Improving fMRI signal detection usingphysiological dataClustered volume acquisition - to reduce the impact of scanner acoustic noise during fMRICardiac gating - to improve the detectionof brainstem activationBrainCochleaAuditoryNerveEardrumThe Peripheral Auditory SystemHeschl'sGyrusRostralCaudalAuditory PathwayInferiorColliculusMedialGeniculateBodyVIIINerveCochleaMedullaCochlearNucleusPonsAuditoryCortexp = 0.001Music to Both Ears Heschl’sGyrus2 x 10-9~~TIME(sec)MusicOn(30 s). . .03060 90MusicOff(30 s) (30 s)MusicOffInferiorColliculiCardiac GatingElectrocardiogram~ 2 secSynchronize image acquisitions to the subject’scardiac cycle.0Cardiac gating introducessignal variations. RegionSignalInRegionTime0Ungated(TR = 2 sec)Gated(TR 2 sec)Gated & Corrected(TR 2 sec)Example: heart rate = 60 - 70 bpm S1 = signal at 70 bpm S2 = signal at 60 bpmPercent S2 - S1Change S1= x 100 = 5%With gating, TR is not constant because heart rate fluctuates.The variations in TR cause image-to-image variationsin signal. The signal variations are comparable to (or greater than)the changes associated with activation.Where do the signal variations come from?Signal ~ 1 - exp(-TR/T1)TR (sec)Signal70 bpm 60 bpm1.7 2.0S2S1How can gating-related signal variations be removed?For any given voxel,Sn = An [1 - exp(-TRn/T1)] (1)Sn is the signal in the nth image.TRn is the inter-image interval (measured during experiment).An includes signal changes associated with activation. TRnImagen-1ImagenModel the signal variationsCorrect the signal based on the model (use eq. (2)).Corrected Sn = Sn (2)[1 - exp(-TRav/T1)][1 - exp(-TRn/T1)]The correction effectively removes any signal variations explained by the model.Consider a correction that removes these variations.TRav = the average inter-image interval during the experimentT1 is chosen to minimize variations in the corrected signal.0Cardiac gating introducessignal variations. However, these can beremoved.RegionSignalInRegionTime0Ungated(TR = 2 sec)Gated(TR 2 sec)Gated & Corrected(TR 2 sec)Gated & CorrectedUngatedInferiorColliculus(Guimaraes, et al., 1998,Human Brain Mapping, 6:33-41)InferiorColliculiInferiorColliculiHeschl’sGyrus0.5 cmAcoustic Noise During fMRI:pump noise - produced by the liquid helium pumpgradient noise - produced by the gradient coils each time an image is acquired250Acoustic Noise420-2-40 50 100 150 200 250Time (msec)Sound Pressure (Pa)-501.5 T 3 TSound Pressure Level (dB)Frequency (kHz)806040200.1 1.0 10AirHandlingPumpGradients(Pump Off)10000 50 100 150 200Time (msec)Sound Pressure (Pa)-5015100-5-105-15Sound Pressure Level (dB)Frequency (kHz)806040200.1 1.0 10AirHandlingPump1000GradientsAsymmetricSpin EchoGradientEchoPumpGradientsPumpGradientsAcoustic conditions differ from mostauditory studiesDifficulties hearing sound stimuliSuppression of fMRI activation inauditory areas Problems Posed by Acoustic Noise:Effect of Gradient Noise on Auditory Activation(Music to Both Ears)Right Left3 images / sec(6 Slices)1 / sec(2 Slices)0.5 / sec(1 Slice)Heschl’sGyrusNeuralActivitySound Level (dB SPL)0 10 30 50 70 90Neural Activity vs. Sound Levelin the Auditory NerveSound Stimulus On10 secResponseto SoundStimulusResponseto GradientNoiseClustered Volume AcquisitionVolumeAcquiredClustered Volume Acquisition(TR = 8 sec)Distributed Volume Acquisition(TR = 8 sec)Subject 1Subject 2Apapted from Edmister et al.Human Brain Mapping 7: 89-97
View Full Document