Experienced that intermittent DICOM crash inside the DCMTK library.
Traced it down to DCMTK logging certain events routinely, even without error encountered, so
I am trying to disable the DCMTK::logger right now...

Apparently, a common problem, see:
 http://forum.dcmtk.org/viewtopic.php?p=12206&sid=ed45aa727c4ec83992bed097b713920f

working on disabling logger completely.

Recompiling the whole DCMTK by redefining

#define OFLOG_TRACE(logger, msg)
#define OFLOG_DEBUG(logger, msg)
#define OFLOG_INFO(logger, msg) 
#define OFLOG_WARN(logger, msg) 
#define OFLOG_ERROR(logger, msg)
#define OFLOG_FATAL(logger, msg) 

Finished the Automatic Build Test for Build 125. 

Fixing defect:  On some volumes, when drawing on the single slice ROI, and the mouse cursor croses the right edge of the slice, the ROI voxels apper to be drawn at the top edge.

In parametric map calculation framework, switched the code from TBB:tasks constructs into parallel_for construct substantially reducing the size and complexity of the code.

Added new functionality: Presenly Parametric Model calculation would NOT initialize on the PET volumes.
Update: Such logic is simply not present for Real-valued volumes (=PET).  Requires a separate implementation. But this is critical for the implementation of MRTM, SMRTM models.

Enabled  ViewTAC dialog (real time viewing of TACs as the cursor moving over) for 4D Parametric Maps and Real-valued volumes (that includes PET).

Fixed defect when Wait Cursor is not display while obtaining TAC from a 4D volume.  For large volumes such as PET this causes a problem as it is unclear if operation is being performed or app just froze.

Working on several defects related to display "hourglass" cursor in some registration scenarios, timee scrolling etc.

Switched BW (1 bit ROI) compression to 16-bit base to gain extra compression and speed while drawing the ROI.

Meeting with Yu-Shin. Installed and ran the FireVoxel on her MacWin configuration on the acquired PET dataset. Discussed priorities for the next week:  PET-to-MPRAGE registration, PET averaging function for specified frames, further TAC support for Excel import, parametric models without Input funciton SRTM, SRTM2, MRTM.
The need for the MNI-Altlas registration was raised again.

Working on the registration of PET volume to MPRAGE volume using the existing registration tools of the FireVoxel.  PET is cropped to keep only relevant voxels. MPRAGE is transformed into the Axial projection.

Attempting to further increase decompression speed for highly compressed volumes, in light of the high speeds required by overlayed ROI drawing. Seamlessly combining the prediction residual decoding with the Huffman decoding.

This provided an additional x1.25 speed up in the decompression of the volume.

Substantially sped up huffman decompression, by replacing the bit-testing loop for all codes under 12-bit long. Instead a direct table is used which covers bulk of the codes, but is still small in size.
This allowed to speed up decompression by x3 times. This is extremely important,  because when user draws an ROI over signal intensity volume, a lag will appear in the ROI drawing if the decompression is not fast enough.

Added a new feature: when 3 point projection View is initiated, the other 2 orthogonal projections start in a Single, middle-slice view.

Fixing defect: Saving dynamic document (f.e. Ktrans) to FireVoxel document format, unable to load the saved 4D volume

Released an unofficial Build 125 with the working K-trans model for prostate dynamic dataset