Click here to


Are you sure ?

Yes, do it No, cancel

Improve IVIM&DCE Map Using Convex Optimization with Higher-Order Total-Variation Regularizations

R He , Y Ding , A Mohamed , H Elhalawani , R Ger , B Elgohari , J Wang , K Brock , C Chung , K Hutcheson , S Frank , S Lai , C Fuller*, UT MD Anderson Cancer Center, Houston, TX


(Sunday, 7/29/2018) 3:00 PM - 6:00 PM

Room: Exhibit Hall

Purpose: Improving MR maps of IVIM and DCE in the head and neck cancer patients is particularly challenging because of noise and involuntary motion. We implement an efficient method towards recovering data form noise while keeping motion effects reduced.

Methods: In acquisition of IVIM and DCE data, either b-value series or time-series will be generated, and the parameter map is estimated by fitting the series at each voxel to the MR signal models respectively. However there are factors such as noise and motion limited the confidence level of the maps. This limitation can be addressed by jointly exploiting the correlation in space and in the b-value/time-course dimension. Towards that, an iterative approach using convex optimization with higher-order total-variation regularizations was adopted to allow IVIM and DCE reconstruction from head and neck data more reliably. It's well known that total variation regularization can help in getting rid of noise and motion, but first order total-variation may lead to staircase which could be annoying in spatial visualization, higher-order total-variation regularizations can overcome staircase while keep the advantage of total-variation. Currently there are no efficient and elegant approaches towards higher-order total-variation regularizations in convex optimization, we implement an algorithm that convex optimization with non-smooth regularizers is fulfilled in a flexible scheme in which any higher-order total-variations can be involved.

Results: We demonstrate that with the combination of first-order (in spatial domain and series direction) and second-order (in spatial domain) total-variations on the original data for filtering, IVIM and DCE parameter maps can be reconstructed with considerable improvement in quality comparing to the conventional approaches.

Conclusion: Higher-order total-variation regularizations on the filtering of IVIM/DCE data is useful in improving the quality of parameter maps.


Not Applicable / None Entered.


Not Applicable / None Entered.

Contact Email