Defense Date


Document Type


Degree Name

Doctor of Philosophy



First Advisor

Alen Docef


The signal resulting from the IR-FSE sequence has been thoroughly analyzed in order to improve the accuracy of quantitative T1 mapping of the human brain. Several optimized post-processing algorithms have been studied and compared in terms of their T1 mapping accuracy. The modified multipoint two-parameter fitting method was found to produce less underestimation compared to the traditional multipoint three-parameter fitting method, and therefore, to result in a smaller T1 estimation error. Two correction methods were proposed to reduce the underestimation problem which is commonly seen in IR-FSE sequences used for measuring T1, especially when a large turbo factor is used. The intra-scan linear regression method corrects the systematic error effectively but the RMSE may still increase due to the increase of uncertainty in sequences with large turbo factors. The weighted fitting model corrects not only the systematic error but also the random error and therefore the aggregate RMSE for T1 mapping can be effectively reduced. A new fitting model that uses only three different TI measurements for T1 estimation was proposed. The performance for the three-point fitting method is as good as that of the multipoint fitting method with correction in the phantom simulation. In addition, a new ordering scheme that implements the three-point fitting method is proposed; it is theoretically able to reduce the total scan time by 1/3 compared to the TESO-IRFSE sequence. The performance of the three-point fitting method on the real human brain is also evaluated, and the T1 mapping results are consistent to with the conventional IR-FSE sequence. More samples of true anatomy are needed to thoroughly evaluate the performance of the proposed techniques when applied to T1 mapping of the human brain.


© The Author

Is Part Of

VCU University Archives

Is Part Of

VCU Theses and Dissertations

Date of Submission

December 2011

Included in

Engineering Commons