I'm a NYC-based radiologist.
Hoping to find someone/people familiar (or willing to become familiar) with the DICOM file standard/format which is used in medical imaging.
DICOM files are great for what they are; however they have 2 enormous down-sides:
- 1) Size: They are often hundreds of megabytes big making them difficult to share
- 2) Usability: They require special viewers to be visualized.
I want to create a web application that allows the end-user to:
1. Create an account
2. Upload a DICOM file (from desktop or CD-ROM, ideally it would be a browser-based HTML5 uploader as this company has done: http://mphrx.com/blog/?p=7)
3. Strip all metadata (no identifiable patient information, none).
4. Convert/compress each series of consecutive .dcm images into lossless MP4 videos of maximum bit-depth (apparently MP4 can handle up to 14-bit depth) and resolution. From what I've gathered, www.FFMPEG.com is the best way to do this. I also wonder if using the H.265/HEVC codec (rather than H.264) is possible to achieve better compression/file-size.
5. Allow the user to "play" these optimized videos (scroll forward/backward, zoom, adjust brightness/contrast, etc.).
6. Allow the user to share the videos (downloadable and email-able)
Here is a brief proof of concept screen-capture video I made myself by using multiple 3rd party converters and iOS video applications to convert DICOM to MP4:
There are multiple desktop software (www.dicomapps.com) and DICOM viewers (www.osirix-viewer.com/) that allow the user to convert/export DICOM files as MP4 videos while adjusting the frame rate, quality, pixel size, etc. in 1-click. However, I have yet to see anyone offer this functionality in a web-based application which is the goal here.
I've reached out to a number of developers and data compression scientists to determine the best way to go about this project. So far the most encouraging answers I've received are:
1. "DICOM is a format for multiple embedded images. They aren't necessarily going to be the same type, though if you are sure they are you can skip this step. What I'd recommend is DICOM to DICOM where the 2nd DICOM data is in a standardized format (like lossless jpg). Then dump the contents keeping track of the sequence in the file names (i.e. Img0001.jpg, Img0002.jpg). You can use the GDCM programs for these operations example (http://manpages.ubuntu.com/manpages/lucid/man1/gdcmraw.1.html). Then the next step is converting to MP4 to achieve higher compression rates (so each slice of a study is not a still frame within a HD video). I would use FFMPEG.org to convert to MP4. That will take a sequence and give you the MP4, I think. Using something like this: "ffmpeg -f image2 -r 30 -i %09d.jpg -vcodec libx264 -profile:v high444 -refs 16 -crf 0 -preset ultrafast a.mp4." One more thing you may have many duplicates and you may want to throw a dedup step in there. Once you get this working smoothly command line creating an app is just the standard command line to app."
2. "I'd suggest to use Java, DCM4CHEE for parsing/converting images and FFMPEG as initially suggested by you for MP4."
I've spent months researching DICOM and MP4 and data conversion, etc, but I don't have the technical/programming knowledge to make it happen on my own (yet) and I need your help.
Please contact me if you think you can achieve the above deliverables for a reasonable cost and allotted time.