A COMPARISON OF THE ROWE AND THE LEE MODELS FOR COMPLEX-VALUED FUNCTIONAL MAGNETIC RESONANCE IMAGING by

A COMPARISON OF THE ROWE AND THE LEE MODELS FOR COMPLEX-VALUED FUNCTIONAL MAGNETIC RESONANCE IMAGING Zhaoyang Teng Marquette University, 2010 In 2005, Rowe published an fMRI magnitude and/or phase activation model that utilized both of the real and imaginary data in each voxel [1]. This model described the magnitude and phase time course as varying linearly with different design matrices where the unknown quantities are estimated voxel-by-voxel. This model follows previous work by Rowe and colleagues [1, 9, 10]. In 2007, Lee as the lead author published a similar model to the Rowe model. Lee (2007) described his model as being equivalent to the Rowe model when the design matrices are the same for the magnitude and phase in addition to having the same contrast matrices [2]. In 2009, Rowe published a letter to the Editor of Magnetic Resonance in Medicine in which he described the Lee (2007) model as elegant and computationally efficient, but Rowe did not completely agree with the properties of the Lee model [3]. In his letter to the Editor, Rowe therefore described four items regarding Lee’s model that need to be clarified in addition to describing its relationship to his model. In Rowe’s letter to the Editor he specified these four points and provided counter examples to show the non-equivalence of these two models. A response to Rowe’s letter to the Editor was published in which Lee agreed with the first of Rowe’s items, but disagreed with the remaining three items [4]. In this paper, I will summarize the Rowe model and the Lee model, and discuss the related points in terms of Rowe’s letter to the Editor and the response letter from Lee.