A Preliminary Study of Major Depressive Disorder Using Simultaneous PET/fMRI with Two MID Tasks in a Single Scan
Fuyixue Wang1, Paul Hamilton2, Brian Knutson2, Ian Gotlib2, Matthew Sacchet2, Hershel Mehta2, Christina Schreiner2, Dawn Holley3, Fred Chin3, Bin Shen3, Greg Zaharchuk3, Mehdi Khalighi4, and Gary Glover3

1Department of Biomedical Engineering, Tsinghua University, Beijing, China, People's Republic of, 2Department of Psychology, Stanford University, Stanford, CA, United States, 3Department of Radiology, Stanford University, Stanford, CA, United States, 4Applied Science Lab, GE Healthcare, Menlo Park, CA, United States


The release of dopamine during reward tasks is modulated by major depressive disorder (MDD). In this study of MDD, we used simultaneous PET/fMRI to detect the neurochemical changes of dopamine and neurovascular activity through BOLD contrast during two sequential monetary incentive delay tasks in a single scan. Several modeling methods were proposed and evaluated for dynamic PET data. Six participants with MDD were studied. The results of the group analysis of PET and fMRI show significant effects of dopamine release in ventral striatum bilaterally, close to the nucleus accumbens, and significant BOLD signals in putamen bilaterally during reward tasks.


The release of dopamine (DA) during reward tasks is modulated by major depressive disorder (MDD) 1. This multi-modal study of MDD used PET and fMRI simultaneously to detect and measure the neurochemical changes of DA by PET and neurovascular activity through BOLD contrast during monetary incentive delay tasks. Specifically, in this pilot study, different modeling methods of dynamic PET data were proposed and evaluated for concurrent PET-fMRI study with two sequential monetary incentive delay (MID) tasks in a single scan.


Data acquisition: We simultaneously acquired PET and fMRI data on a PET/MR system (GE Healthcare, Milwaukee). Six participants with major depressive disorder underwent a 50min or 40min PET scan with [11C] raclopride and a 32min MR scan simultaneously. During the scan, participants performed two 10-min MID tasks (high ($5) and low ($1) stakes) sequentially, each divided into two conditions including reward (winning or not losing money) and punishment (losing money).

fMRI analysis: Slice-timing correction, motion correction, spatial smoothing and a quadratic detrending were performed for all fMRI data. Statistical analysis was performed using a two-stage mixed effect model. In the first stage, voxel-wise regression analysis of the data of each subject was performed. The hemodynamic response was modeled by convolving the task design with a canonical hemodynamic response function 2. Each subject's brain data were normalized to the standard template provided by the Montreal Neurological Institute (MNI) template using SPM8 3. In the second stage, one-sample t tests were calculated over images of the interested contrasts to obtain group activation maps.

PET analysis: For PET data, statistical analysis was performed using general linear model (GLM). The task related regressors were derived from the results obtained in the previous experiments in which variations due to endogenous dopamine release in the striatum were modeled using two exponential functions with time constants of 3 min. To detect task-related signal changes in response to two sequential MID tasks, three methods, quadratic fitting, gamma fitting of the average signal of striatum region with 1st or 2nd order detrend, have been tried by including them in the GLM as covariates of no interest. They were compared with the conventional kinetic model LSSRM 4.

Kinetic simulations were performed to simulate the time-activity curves (TACs) of the whole brain in order to evaluate the various modeling methods. In the first set of simulations, different amounts of task-related signal decreases were added to the TACs of voxels in striatum region and the proportion of activated voxels were calculated to examine the sensitivity of detecting ligand displacement. The second set of simulations examined the effect of noise on detection of signal using different modeling methods, and tested the false positive rate when there was no task added.

The analysis of the human studies used a GLM model constructed with different modeling methods as described above to obtain the activation maps during the two MID tasks. Group analysis was also performed using one-sample t test.


The results of kinetic simulations are shown in Fig. 1. The sensitivity of all methods to detect signal changes in striatum increased with the decrease of noise and the rise of signal changes. Fig. 2 shows examples of TACs for striatum and cerebellum and their corresponding fitting curves using different methods in the simulation and the human experiment. Single subject t statistical maps acquired from the three methods are compared in Fig. 3.

Fig. 4 illustrates the results of group analysis of both PET (using quadratic fitting) and fMRI. The significant effects due to dopamine release were found in ventral striatum bilaterally, close to the nucleus accumbens in both high and low stake tasks. Significant BOLD signals were located in putamen bilaterally when subjects were studied during reward tasks. But significance was only found at right putamen during high stake punishment tasks.

Discussion and Conclusion

The PET kinetic simulations demonstrate the feasibility of all the methods to detect task-related signal changes in striatum with relatively low false positive rate. The agreements of the fitting curves and the single subject t statistical maps indicate the similar estimation characteristics of the three modeling methods.

Compared with the conventional kinetic model, quadratic and gamma fitting for PET data are more simple and flexible for different experimental designs and task strategies of concurrent PET-fMRI, and only quadratic fitting shows activations in striatum in group analysis in this study. The results of group analysis demonstrate the dopamine release and fMRI activations in the ventral striatum during the MID reward tasks for MDD patients.


We acknowledge GE Healthcare for assistance. This work was supported by NIH EB01589, Weston Havens Foundation. The authors thank Jingyuan Chen for help.


1. Knutson B, Bhanji J P, Cooney R E, et al. Neural responses to monetary incentives in major depression. Biological psychiatry, 2008, 63(7): 686-692.

2. Friston K J, Fletcher P, Josephs O, et al. Event-related fMRI: characterizing differential responses. Neuroimage, 1998, 7(1): 30-40.

3. http://www.fil.ion.ucl.ac.uk/spm/

4. Alpert N M, Badgaiyan R D, Livni E, et al. A novel method for noninvasive detection of neuromodulatory changes in specific neurotransmitter systems. Neuroimage, 2003, 19(3): 1049-1060.


FIG. 1. PET simulation study. a) Proportion of activated voxels (p<0.05) in striatum with different amounts of simulated task-related signal changes. b) Proportion of activated voxels (p<0.05) in striatum with and without simulated task-related signal changes at various noise levels.

FIG. 2. PET simulation study and human experiment. Time activity curves of voxels in striatum and cerebellum respectively and the corresponding fitting curves using different methods in the simulation (a) with noise level of 12% and the human experiment (b).

FIG. 3. PET single subject in the human experiment. Activated voxels (t statistic maps were thresholded at significance level of p<0.05) spatially normalized to the MNI space displaying dopamine release in response to high stake (a) and low stake (b) MID tasks using the three different modeling methods.

FIG. 4. PET & fMRI Group analysis. The t statistic maps (thresholded at p<0.005, combined with cluster size threshold of 100 voxels) of group analysis of PET (quadratic fitting) and fMRI during the two MID tasks (high and low stake), each divided into two conditions including reward and punishment.

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)