Signal Modeling in Functional MRI

Kruggel F.1, Rajapakse J.C.2, Svensen M.3

1 University of California, Irvine, 2 Nanyang Technological University, Singapore, 3 Max-Planck-Institute of Cognitive Neuroscience, Leipzig

Signal description aims at modeling the response shape by parameters and relating these parameter values to variables of the stimulation context. A flexible nonlinear regression context was proposed to model the hemodynamic response (HR) as a sum of a deterministic function and a stochastic part. By adaptation of the model function to the observed HR, we obtain parameters that characterize the HR shape, such as the gain (height of the HR), the lag (time from stimulation onset to maximum) and dispersion (temporal width of the HR). These parameters are compared with the stimulation context in a subsequent regression model.

An item recognition task proves this concept in a complex event-related experiment. Subjects were shown a cue set of 3-6 letters. After a variable delay length (2.0-7.0 s) a probe letter appeared, and the subject was asked to respond via a button press whether the probe letter belonged to the previously presented set. Subjects completed 192 different combinations. Because cue and probe phases were expected to elicit HRs, we used a sum of two Gaussian functions as a model function. The cue phase response was found to depend on the set size only, with a gain increase of up to 15% per set item in activations along the inferior frontal sulcus. The lag of the probe phase response depended obviously on the delay time, but on the set size (+80 ms per item) and on the hit/foil manipulation (+200 ms). Responses in both phases decreased during experimental time by more than 50% in some focal activations, indicating an on-going parallel optimization of the task.

activated_areas modelled_timecourses

Left: Activated brain regions in a fMRI experiment on working memory. Right: A few trials from the spatially averaged time course of the signal in two sample regions. Dotted lines denote the visual stimulation. Note the variable delay time between cue and probe phase. Thick dots correspond to the measurements, and black lines to the estimated waveforms.


Kruggel F., von Cramon D.Y. (2001) Nonlinear Regression Analysis of the Hemodynamic Response in Functional MRI. Pattern Recognition Letters 22, 1247-1252.

Kruggel F., Zysset, S., von Cramon D.Y. (2000) Nonlinear Regression of Functional MRI Data: An Item Recognition Task Study. NeuroImage 12, 173-183.

Kruggel F., von Cramon, D.Y. (1999) Physiologically Oriented Models of the Hemodynamic Response in Functional MRI. In: Kuba A., Samal M., Todd-Pokropek A.T. (eds.), Information Processing in Medical Imaging (Visegrad), Lecture Notes in Computer Science Vol. 1613, pp. 294-307. Springer, Heidelberg.

Kruggel F., von Cramon D.Y. (1999) Modeling the Hemodynamic Response in Single Trial fMRI Experiments. Magnetic Resonance in Medicine 42, 787-797.

Kruggel F., von Cramon D.Y. (1999) Temporal Properties of the Hemodynamic Response in Functional MRI. Human Brain Mapping 8, 259-271.

Kruggel F., Zysset S., von Cramon D.Y. (1999) Modeling Multiple Hemodynamic Responses in Event-Related Functional MRI Experiments. In: Gedeon T., Wong P., Halgamuge S., Kasabov N., Nauck D., Fukushima K. (eds.), 6th International Conference on Neural Information Processing (Perth), pp. 903-909. IEEE Computer Society Press, Piscataway.

Rajapakse J.C., Kruggel F., Maisog J., von Cramon D.Y. (1998) Modeling the Hemodynamic Response for Analysis of Functional MRI Time-Series. Human Brain Mapping 6, 283-300.

Rajapakse J.C., Kruggel F. (1997) Neuronal and Hemodynamic Events from Functional MRI Time-Series: A Computational Model. In: Kasabov N., Kozma R., Ko K., O’Shea R., Coghill G., Gedeon T. (eds.), Progress in Connectionist-Based Information Systems (Christchurch), pp. 30-34. Springer, Singapur.

Svensen M., Kruggel F., von Cramon D.Y. (1999) Markov Random Field Modelling of fMRI Data Using a Mean Field EM-Algorithm. In: Hancock E.R., Pelillo M. (eds.), Workshop on Energy Minimization Methods in Computer Vision (York), Lecture Notes in Computer Science Vol. 1654, pp. 317-300. Springer, Heidelberg.

Svensen M., Kruggel F., von Cramon D.Y. (2000) Probabilistic Modelling of Single Trial fMRI Data. IEEE Transactions on Medical Imaging 19, 25-36.