Utilizing Hemodynamic Delay and Dispersion to Detect fMRI Signal Change Without Auditory Interference: The Behavior Interleaved Gradients Technique G.F. Eden, 1 * J.E. Joseph, 1 H.E. Brown, 1 C.P. Brown, 1 and T.A. Zeffiro 1,2 A major problem associated with the use of functional magnetic resonance imaging (fMRI) is the attendant gradient noise, which causes undesirable auditory system stimulation. A method is presented here that delays data acquisition to a period immedi- ately after task completion, utilizing the physiological delay and dispersion between neuronal activity and its resulting hemody- namic lag. Subjects performed finger movements with the gradients off, followed by a rest period with the gradients on. This resulted in task-related signals comparable to those ob- tained with concurrent task performance and image data acqui- sition. This behavior interleaved gradients technique may be particularly useful for the studies involving auditory stimula- tion or overt verbal responses. Magn Reson Med 41:13–20, 1999.  1999 Wiley-Liss, Inc. Key words: neuroimaging; motor cortex; auditory; behavior interleaved gradients; single-trial fMRI Even though the efficacy of functional magnetic resonance imaging in detecting task-related signal change reflecting sensory and cognitive processing has been demonstrated, fMRI data acquisition involves high levels of potentially contaminating acoustic noise. Language or auditory process- ing studies are particularly susceptible to interference from this source. Gradient noise can adversely affect detection of task-related signal change in numerous ways. First, the background noise generated by the scanner may alter difficulty levels in tasks involving acoustically presented stimuli such as tone discrimination or verbal processing. This acoustic interference may not be consistent across conditions. Second, gradient noise may also induce hemo- dynamic changes in regions involved in auditory process- ing. Third, when using non-auditory stimuli in visual, motor, or tactile tasks, acoustic noise may distract the subject from the task at hand. Some of these potentially contaminating effects might be mitigated if the signal changes induced by the gradient noise were linearly additive with task-related signal changes. For example, many sensory and cognitive studies using fMRI are based on a subtraction analysis technique. This approach assumes that the physiological changes induced by the gradient noise are similar in both the control and task conditions and therefore cancel. Should this ‘‘subtraction’’ assumption be incorrect, noise-induced signal change might be misidentified as task-related signal change. Furthermore, the range of modulation between the resting and activation states may be artificially decreased due to saturation of auditory cortical activity by the relatively loud gradient noise (1,2). In auditory experiments, attempts have been made to reduce the effects of the gradient noise by acoustically insulating the subject. To date there exist no wholly successful methods for noise reduction, with partial suc- cess resulting from the use of earplugs and acoustic shielding. This poses a serious problem for fMRI compared with other neuroimaging techniques, such as positron emission tomography (PET), that acquire image data in a comparatively quiet environment. As noise abatement does not seem technically feasible at present, an alternative approach involves confining gradient noise to a time in which it will have minimal contaminating effects on task-related signal change (3,4). This is accomplished by utilizing the hemodynamic delay between neural activity and its resulting hemodynamic response. In response to neural activity, there is an increase of blood flow and oxygen delivery (5,6). Because the oxygen utilization is far less than the amount delivered during blood volume changes, there is a net deoxyhemoglobin concentration decrease, resulting in the blood oxygenation-level depen- dent (BOLD) contrast response. This response is delayed in time by 4–8 sec and lasts far longer than its antecedent neural activity (7–9). However, as the delay and dispersion times are approximately known, this phenomenon may be used to advantage to capture the task-related signal long after task completion. In the present method, the gradients were off during periods of task execution and then immediately switched on for image acquisition, effectively interleaving task per- formance and its detection with echo planar imaging (EPI). We call this the behavior interleaved gradients technique. In addition to this delay, the response is also temporally dispersed, with a brief burst of neuronal activity resulting in a hemodynamic response with a longer time scale. This delay and dispersion allows detection of task-related activ- ity after its corresponding neural activity has ceased. In this study, we used a simple motor task, known to produce robust signal changes, to compare the interleaved data acquisition approach with more conventional techniques in which image data acquisition and task performance are concurrent. MATERIALS AND METHODS Six right-handed control subjects (five females and one male; mean age 28 years, range 24–33 years) participated in this study. Hand dominance was determined with the 1 Georgetown Institute for Cognitive and Computational Sciences, Georgetown University Medical Center, Washington DC 20007. 2 Sensor Systems, Sterling, Virginia. Grant sponsor: Charles A. Dana Foundation; Grant sponsor: Department of Defense; Grant number: DAMD17–93-V-3018. *Correspondence to: Guinevere Eden, Georgetown University Medical Center, Washington, DC 20007. E-mail: eng@giccs.georgetown.edu Received 6 April 1998; revised 27 August 1998; accepted 8 September 1998. Magnetic Resonance in Medicine 41:13–20 (1999) 13  1999 Wiley-Liss, Inc.