have made geospatial users happier) or microns rather than millimeters.

Many whole slide images also have a magnification, where 20 is often 0.5 microns, but not necessarily exactly. It should be the magnification of the objective lens used in data collection, but this ignores the effects of the eye piece and possible the scale of the data sensor.

[elizaan]

so, in the case of magnification calculation, I see something like the
below:
mag =. 0.01/ mm_x (I am referring to this: (https://github.com/girder/large_image/blob/724dc6781a954c9bb7586be4ad4228addfd7b2b7/sources/tiff/large_image_source_tiff/__init__.py#L590)

I do not understand why 0.01 is divided by mm_x; it calculates how many pixels are in 1 mm, but how is it meaningful to the traditional magnification process in the whole slide image scanning? my question is why 0.01 is divided by mm_x

[manthey]

Many algorithms are said to work at a specific magnification (e.g., 20x). Ideally, this would be recorded from the optics of the microscope, but in reality the algorithms are really tuned for a specific pixel size (e.g., 0.5 μm/pixel or 0.0005 mm/pixel). This equation is just the approximation of magnification based on pixel size (that is, 10x is 1 μm/pixel, 20x is 0.5 μm/pixel).

[elizaan]

I also do not understand how the initial value for magnification is set, how the max magnification is set,

also item/tiles API should return what magnification, max or initial?
{
bandCount: 3
levels: 12
magnification:?
mm_x:
mm_y:
sizeX: "310804"
sizeY: "278036"
tileHeight: 256
tileWidth: 256
}

and should maginifaction = always 0.1/ mm_x?

[manthey]

From a UI perspective, if we have a magnification, the zoom level will correspond to it. If we don't, the UI just defaults to "20x". If we don't know the resolution of an image, the controls could be more accurate reading something like "full res, half res, quarter res, etc."

[elizaan]

Sorry if my questions were a little bit clumsy. I was actually asking more specific UI setup. For example, the API returns a value of 15x, now when the imageview first renders the image, the slider should show what value? And the button should show the ranges of magnification.

[manthey]

The UI shows a fixed list of magnifications (https://github.com/DigitalSlideArchive/HistomicsUI/blob/master/histomicsui/web_client/panels/ZoomWidget.js#L68) up the the magnification of the source image, plus a power-of-two higher with options to extend it further.