I’m considering integrating Osimis into our ISB Cancer Genomics Cloud webapp. The goal is to enable viewing radiology images that are associated with some cohort of patients that a researcher has generated by some exploration and analysis of the TCGA, TARGET, CCLE and other data sets. We have extracted DICOM metadata from TCGA radiology images into Google BigQuery, and from that can identify the DICOM StudyInstanceUID of studies associated with a given cohort. The expectation is that, for a given cohort, our webapp will present a table of associated studies, and from which a researcher can select a study to view. In particular, we do not need to expose Orthanc Explorer, just an instance of Osimis when a user clicks on a study in the table.
For this purpose, I would like to access a study in the Osimis viewer using the study’s DICOM StudyInstanceUID, e.g. something like:
Please keep the Orthanc project updated: Your integration project for ISB-CGC indeed looks extremely interesting to us, as we fully support any open science initiative on the top of our work.
And Sebastien, we will definitely keep you updated, and acknowledge Orthanc.
Now, while I have your attention: When I drag a series into the Osimis display window, the scale at the bottom shows the progress of instance loading. What is the distinction between red and green? It appears that when a segment turns red, the corresponding instance is loaded. What does it mean when that segment later turns green?
Someone can correct me if I am wrong. When you drag a series to the window it will go read as the series images are loaded, once they are all loaded and in full quality the bar will go green letting you know all images are there.
What I see is that, yes, the segments go red as the corresponding instances are loaded. Then the segments go green individually in some arbitrary order.
I think you are suggesting that going green means full resolution. In fact, I demonstrated Osimis to my group today and we tried to see whether “going green” might meant an improvement in image quality from the initial “red quality”, but we could not see any change in image quality.
Not a big deal, except that when I add Osimis functionality to our documentation I would like to be able to explain the red-green thing!
In the Osimis Web viewer, images are tagged as “red” if they are downloaded in the JPEG format (destructive compression), whereas “green” is used to indicate the use of the PNG format (lossless compression). Furthermore, JPEG images are encoded in 8 bits per color channel, whereas the source DICOM images are generally encoded in 12-to-16 bits, which can only be represented by PNG.
As a consequence, even if the degradation in quality is mostly unnoticeable in “red” images, it is nonetheless affected by compression artifacts and by a smaller dynamic range. These approximations disappear once the image becomes “green”, that are pixel-to-pixel equivalent to the source DICOM image. The use of “red” JPEG images allows to highly reduce the network bandwidth, in order to quickly provide a good-quality approximation of the full-resolution “green” image, hereby greatly enhancing user experience.
From a medical perspective, (tele)radiologists should thus only do a diagnostic/therapeutic use of the images once they are tagged as “green”.
Thanks for the explanation. This means that Orthanc is saving each instance (at least the image part) as both a JPEG and a PNG? I think the original DICOM file has only a lossless image. Is that correct?
A couple of other questions regarding Osimis:
-Can you explain what the “Presets” are doing? I realize that there is some kind of filtering of intensity values going on, but have figured out quite what’s happening.
-The pixel probe tool seems to report intensity values of monochrome instances in “SP” and “MO” units. What are those?
-Finally, I note that the viewer is completely anonymous…there is no indication of Orthanc or Osimis. Is this by design? Nevertheless, we want to give credit to you folks. When we open an instance of Osimis. How would you like us to do that?
In working with Osiris, I noticed that when an elliptical ROI is drawn on an RGB color instance, no mean or standard deviation values are displayed. However, when a rectangular ROI us drawn on the same RGB instance, a single mean value and a single SD value are displayed. On a monochrome instance, each ROI shows single values for mean and SD.
Would you kindly respond to some questions which I asked earlier?
-Can you explain what the “Presets” are doing? I realize that there is some kind of filtering of intensity values going on, but have not figured out quite what’s happening.
-The pixel probe tool seems to report intensity values of monochrome instances in “SP” and “MO” units. What are those?
-Finally, I note that the viewer is completely anonymous…there is no indication of Orthanc or Osimis. Is this by design? Nevertheless, we want to give credit to you folks when we open an instance of Osimis. How would you like us to do that?
This means that Orthanc is saving each instance (at least the image part) as both a JPEG and a PNG? I think the original DICOM file has only a lossless image. Is that correct?
No, Orthanc only saves the source DICOM file, together with a summary of its tags formatted as a JSON file. In other words, Orthanc saves no JPEG or PNG version of the image. You can find more technical information in the Orthanc Book:
Note however that the Orthanc Web viewer does manage a cache on the filesystem that contains JPEG/PNG versions of the DICOM images, to speed up teleradiology applications.
