Build 461

 1. Continued work on UNet3D for pancreas: replace the binary sigmoid DICE loss by the 2-class softmax DICE loss. That worked very well with binary threshold issue gone away.

2. First variant of "genetic shape crossover": augmentation technique of generating new pancreas shapes. Gained +2% DICE on NIH dataset.

3. Implemented Measure>ROI mass center distance.

4. Fixed N4 failure problem - including VS 2013 redistribution into FireVoxel now.

5. Pancreas AI work.

6. Very efficient separable convolution. Updated VolumeSmooth dialog to include "Separablem option". N3 was updated and old version eliminated.

7. Compilation with extra Security flags.

Build 460 is released.

 1. Defect corrections.

2. UNet framework for pancreas and supporting functions.

3. CerebrA brain atlas integration.

 Build 459A is released.

1. Modified layer name construction so no path is present, just the title.

2. Measure>Calculate TV:  now works for all enabled layers.

Build 459 is released.

 1. Update to the "DICOM folder -> to a folder of NII files"  GUI function.

2. Update to the "DICOM folder -> to a folder of FVX files"  GUI function.

3. Fixed a memory leak in the LoadDicomRecord function.

4.  Fixed the FireVoxel.key generation (multiple spaces problem)

5. Updated ChangeVolume Resolution  function.

6. Fixed defect in DICOM>Convert DICOM folder to NIFTI.

7. Fixed error in PFRAME VA_GetNonVoidMask( PFRAME V )

8. Experimental implementation of Genetic Algorithm (GA) for optimizing the GOMR  maps using TV-measure. Gained 2% on top of Laplacian smoothing.

9. Implemented Measure>Calculate Volume TV-2D

10. Implemented Dynamic Modelling > Output N-tuple parametric fille.

11. Removed Volume>Smooth across dynamic dimension. This was superseded by more advanced smoothing functions.

12. Fixed the DICOM folder > to FVX folder function.

13. Enabled loading NIFTI files when both Q- and S-forms are 0.

14. Workflows>Brain MRI   :  EdgeWave Express:    collapsed 3 items {Axial,Sagittal,Coronal} into a single one with the automatic orientation check.

Build 458 is released.

 1. When parsing the DICOM tree: we determine if the files were deidentified and in that case we add the PatientID to distinguish between different patient's records.

2. Corrected defect:  New layer addition and paintbrush usage now set the DocumentModified flag. So there is a warning when the User tries to close the document.

3. FLEXR3D processing dialog: added an option to save FFF (flow field file).

4. MainMenu>Tranform:  added an option to transform using FFF file.

5.  FLEXR3D\FLEXRMC added warning when SSC measure is selected to ensure the feature keypoints are enabled.

6.  Added UI Function>Test :  Combine Probability Maps to Atlas.

7.   Added UI Function>Test :  Sharpen template using Atlas

8. FLEXR cross-correlation similarity measure:  on large appertures (~16vox) obtained a 20% speed up by re-arranging the memory access during parallel execution.

9.  FLEXR SSC measure calculation : removed a major memory leak.

10. Further optimization of Crosscorrelation measure (FLEXR):  using the x86 superscalar instructions and 32-bit floating precision, achieve 32 threads X 16 superscalar operations = x512  parallelization.

11. Removed NGF (Normalize Gradient Field) similarity measure from FLEXR3D\FLEXRMC dialog options.

12. Enable the SSC similarity measure in FLEXR-MC dialog box.

13. Model 17 (IVIM segmented fit):  substantial update with improved parameter ranges etc.

14. Model 18 (IVIM biexp fit): provided update for the ranges of fitted parameters.

15. Parametric modelling dialog: when model changes - the Output parameters selection gets all unchecked to avoid confusion.

16. Corrected behavior when the new layer is added and it has diferent spatial attributes.  Now it ensures that everything matches to the base layer (resolution, slice position, orientation, etc).

17. Merged PWR.DLL and VolumeAnalyze.dll (VolumeAnalyze is removed from distribution now). This is done for linkage efficiency and removing redundancies.

Build 457 is released.

 1. Open DICOM folder:  fixed crash when the folder is empty. Provided diagnostic.

2. Fixed crash when multi-frame file is invalid. Provided diagnostic.

3. Corrected defect:  when the ROI or other layer with different resolution from base is added, now ALL spatial information is properly copied to the new layer.

4. Fixed hard to find defect:  when function FindLayerContainingName was called, the whole name of the layer including PATH was checked. Corrected: only the name title is checked, because the path contained the searched name (like Atlas)

5. Fixed defect when loading zip-compressed headers embedded in FVX files.

Build 456 is released.

0. DICOM load, STUDY construction was corrected.

 1. Defect fixed: Parametric map calculation is skipped for a voxel that has a VOIDVOX at some point.

2.  Defect fixed: exclude VOIDVOX from GOMR calculation

 Build 455A is released.

1.  Model #25 (T2-biexponential fit) is corrected to display a single curve fit correctly.

2. ABT is added for this model

 Build 455 is released.

1. Seamless integration of Multi-frame DICOM support. MF-file is automatically split into a single frame files in the temporary folder and the DICOM tree works on single-frame files.

 Build 454A is released.

1. Corrections for optional loading of FvxANN.dll

Build 454 is released.

 

 1. FLEXR 3D\MC : introduced the Regularization 32/64-bit precision option.

2. Fixed defect in optional loading of FvxANN.dll

Build 453 is released.

1. FLEXR3D:  Implemented the SSC (Self Similarity Context) registration measure.

2. Performed corrections in "Normalize by ROI" dialog box.

Build 452 is released

 1. Internal coding: Parallel Batch unified calculation

2. Model #17:  allowed volumes with two  large b-values for calculation (before it was 3).

3. Cylinder Paintbrush option.

Build 451 is released.

 1. Defect: corrected MST (minimum spanning tree) function in coordinate calculation VA_CoordFromOffs.

2. MST function speed optimization.

3. Fixed defect:   AVX512 instruction detection was always returning TRUE.

4. Parallel Batch functions (x4) consolidation.

Build 450 is released.

 1. DICOM Smoothing batch tool: performed correction for site#3 by folding STUDY UID.

2. Temporal Shape Means smoothing - implemented 2D option for a major acceleration.

3. Internal : Add Layer using Header Volume: fixed the crash defect when there was a mixture of 3D and 4D layers.

4. Neck Multimodality workflow: Updated the parameters for CT->GRASP FLEXR registration.

Build 449A is released.

 1. Dynamic Model #10: rebuilt from scratch to use modern framework.

2. Duplicate Model 11 is removed from the interface.

Build 449 is released.

 1. >ROI>Threshold To ROI  GUI function: parallelized and added a progress bar. It was slow on really big REAL images (300MB+).

2. Increased the REAL Image compression threshold: this is so the images are only compressed when it is worth more. This really speeds up the viewing.

3. Corrected the ParalleBatch {Length,NumberOfBatches} calculation function.

4. FLEXR3D\FLEXRMC dialogs : added an option to Inflate target ROI (similar to AutoFocus3D)

5. FELXR: Minimum Spanning Tree : added the progress bar.

6. Cross Correlation function : upgraded from AVX2 to AVX512

7.  Implemented Hip registration workflow.

8. Fixed defect: Layer Control Dialog: SaveAll (fixed the orientation issue when saving to NIFTI volumes)

9. Fixed defects in API:VolSetSlice function.

10. Fixed defects in Temporal Shape Means function

11. DICOM Folders : Conventional Smoothing to Target DICOM Folder.

12. DICOM Folders : TSM Smoothing to Target DICOM Folder.

13. Remove Multipass from FLEXR dialogs.

Build 448A is released.

 1. GUI: Implemented the Volume>"Multilayer Document PCA" function.

2. "Batch Fix DICOM tags...": provided a better diagnostics message.

Build 448 is released.

 1. FLEXR: Introduce vicinity rotation invariance for similarity measures. Applied it to the X-correlation measure.

2. Brain Atlas GUI: implemented menu items for Stage1 and 2SW-Stage2 processing.

3. FLEXR: Removed OptiMap measure from Menu choices.

4. Re-implemented important IsRoiEmpty function so it works very fast and without decompression.

5. Removed all the items related to Ultra Sound processing.

6. Incorporated the IsRoiSliceEmpty fast check into zooming pipeline. It was x6 time faster for Mindboggle atlas with 70 ROIs.  Also incroporated into FilmView for a similar speedup.

7. Regular Grid keypoint generation:  corrected the adjacency checking to comply with keypoint positision in voxel corner (not vertex).

8. Complete rework of Roi Surface keypoints. Vertice correction and speed up. 

9. Autofocus3D:CalculateMeasure: removed sub-code for SignalDifference using RL-encoding.

10. Corrected FLEXR ROI padding calculation to include all measures.

11. Removed FLEXR legacy algorithm

12. Corrected crash defect : in Affine Reslicing when the resulting volume exceeds 4GB.

13. FLEXR\Similarity: adjusted vicinity calculation in case single slice (2D) image.

Build 447 is released.

 1. Re-implemented the dense Flow field generation, so it works correctly when the initial transform is general affine (as opposed to scale+translation)

2. FLEXR : implemented the VOID voxel concept in Target and Source images. Modified all registration metrics to work properly with VOID voxels. This is especially important for Atlas representation.

3. Atlas based registration : added the workflow item.  Atlas is automatically split into the regions that appear as the individual layers.

4. Made corrections in Neck Multimodality workflow - Step 15.

Build 446B is released.

 1.  Neck Workflow :  DWI Body segmentation automatically advances to starting timepoint.

2. FLEXR:  Local Mutual Info - progress bar added and memory usage substantially decreased.

3. FLEXR : prevent running when no keypoints were specified.

Build 446A is released.

1. Corrections FLEXR when registering 4D volume to 3D.

2. FLEXR-MC dialog simplifications.

Build 446 is released.

 1. Minimum Spanning Tree algorithm update with keypoint re-arrangement in the order of the Tree width-first traverse.

2. FLEXR: developed a multipass version of this algorithm running on 1/3 of RAM.  Multipass allows to start minimum spanning tree from several distant positions to compensate for tree approximation.

3. NW flow field generation:  provided an AVX256 implementation for x2 speedup.

4. NW flow Sparse-to-Dense:  implemented algorithm for filtering out remote keypoints for speedup. This was based on evaluating Y-Z exponents and introdducing some minor artefacts.

5. Volume-to-concentration conversion: fixed defect when VOID voxels were treated as regular ones calculating an infinite concentration as a result.

6. Completely re-engineered sparse-to-dense flow field generation for upto x20 speed up (on CT Lung 5 mm grid)

7. Workflow>Breaset MR BPE : implemented Whole Breast Segmentation function

8. File>DICOM>Dicom Tree to Folders   : fixed defect when composite SERIES were not able to convert to folder.

9. VA_VolSegHist:  fixed defect. This function was returning thresholds measured in Histogram bins. Corrected to signal.

10.  Workflow>BPE breast :  added a one-click function for FGT segmentation.

11.  Workflow>BPE breast: added function for 3D motion correction.

12, MainMenu>Dynamic Analysis:   enabled work without supplying the T1-map so only T1W conversion is enabled.

13. Simplified the FLEXR 3D/MC menues.

Build 445 is released.

1. VA_RAW_VolSaveFile function: optimized when CurrTinx!=1 only the required timepoint is decompressed.

2. New GUI function: Convert folder of FVX files into a target folder with separate subfolder for each FVX file. For each FVX subfolder save all layers to NII files.

3. Provided FLEXR RAM estimation function and comparison to the system resources.

4. Updated MR Lung segmentation algorithm

Build 444 is released.

 1. FLEXR fixed defect so the registered volumes have the same Image Position\Orientation as the Target volume.

2. Updated the Machine GUID scrambling algorithm.

3. Neck multimodality workflow:  added the DWI Body Segmentation and DWI->GRASP registration.

4. Improved the FLEXR masking algorithm, both Smask and Tmask.

5. FLEXR perform corrections is Target->Source output option.

6. Fixed defect in bool VA_VolOverRoi_AvgMinMax, that was affecting the Optimal Intensity Window.

7. FLEXR3D works now when the Source volume is 4D, so it would reslice all timepoints and produce a 4D layer.

8. Fixed defect in DisplayVoxelInfo (status line), so the VOID voxel location show the dynamic index too.

9. Provided a first version of Neck Multimodality 15 step workflow combining PET, CT, GRASp,DWI.

10. Compare documents\compare volumes : provided the comparison function which works on compressed image data, so there is no need to decompress for the comparison.

11. Huffman compression of DICOM headers: upgraded to 64-bit and corrected when the longest H-code exceeded 24-bit.

12. Updated the Machine GUID scrambling scheme.

Build 443 is released.

 1. EdgeWave UI dialog box: fixed defect when Absolute Core SI was not accepting negative values (f.e. Real CT).

2. FLEXR3D dialog UI: added the Initial Transform option

3. FLEXR:InitialAlignment - substantial algorithm re-engineering.

4. Neck Multimodlaity Workflow: Added GUI function for the initial Ct to GRASP alignment.

5. FLEXR3D: implemented the FLEXR with Initial Transform file (vtf).

6. Expanded the FVX License key file to  include Windows Machine GUID.

7. Corrected the WC\WL setting in Volume View Filter dialog.

 Build 442 is released.

1. Developed a new tool for checking precision of motion correction "Motion Profile Diagram". This is based on Gene Kim's 2007 JMRI paper.

2. Extended the logic of loading DICOM 4D Volumes based on Series Description tag.

 Build 441 is released.

1. Volume Layer: ViewFilter dialog box (INT,REAL) : introduced new "Adaptive Window Level\Width option". It optimizes mapping to screen for every zoom. Very  useful for comparing registration results on dark timepoints such as last b-values.

2. LayerControl dialog : implemented  "Interlaced View" option {X,Y,XY} so the layers are seen in overlacing manner.

3. Volume>Split 4D volume into layers" : for 4D volumes splits each separate timepoint into a document layer. Useful for viewing motion correction results.

4. Upgraded to FVX file format v17 due to above (#1,2) changes.

Build 440 is released.

 1. Internal: upgraded VolumePad function to work with 4D volumes.

2. Implemented prototype of NLM-D denoising (when TACs are compared for each location), as opposed to perform NLM for each timepoint separately.

3. When saving volumes with Void orientation to NIFTI, display a warning and save in default orientation.

4. Implemented new denoising algoritm: Temporal Shape Means (with GUI)

5. Implemented new denoising algoritm: Spatial Temporal Shape Means (with GUI).

6. Dynamic framework can now output the Model volume for every voxel (modified GUI by adding a check to the dialog box).

7. Numpy load COMPLEX64 support was added.

8. Fixed FLEXR3D crash on single slice volumes (regular grid generation).

9. Fixed crash in Gaussian smoothing function applied to ROI.

10. BPE map and report generation: changed to average Slope (from Start-End slope). Allowed function to work on any number of timepoints.

11, Added support for additional variety of SIEMENS b-value encoding in SEQUENCE_NAME tag.

Build 439 is released.

 1. Non-local Means denoising : expanded to optionally process all timepoints.

2. Implemented a GUI dialog for loading {2,3,4} dimensional matrices as volumes (File>Open RAW data).

3. Numpy format support under the File>Open RAW data

Build 438A is released.

1. FLEXR:Regularization :  speedup parallelizing the  Cost1,Cost2 updates.

2. DICOM load timings: updated the logic so the the timing DICOM tags are skipped

if they are present but have empty value. 

3. Build test unit for the FLEXR with Jacobian calculation