Why light(white) pixels disappeared?

[Alistair]

I use storescp.exe to import images from CR. All works perfectly except one problem: light (almost white) pixels disappear from images. When we see these images at CR stations, they are OK.

Image (top left part of image): http://s54.radikal.ru/i143/0911/f3/14492431ab7f.jpg

Image info: 
transfer syntax     : LittleEndianExplicit
SOP class           : ComputedRadiographyImageStorage
columns x rows      : 2920 x 2320
bits per sample     : 12
color model         : MONOCHROME2
pixel aspect ratio  : 1.00
maximum pixel value : 3000
minimum pixel value : 199

[Marco Eichelberg]

This is nothing to do with storescp. Essentially it means that your CR station interprets the pixel data differently from the other system (whatever you are using to display the images received with storescp). Most likely an incorrect implementation of the grayscale pixel processing pipeline on one of the two systems.

[Alistair]

Can you advice me how to analyze(debug) this situation? It's very strange, because this CR can export images to another places without any problems. I can import images from another places also succesfully. So, i have problem only with this CR.
Is it any ideas about changing configuration of CR(Agfa) or storescp parameters?

[Marco Eichelberg]

It is very unlikely that the StoreSCP parameters have anything to do with the effect you are perceiving, because StoreSCP will never change anything (except for encoding options DICOM explicitly permits to change), it will just store what it receives over the network. You might want to try the --bit-preserving command line option (which guarantees that StoreSCP does not change even a single bit of the incoming dataset) and/or the --implicit option which forces StoreSCP to only negotiate the "oldest, simplest" encoding out of those offered by the CR.

Run the resulting image through dcmj2pnm, preferrably with the +Wm command line option to set a halfway reasonable window center/width. This is what our toolkit's image processing code would make out of the CR image. If the image looks quite different in the viewer you are using, this is an issue of correctly interpreting the image data.