Yinghua Tao BEng, Presenter: Nothing to Disclose

Howard A. Rowley MD, Abstract Co-Author: Research Consultant, Eli Lilly and Company Research Consultant, W.L. Gore & Associates, Inc Research Consultant, Medpace, Inc Research Consultant, H. Lundbeck A/S Research Consultant, Bayer AG Research Consultant, General Electric Company Speaker, Bracco Group Researcher, Guerbet SA

Kari A. Pulfer, Abstract Co-Author: Nothing to Disclose

Guang-Hong Chen PhD, Abstract Co-Author: Research funded, General Electric Company Research funded, Siemens AG Research funded, Varian Medical Systems, Inc Research funded, Hologic, Inc

Temporal dynamic information is crucial in determining perfusion parameter maps, we investigated the impact of temporal average based HYPR (HighlY Constrained BackPRojection) in reducing radiation dose in CT cerebral perfusion while maintaining temporal information, and ultimately perfusion map accuracy.

Two HYPR schemes were investigated: mask-subtracted and non-subtracted. Temporal averaging was used in both schemes. Five clinical CT cerebral perfusion datasets at full dose (80 KVp, 200 mAs, 2.8 sec ) and one high and low dose canine perfusion dataset (80 KVp, 100 and 25mA, 1 sec )  acquired using GE discovery VCT were studied under an IRB approved protocol. Perfusion map accuracy and noise were compared between original and HYPR images using both approaches.

For clinical full dose datasets, map accuracy was preserved in non-subtracted scheme (error<10%) up to temporal averaging of 15 time frames, but was only preserved up to averaging of 6 frames in mask-subtracted scheme. This agreed with the peak value error of  tissue enhancement curve between HYPR and original images. Using non-subtracted scheme, peak value was underestimated by less than 3% using averaging of 5, 10, and 15 time frames, and noise (standard deviation) reduction was 27.2%±0.05, 34.0%±0.06,  and 35.4%±0.07. Using mask-subtracted scheme, peak value was underestimated by 11.14±3.05, 22.93±10.89, 43.53±19.99, and 62.64±23.72, using averaging of 4, 6, 8, and 10 time frames, and noise reduction was 33.6%±0.06, 38.2%±0.08, 40.7%±0.08, 41.3%±0.07. For canine low dose dataset, both the HYPR non-subtracted scheme with averaging up to 50 time frames and mask-subtracted scheme with averaging up to 6 time frames improved map quality, and the mask-subtracted scheme was comparable to high dose dataset. 

HYPR non-subtracted scheme allowed for more temporal averaging without loss of perfusion map accuracy than mask-subtracted scheme. However, taking both map accuracy and noise reduction into account, both schemes allowed for similar noise reduction factor and preserved map accuracy. In addition, mask-subtracted scheme was better at reducing temporal fluctuations. Low dose data indicated that radiation dose reduction factor of 4 was achievable with maintained perfusion map accuracy.

High quality, low radiation dose CT cerebral perfusion can be obtained using HYPR post-processing schemes.

Tao, Y, Rowley, H, Pulfer, K, Chen, G, Comparison of Two HYPR Schemes in CT Cerebral Perfusion Map Accuracy and Their Impact on Radiation Dose Reduction.  Radiological Society of North America 2011 Scientific Assembly and Annual Meeting, November 26 - December 2, 2011 ,Chicago IL. http://archive.rsna.org/2011/11034613.html