Difference between revisions of "SidongLiu Update"

From NAMIC Wiki
Jump to: navigation, search
Line 101: Line 101:
 
* 23 Jun - 27 Jun (NA-MIC 2014 SUMMER PROJECT WEEK)
 
* 23 Jun - 27 Jun (NA-MIC 2014 SUMMER PROJECT WEEK)
 
** CAD Toolbox for Neurodegenerative Disease Diagnosis (http://www.na-mic.org/Wiki/index.php/2014_Summer_Project_Week:CAD_Toolbox_for_Neurological_Disorders)
 
** CAD Toolbox for Neurodegenerative Disease Diagnosis (http://www.na-mic.org/Wiki/index.php/2014_Summer_Project_Week:CAD_Toolbox_for_Neurological_Disorders)
 +
** Serve as NA-MIC tutorial contest judge
 
** Process the dataset of patient2_1 with DWIConverter
 
** Process the dataset of patient2_1 with DWIConverter
 
** Prepare the DICTA 2014 abstract
 
** Prepare the DICTA 2014 abstract
Line 107: Line 108:
 
** Reprocess the dataset of patient2_1 with DWI Converter
 
** Reprocess the dataset of patient2_1 with DWI Converter
 
** Test 'DICOM-TO-NIFTI-CONVERTER' matlab toolbox (http://www.mathworks.com/matlabcentral/fileexchange/41874-dicom-to-nifti-converter/content/dicom2nifti.m). However, this toolbox is not usable for our SIEMENS data, because it only work with DICOM files from Phillips Achieva 3T R2.6 with  
 
** Test 'DICOM-TO-NIFTI-CONVERTER' matlab toolbox (http://www.mathworks.com/matlabcentral/fileexchange/41874-dicom-to-nifti-converter/content/dicom2nifti.m). However, this toolbox is not usable for our SIEMENS data, because it only work with DICOM files from Phillips Achieva 3T R2.6 with  
% - Slice orientation  : transverse
+
***  Slice orientation  : transverse
% - Patient position  : head first
+
***  Patient position  : head first
% - Patient orientation: supine
+
*** Patient orientation: supine
% - Fold-over direction: AP
+
*** Fold-over direction: AP
% - Fat shift direction: P
+
*** Fat shift direction: P
% - Gradient resolution: high (DTI)
+
*** Gradient resolution: high (DTI)
% - Gradient overplus  : yes (DTI)
+
*** Gradient overplus  : yes (DTI)
% - Sort images        : b=0 volume first (DTI)
+
*** Sort images        : b=0 volume first (DTI)
 
** Make the submission for DICTA 2014
 
** Make the submission for DICTA 2014
  

Revision as of 14:16, 9 July 2014

Home < SidongLiu Update
  • Feb 28
    • MICCAI paper submissions
  • Mar 12 - 14
    • Start to work at SPL, 75 Fransic St
  • Mar 17 - 21
    • EMBC paper submission
    • Implement a new functionality for peritumoral tract exploration
  • Mar 24 - 28
    • MICCAI paper reviews
    • SNMMI abstract papers accepted
    • Reformat the Hausdorff outputs
  • Mar 31 - Apr 4
    • Postdoctoral fellowship application
    • Start to work at SPL, Boylston St
  • April 7-11
    • CMIG journal revised version submission
    • ACM MM abstract preparation
    • Work on DTI Editor module ("tract extractor")
  • April 21-25
    • Made progress to the Mosaic Viewer module. Now it could display and manipulate multiple independent 3D viewers at the same time
    • Discussed DTI Challenge preparation. Need to install Evernote and Live Minutes to enhance collaboration
    • Next step would be to finalize the Mosaic Viewer by creating a scene view for the it
    • Submitted the Grants-in-Aid application
    • The original test dataset is a HARDI dataset with 69 gradients plus 4 baselines, and we extracted 30 volumes whose b-values equal to 1000.
  • April 28 - May 02
    • Finalized the Mosaic Viewer module.
    • Successfully loaded and corrected the Brain Visa surface files into Slicer. Next step would be streamline the process. There might be three potential solutions:
      • Build CLI modules (SlicerToBrainVisa Import and Export) to integrate Brain Visa morphologist tools in Slicer. Check NiftyReg extension as an example (https://github.com/spujol/niftyregExtension) and Hello World tutorial to learn how to integrate .cxx plugins (http://www.slicer.org/slicerWiki/images/2/22/ProgrammingIntoSlicer3.6_SoniaPujol.pdf). Note this tutorial might need to be updated to Slicer 4.
      • Build a Python wrapper which point to the Brain Visa executable files and run Brain Visa in Slicer.
      • Write a simple script to 1) convert nhdr to Nifty; 2) filp left / right by changing RAS to LPI coordinate system; 3) run Morphologist pipeline in Brain Visa; 4) load output .ply file in Slicer; and 5) apply 'BrainVisatoSlicer' transform if the coordinate differences were independent to the datasets.
    • Test the DWIConverter module to convert DICOM to nhdr using the datasets '...high-b.raw.gz' dataset and compare the output to '...high-b.nhdr'.
    • New tasks regarding the test dataset:
      • Create the tractography for extracted dataset only using single tensor algorithm
      • Compare the result to that of original dataset using UKF extension
      • Run the DTIPrep extension on both datasets (original and extracted)
  • May 05 - May 09
    • Reconstruct the tensors using eigen values and eigen vectors. Slicer could not handle eigen vectors directly, so we ned to calculate the 7 coefficients in the tensor matrix.
    • Our multi-channel pattern analysis paper was accepted by CMIG.
  • May 12 - May 16
    • Run DTIPrep for all the datasets.
    • Clean up and classify all the processed datasets.
    • Send email to Francois checking what are the differences VC and QCed results.
    • Send BrainVisa screenshots to Sonia, describing the sulci recognition problems.
    • Run UKF tractography with 1-6 seeds per voxel, 2 tensors, other parameters by default.
  • May 19 - May 23
    • Process the new dataset 'test1'
    • Solve the DTIPrep problems.
      • 1. Why there are 4 baseline images in the original dataset, but only one is left after processing? All the baselines are merged into one image.
      • 2. How to see the sphere after quality check? We need to first save the results, and then load the XML to DTIPrep.
      • 3. Why the gradient directions are changed after processing? This might be a bug, resulted from rounding errors.
      • 4. Why there are holes in the QCed DTI dataset? We need to reset the parameter 'DTI_bCompute - DTI_baselineThreshold'.
  • May 26 - May 30
    • Submit propagation matrix fusion paper to ICARCV 2014.
    • Rerun DTIPrep on 'test1'. After testing with different settings, we found when this parameter is set to less than 50, the holes disappear.
  • Jun 2 - Jun 6
    • Run BrainVisa on T1 and T2 registered data, use label map as a mask to filter out the lesion from the generated surface. Try 'Resample Scalar/Vector/DWI Volume' module, set 'Interpolation Type' to ws.
    • Run 'Slicer Surface Toolbox' to refine BrainVisa surface. Chcek 'splitting' box in Normals.
    • Convert DICOM to FSL (nifty) format.
    • Invited to be NA-MIC tutorial contest judge.
    • Attend Junichi's experiment
  • 09 Jun - 13 Jun
    • Test DWIConvert on Sample Data - DWIVolume
      • NrrdToFSL: Slicer-build/lib/Slicer-4.3/cli-modules/DWIConvert --conversionMode NrrdToFSL --inputVolume ./dwi.nhdr --outputVolume ./dwi_fsl.nii.gz --outputBValues ./bvals.txt --outputBVectors ./bvec.txt
      • FSLToNrrd: Slicer-build/lib/Slicer-4.3/cli-modules/DWIConvert --conversionMode FSLToNrrd --inputVolume ./dwi_fsl.nii.gz --outputVolume ./dwi_reconverted.nrrd --inputBValues ./bvals.txt --inputBVectors ./bvec.txt
    • Test DWIConvert on 'test1' dataset
  • 16 Jun - 20 Jun
    • J1 extension funding application
    • Process the dataset of patient2 with BrainVisa
    • Process the dataset of patient3 with BrainVisa
    • MICCAI 2014 Machine Learning Challenge result was revealed and we ranked 10th in 50 teams.
  • 30 Jun - 04 Jul
    • Reprocess the dataset of patient2_1 with DWI Converter
    • Test 'DICOM-TO-NIFTI-CONVERTER' matlab toolbox (http://www.mathworks.com/matlabcentral/fileexchange/41874-dicom-to-nifti-converter/content/dicom2nifti.m). However, this toolbox is not usable for our SIEMENS data, because it only work with DICOM files from Phillips Achieva 3T R2.6 with
      • Slice orientation : transverse
      • Patient position : head first
      • Patient orientation: supine
      • Fold-over direction: AP
      • Fat shift direction: P
      • Gradient resolution: high (DTI)
      • Gradient overplus : yes (DTI)
      • Sort images : b=0 volume first (DTI)
    • Make the submission for DICTA 2014
  • 07 Jul - 11 Jul
    • DS-2019 extension, turns out that no need to extend it
    • Mosaic Viewer first workable version
    • ICARCV paper accepted