MR Estimation of Longitudinal Relaxation Time (T1) in Spoiled Gradient Echo Using an Adaptive Neural Network

Introduction: The acquisition of high-resolution T1 maps in a clinically feasible time frame has been demonstrated with Driven Equilibrium Single Pulse Observation of T1 (DESPOT1) [1-3]. DESPOT1 derives the longitudinal relaxation time, T1, from two or more spoiled gradient recalled echo (SPGR) images acquired with a constant TR and different flip angles [2, 3]. In general, T1 can be estimated from two or more SPGR images [2, 3]. Estimation of MR parameters (T1, M0, etc.) from these sequences is challenging and susceptible to the level of noise in signal acquisition [2, 3]. Methods such as Simplex Optimization, Weighted Non-Linear Least Squares (WNLS), Linear Least Square (Gupta’s LLS or GLLS), and Intensity based Linear Least Square (ILLS) method have been employed to estimate T1 [2-6]. In both linear and non-linear methods, the estimated T1 values are dependent on defining the weighting factors, which may result in a biased estimate of T1 [3-6]. Herein, an adaptive neural network (ANN) is introduced, trained and evaluated as a T1 estimator from the SPGR signal. The ANN was trained using an analytical model of the SPGR signal in the presence of different levels of signal to noise ratio (SNR). Receiver Operator Characteristic (ROC) analysis with K-fold cross-validation (KFCV) were employed for training, testing, and network optimization.