Back to Projects List
4D MRI via retrospectiv stacking of 2D Slices
- Gino Gulamhussene (University of Magdeburg)
- Marko Rak (University of Magdeburg)
- Michael Halle (Brigham and Women’s Hospital and Harvard Medical School)
- Michael Kelm (Siemens Healthcare GmbH)
- Anneke Meyer (University of Magdeburg)
- Steve Pieper (Isomics, Inc.)
- Tina Vajsbaher (University of Bremen)
- Christian Hansen (University of Magdeburg)
Creating a program to generate 4D MRI sequences applying the retrospectiv stacking method on 2D MR slices.
The available data is comprised of an time resolved alternating sequence of navigator and data slices and a pure sequence of time resolved navigator slices.
All navigator slices were aquired at the exact same location and the data slices were “scanning” the complete liver in a cyclic manner.
To generate a 4D MR sequence from that the program has to collect all data frames that were aquired during the same breating phase, i.e. not at the same time but at different times during the same breathing phase.
To find these, the navigator slices are utilized. Finding correspondances between the navigator slices of the pure navigator sequence and the once of the alternating sequence means to find similar or same breathing phases.
Thus we find all corresponding data slices giving the 3D liver at the specific breathing phase.
- Creating a program to generat sequences of 3D MR images from described 2D Slices.
Approach and Plan
- Using python
- Handling the Slice Data as one big 3D Data Array
Progress and Next Steps
- got insight in available and relevant DICOM tags for the data at hand (big thanks to Joost for the DICOM Explorer)
- hit a roadblock when data appeared to be faulty
- wrote a python script sorting the data by aquisition time, turns out data is faulty after all (thanks to Joost again)
- figured out how to change the MRI sequence back in Germany
- acquired lots of data
- started program to stack the volumes
progress during project week
- made huge progress on the code front
- data packing for faster processing
- config for proband to remember user clicked ROIs, ref. index etc.
- similarity measure, ROI management, file Management, slice binning
- bug bashing
- found out, that the MRI did chang the acquisition order of the slices which changes some assumtions and has to be accounted for
- finish the program
- annotate 3D samples of the 4D Data for ground truth using 3D Slicer
- actually use the data
Data Sorting scheme
Background and References
[VON SIEBENTHAL, Martin, et al. 4D MR imaging of respiratory organ motion and its variability. Physics in Medicine & Biology, 2007, 52. Jg., Nr. 6, S. 1547.⟧(http://iopscience.iop.org/article/10.1088/0031-9155/52/6/001/meta)