FireVoxel Build 166 is released

Download link:

https://drive.google.com/file/d/0B2lI9iEKOqv6eG9nQ0xGazlIdjA/view?usp=sharing

VA grant batch processing dialog, latest version:

Corrected the dialog logic (pasting appropriate dz suggestion) inside the Dialog SpecifyResolution.

DICOM Save:  changed the default compression mode to "Uncompressed".  Enabled load of the result by ImageJ (requires uncompressed).

DICOM File Save:  added tags that were missing in previous versions:
TAG_SERIES_NUMBER
TAG_INSTANCE_NUMBER (former TAG_IMAGE_NUMBER)
TAG_MODALITY

HL-case:  PET-FDG.   Reconstruction artefact due to the Gross motion between timepoints #34 and #36.  It appears that at T#35 the reconstruction has combined the head signal from several grossly distant positions, resulting in blurred image and halo (well visible on left-top corner T#35).
This timepoint is too blurred to be motion corrected but result is reasonable. Everything else is motion corrected quite well.

HL - T#34 (2440 sec)

HL-T#35 (2740 sec)

HL- T#36 (3040 sec)

Comparison of   "IVIM- free fit" on the sample dataset obtained from Henry:  Eric's vs. FVX

Eric's IVIM-free fit.  Residual = 2.33

FireVoxel IVIM-free fit.  Residual=2.006

Comparison of   "IVIM-segmented fit" on the sample dataset obtained from Henry:  Eric's vs. FVX

Eric's IVIM-segmented fit.  M0=Tac[b0].  Residual = 2.3

FireVoxel IVIM-segmented fit.  M0 is fitted.  Residual=2.13

PET-FDG motion correction.  Most challenging case (MB-15) of the last batch.  Before and after motion correction, reference BrainMask from the last timepoint (#38) is overlaid on the maximum displacement timepoint (#27)

Before

After

MainMenu>ROI>Clip Volume by ROI:   completely parallelized the implementation with great performance gain (on long 4D sequences).

FireVoxel Build 165 is released.

Download at:  https://drive.google.com/file/d/0B2lI9iEKOqv6RE1sSno1a19yc00/view?usp=sharing

1. FireVoxel file save - disabled the PHI warning,

2. Fixed crash defect:  4D volume overlaid by 3D-ROI as active layer.  Calling EdgeWave function was producing crash.

3. Defect fix: EdgeWave algorithm on real-valued volumes:  provided correct conversion of units from the dialog box to the actual procedure.

4. Affine transform reslicing: substantially increased the parallelization resulting in about x2 speed improvement on the developmental 6-core platform.

5. SIG_DIFF registration measure:  changed accumulation from L2 to L1 norm.

6. UI - VolumeSmooth function:  Removed erroneous  "ROI required function"

7. VolumeSmooth (Median):  corrected the fixed memory allocation for the sorting, so the smoothing of double-valued volumes does not crash anymore.

8. Fully parallelized the VolumeSmooth procedure so it runs with the 100% load on 12-thread system.

9. Implemented initial variant of the "DICOM File Extractor" dialog box.

Implemented initial variant of the "DICOM File Extractor" dialog box.

Next iteration of CT file extractor dialog:

Initial Prototype of CT DICOM file extractor for VA grant.

Fully parallelized the VolumeSmooth procedure so it runs with the 100% load on 12-thread system.

VolumeSmooth (Median):  corrected the fixed memory allocation for the sorting, so the smoothing of double-valued volumes does not crash anymore.

UI - VolumeSmooth function:  Removed erroneous  "ROI required function"

DICOM Tree functionality:  decoupling the DICOM volume objectification functionality from the DICOM-Folder-Tree dialog box.  This is so that functionality could be used by other DICOM-related dialogs.

Register 4D volume to 3D:  major code re-engineering.

Registration of 3D/4D volume to ROI.  This frequent and useful scenario is greatly sped up by adding the "Target ROI" identical to the binary Target and inflated by 1 voxels. This improvement is only effective for the URAL-measure scenarios.