MGH DCE analysis use case diagram

Q1 Progress Report

Here's the summary of the action items and responsible individual in parenthesis, for the remainder of the year (which ends Aug 30):

1. PkModeling:

1.1. Compare the analysis results with the internal code on the internal data - confirm direction of change is consistent, provide the summary of the results (John).

1.2. Finish investigation of the problematic combinations of Ktrans/ve parameters in the GE DRO (Jayashree), investigate the role of BAT (Andrey)

1.3. Clean up the pAIF code and integrate with PkModeling master (John/Andrey, then Jim)

2. Image registration tasks: we agreed implementation of an open source alternative to the proprietary optimizer used in DRAMMS will be very helpful for integrating MGH registration tools with Slicer. Jayashree said Yangming will look into this as part of his QIICR activities. (Yangming)

3. High-level chart for a subset of MGH use case, similar to Iowa (Andrey will take a first stab, then coordinate with Jayashree).

4. MGH status/progress wiki pages to be updated with details (John).

5. DWI processing: Slicer extension for bi-exponential DWI ADC fitting: will be useful both for BWH and MGH collaborations (Andrey).

As general principles, we agreed to aim for no carry-forward in QIICR budget into the next year, acknowledgment of QIICR grant number in relevant publications, and coordination of co-authorship for relevant joint publications.

DCE

Completely automated DCE module with fixed T1 (in tissue and blood), population AIF

Completed. An AIF can be specified, computed through a region of interest, and now also via a population AIF based upon Parker, 2006. We have currently batch processed a clinical trial dataset which has 42 patients with anywhere from 5 to 35 visits (over 540 DCE datasets).

DCE module support for 4D nifti

There is no support currently built into the module, which currently requires NRRD as the input format. SimpleITK and Nibabel are currently used in a preprocessing step to convert from NIfTI to NRRD. This is a fairly simple and quick process.

Automated AIF detection

No progress.

Repository Management

1. Rebased github repository on top of Jim Miller's upstream repository to get all change history into our repository.

Further Topics to Explore

1. BAT detection could be improved. -

• The bolus arrival time detection can be fooled by small spikes in the signal. Using some sort of smoothing could help (Savitzky-Golay filtering is implied but not implemented in the current code base).

2. DCE DRO dataset validation. -

• The results of validating the PkModeling module on the DCE DRO are ok, but could be better. A paper by Morgan outlines a couple of schemes for improving the module's accuracy.

3. Motion correction between frames -

• should this be incorporated into the PkModeling module? We currently use FSL's mcflirt. Yangming Ou (MGH) is investigating a new method for joint registration and estimation for DCE.

4. We use a double-echo pulse sequence and correct for R2*. This is important for our analysis.

• should this be incorporated into the PkModeling module? As part of the current batch processing, the PkModeling module is run on both echo1 and echo2, and then a small Python program performs R2* correction before the PkModeling is run for a third time.

5. Blind AIF

• Input: time series. Output: DoubleArray data structure of Slicer that keeps time-resolved contrast agent concentration recovered by the algorithm. PkModeling already allows to prescribe AIF defined as DoubleArray as input.

6. Non-Tofts models

• this does fit into the extension of the PkModeling module. This work should be accompanied by a discussion of how the architecture of PkModeling could be modified to allow registering new models as libraries, without having to modify the main extension.

7. T1 maps - would like to incorporate the ability to use T1 maps

• PkModeling has a separate filter for converting signal intensity into concentrations, this might be a good place to add this as an extra feature, and possibly also for the AIF concentration conversion.