- Email: [email protected]
Lid Closure Mimics Head Movement in fMRI
NeuroImage 16, 1156 –1158 (2002) doi:10.1006/nimg.2002.1140
RAPID COMMUNICATION Lid Closure Mimics Head Movement in fMRI Thomas Stephan,* Esther Marx,* Hartmut Bru¨ckmann,† Thomas Brandt,* and Marianne Dieterich‡ *Center for Sensorimotor Research, Department of Neurology, and †Department of Neuroradiology, Ludwig-Maximilians University, 81377 Munich, Germany; and ‡Department of Neurology, Johannes Gutenberg University, 55131 Mainz, Germany Received December 12, 2001
Motion correction algorithms are commonly used to correct for involuntary head motions in fMRI. We found that lid closure increased signal intensity of the eyes by a factor of 1.6 to 2 in EPI-fMRI. If the eyes were included in the scanning volume, transitions from eyes open to eyes closed in darkness were regularly associated with erroneously detected head translations along the z-axis and head rotations around the x-axis in Talairach’s coordinate system. These artefacts could be avoided by masking the eyes in the images during the estimation of movement parameters. © 2002 Elsevier Science (USA)
INTRODUCTION In an fMRI study on the differential effects of eyes open versus eyes closed in complete darkness, we noticed that the eyes, which were included in the scanning volume, showed remarkable differences in signal intensity, depending on whether they were open or closed. Signal intensity of the closed eyes was increased by a factor of 1.6 to 2 compared to that of open eyes (Fig. 1). The transition from eyes open to eyes closed was associated with the detection of task-related head movement during the realignment stage of data analysis (Friston et al., 1995a). This finding prompted us to study the quantitative interrelation between alternating lid closure and the estimated head movement parameters in four healthy volunteers. METHODS Four healthy subjects (three females, one male; mean age 25 ⫾ 2 years) were scanned on a 1.5 Tesla standard clinical scanner (Siemens Vision) using echo planar imaging (EPI) with a T2*-weighted gradientecho multislice sequence (TE ⫽ 60 ms, voxel size 3.75 ⫻ 3.75 ⫻ 3.75 mm 3, matrix 64 ⫻ 64, interscan interval 4.5 s). Thirty-two transversal slices were po1053-8119/02 $35.00 © 2002 Elsevier Science (USA) All rights reserved.
sitioned to cover the whole brain. This procedure included the subjects eyes in the scanning volume. There was a total of 240 image volumes acquired during two imaging runs for each subject except subject C, who underwent only one imaging run with 120 vol. Subjects lying in complete darkness were asked to alternatingly open and close their eyes at a frequency of 0.022 Hz in response to an acoustic signal given via earphones. The subject’s head was carefully fixed in place with a vacuum headrest, and the subjects were also instructed to avoid any head motion. Data processing was performed on UltraSparc Workstations (SUN Microsystems) and SPM99 (Friston et al., 1995b). The first five images of each imaging run were discarded because of spin saturation effects. Motion correction was performed by realigning every volume to the first one of each scanning session, using the method described by Friston et al. (1995a). This processing step was performed twice for every data set, with and without a binary image mask that allowed the exclusion of the eye area from the computations during parameter estimation. In this way two sets of movement parameters were estimated for each data set: one that was computed using the information of the complete image volumes and the second ignoring the region of the eyes. Additionally, mean signal intensity was calculated for a region of interest (ROI), which included both eyes. RESULTS Signal intensity in the ROI was significantly higher (P ⬍ 0.001) in all subjects with the eyes closed than with the eyes open (mean signal intensity 6264 versus 3445). Lid closure was associated with an erroneous detection of head motion along the head z-axis (translation) and around the x-axis (rotation in pitch) in all subjects. With lid closure the mean of the erroneously detected head motion in pitch downward was 0.3° and of caudal head translation 0.2 mm (Fig. 2, left). This
1156
FIG. 1. Axial head slices (38 mm to 26 mm below anterior commissure) acquired using a standard EPI sequence on a Siemens Vision 1.5T scanner with the eyes closed (top) and the eyes open (bottom). Signal intensity is significantly higher with the eyes closed ( p ⱕ 0.001). The ellipse marks the critical area that was used for ROI analysis and was masked during the estimation of movement parameters.
RAPID COMMUNICATION
1157
1158
RAPID COMMUNICATION
FIG. 2. Original recordings of estimated head movement parameters in three dimensions: translations along x, y, and z-axes; and rotations around x, y, and z-axes. The axes are according to the Talairach coordinate system (Talairach and Tournoux 1988). Black rectangles indicate periods during which the subject’s eyes were open. Each eye closure is associated with erroneously detected head movements along the z-axis (top, left) and around the x-axis (bottom, left). These artefacts are eliminated by masking the orbits during the estimation of movement parameters (right).
artefact disappeared, when the eyes were masked during parameter estimation (Fig. 2, right).
simply masking both orbits during the estimation of movement parameters.
COMMENT
ACKNOWLEDGMENTS
The incidental observation that lid closure modified the signal intensity of the eyes in EPI fMRI is important, for it shows that lid closure has a significant impact on the automatic detection of head motion using algorithms based on comparison of voxel intensities. It causes an undesirable misalignment of the images during motion correction, which may lead to erroneous results in the statistical evaluations of the contents of the images. Especially in situations in which components of the estimated movement parameters are correlated with changes in the task, the application of these parameters during realignment of the images may lead to artefactual activations (Friston et al., 1996). Freire and Mangin (2001) have also shown that activated areas may behave like biasing outliers for the registration method used in the SPM99 motion compensation algorithm. This effect, however, cannot explain the artefacts arising from lid closure in our experimental paradigm since they were suppressed by
We are grateful to Judy Benson for critically reading the manuscript. This work was supported by the Deutsche Forschungsgemeinschaft (Clinical Research Group, BR 639/5-3), Stifter Verband, and the Wilhelm-Sander-Stiftung.
REFERENCES Freire, L., and Mangin, J. F. 2001. Motion correction algorithms may create spurious brain activations in the absence of subject motion. NeuroImage 14: 709 –722. Friston, K. J., Asburner, J., Frith, C. D., Poline, J. B., Heather, J. D., and Frackowiak, R. S. J. 1995. Spatial registration and normalization of images. Hum. Brain Mapp. 2: 165–189. Friston, K. J., Holmes, A. P., Worsley, K. J., Poline, J. B., Frith, C. D., and Frackowiak, R. S. J. 1995. Statistical parametric maps in functional imaging: A general linear approach. Hum. Brain Mapp. 2: 189 –210. Friston, K. J., Williams, S., Howard, R., Frackowiak, R. S. J., and Turner, R. 1996. Movement-related effects in fMRI time-series. Magn. Reson. Med. 35(3): 346 –355. Talairach, J., and Tournoux, P. 1988. Coplanar Stereotactic Atlas of the Human Brain. Thieme, Stuttgart/New York.
RAPID COMMUNICATION Lid Closure Mimics Head Movement in fMRI Thomas Stephan,* Esther Marx,* Hartmut Bru¨ckmann,† Thomas Brandt,* and Marianne Dieterich‡ *Center for Sensorimotor Research, Department of Neurology, and †Department of Neuroradiology, Ludwig-Maximilians University, 81377 Munich, Germany; and ‡Department of Neurology, Johannes Gutenberg University, 55131 Mainz, Germany Received December 12, 2001
Motion correction algorithms are commonly used to correct for involuntary head motions in fMRI. We found that lid closure increased signal intensity of the eyes by a factor of 1.6 to 2 in EPI-fMRI. If the eyes were included in the scanning volume, transitions from eyes open to eyes closed in darkness were regularly associated with erroneously detected head translations along the z-axis and head rotations around the x-axis in Talairach’s coordinate system. These artefacts could be avoided by masking the eyes in the images during the estimation of movement parameters. © 2002 Elsevier Science (USA)
INTRODUCTION In an fMRI study on the differential effects of eyes open versus eyes closed in complete darkness, we noticed that the eyes, which were included in the scanning volume, showed remarkable differences in signal intensity, depending on whether they were open or closed. Signal intensity of the closed eyes was increased by a factor of 1.6 to 2 compared to that of open eyes (Fig. 1). The transition from eyes open to eyes closed was associated with the detection of task-related head movement during the realignment stage of data analysis (Friston et al., 1995a). This finding prompted us to study the quantitative interrelation between alternating lid closure and the estimated head movement parameters in four healthy volunteers. METHODS Four healthy subjects (three females, one male; mean age 25 ⫾ 2 years) were scanned on a 1.5 Tesla standard clinical scanner (Siemens Vision) using echo planar imaging (EPI) with a T2*-weighted gradientecho multislice sequence (TE ⫽ 60 ms, voxel size 3.75 ⫻ 3.75 ⫻ 3.75 mm 3, matrix 64 ⫻ 64, interscan interval 4.5 s). Thirty-two transversal slices were po1053-8119/02 $35.00 © 2002 Elsevier Science (USA) All rights reserved.
sitioned to cover the whole brain. This procedure included the subjects eyes in the scanning volume. There was a total of 240 image volumes acquired during two imaging runs for each subject except subject C, who underwent only one imaging run with 120 vol. Subjects lying in complete darkness were asked to alternatingly open and close their eyes at a frequency of 0.022 Hz in response to an acoustic signal given via earphones. The subject’s head was carefully fixed in place with a vacuum headrest, and the subjects were also instructed to avoid any head motion. Data processing was performed on UltraSparc Workstations (SUN Microsystems) and SPM99 (Friston et al., 1995b). The first five images of each imaging run were discarded because of spin saturation effects. Motion correction was performed by realigning every volume to the first one of each scanning session, using the method described by Friston et al. (1995a). This processing step was performed twice for every data set, with and without a binary image mask that allowed the exclusion of the eye area from the computations during parameter estimation. In this way two sets of movement parameters were estimated for each data set: one that was computed using the information of the complete image volumes and the second ignoring the region of the eyes. Additionally, mean signal intensity was calculated for a region of interest (ROI), which included both eyes. RESULTS Signal intensity in the ROI was significantly higher (P ⬍ 0.001) in all subjects with the eyes closed than with the eyes open (mean signal intensity 6264 versus 3445). Lid closure was associated with an erroneous detection of head motion along the head z-axis (translation) and around the x-axis (rotation in pitch) in all subjects. With lid closure the mean of the erroneously detected head motion in pitch downward was 0.3° and of caudal head translation 0.2 mm (Fig. 2, left). This
1156
FIG. 1. Axial head slices (38 mm to 26 mm below anterior commissure) acquired using a standard EPI sequence on a Siemens Vision 1.5T scanner with the eyes closed (top) and the eyes open (bottom). Signal intensity is significantly higher with the eyes closed ( p ⱕ 0.001). The ellipse marks the critical area that was used for ROI analysis and was masked during the estimation of movement parameters.
RAPID COMMUNICATION
1157
1158
RAPID COMMUNICATION
FIG. 2. Original recordings of estimated head movement parameters in three dimensions: translations along x, y, and z-axes; and rotations around x, y, and z-axes. The axes are according to the Talairach coordinate system (Talairach and Tournoux 1988). Black rectangles indicate periods during which the subject’s eyes were open. Each eye closure is associated with erroneously detected head movements along the z-axis (top, left) and around the x-axis (bottom, left). These artefacts are eliminated by masking the orbits during the estimation of movement parameters (right).
artefact disappeared, when the eyes were masked during parameter estimation (Fig. 2, right).
simply masking both orbits during the estimation of movement parameters.
COMMENT
ACKNOWLEDGMENTS
The incidental observation that lid closure modified the signal intensity of the eyes in EPI fMRI is important, for it shows that lid closure has a significant impact on the automatic detection of head motion using algorithms based on comparison of voxel intensities. It causes an undesirable misalignment of the images during motion correction, which may lead to erroneous results in the statistical evaluations of the contents of the images. Especially in situations in which components of the estimated movement parameters are correlated with changes in the task, the application of these parameters during realignment of the images may lead to artefactual activations (Friston et al., 1996). Freire and Mangin (2001) have also shown that activated areas may behave like biasing outliers for the registration method used in the SPM99 motion compensation algorithm. This effect, however, cannot explain the artefacts arising from lid closure in our experimental paradigm since they were suppressed by
We are grateful to Judy Benson for critically reading the manuscript. This work was supported by the Deutsche Forschungsgemeinschaft (Clinical Research Group, BR 639/5-3), Stifter Verband, and the Wilhelm-Sander-Stiftung.
REFERENCES Freire, L., and Mangin, J. F. 2001. Motion correction algorithms may create spurious brain activations in the absence of subject motion. NeuroImage 14: 709 –722. Friston, K. J., Asburner, J., Frith, C. D., Poline, J. B., Heather, J. D., and Frackowiak, R. S. J. 1995. Spatial registration and normalization of images. Hum. Brain Mapp. 2: 165–189. Friston, K. J., Holmes, A. P., Worsley, K. J., Poline, J. B., Frith, C. D., and Frackowiak, R. S. J. 1995. Statistical parametric maps in functional imaging: A general linear approach. Hum. Brain Mapp. 2: 189 –210. Friston, K. J., Williams, S., Howard, R., Frackowiak, R. S. J., and Turner, R. 1996. Movement-related effects in fMRI time-series. Magn. Reson. Med. 35(3): 346 –355. Talairach, J., and Tournoux, P. 1988. Coplanar Stereotactic Atlas of the Human Brain. Thieme, Stuttgart/New York.