Hassan Bagher-Ebadian1,2, Siamak P. Nejad-Davarani1,3, Ramesh Paudyal1, Tom Mikkelsen4, Quan Jiang1,2, James R. Ewing1,2
1Neurology, Henry Ford Hospital, Detroit, MI, USA; 2Physics, Oakland University, Rochester, MI, USA; 3Biomedical Engineering, University of Michigan, Ann Arbor, MI, USA; 4Neurosurgery, Henry Ford Hospital, Detroit, MI, USA
Estimating the longitudinal relaxation time, T1, from spoiled-gradient-recalled-echo (SPGR) images is challenging and susceptible to the level of noise-to-signal ratio (SNR) in acquisition. Methods such as Simplex-Optimization, Weighted-Non-Linear-Least-Squares, Linear-Least-Square, and Intensity-based-Linear-Least-Square have been employed to estimate T1. In linear and non-linear methods, the estimated T1 values are dependent on defining the weighting factors, which may result in a biased estimation. Herein, an adaptive neural network is trained and compared with different techniques using an analytical model of the SPGR signal in the presence of different levels of SNR. Receiver-Operator-Characteristic analysis and K-fold-cross-validation were employed for validation, testing, and network optimization.