Clinical pediatric populations have a lower success rate for completing fMRI runs than typically developing children matched on age, IQ, and gender. Adolescents completed more fMRI runs than younger children. The success rate for children with ADHD while on or off their prescribed MPH dose did not differ significantly, although 10–12-year-olds on MPH achieved a higher success rate than 7–9-year-olds on MPH. Completion of an entire fMRI battery was equivalent for the Epilepsy group (non pre-surgical and pre-surgical) and their matched controls for 4–12-year-olds, whereas 7–9-year-olds with ADHD (on or off MPH) and 7–12-year-olds with ASD were less successful than their TYP controls. When examining whether task difficulty contributed to success rates, we found lower success rates for completing an fMRI battery for the 7–12-year-old TYP group in the Epilepsy study than in the ADHD or ASD studies. The TYP group in the ASD study was also less likely to complete at least one fMRI run compared to the TYP groups in the Epilepsy or ADHD studies. When examining task design (event-related versus blocked) within the TYP ASD group, we did not find a difference in any of the success rates. Excessive head motion was the leading cause for an unsuccessful run across all groups.
Acquiring valid fMRI data in pediatric patient populations requires accounting for the success rates of different clinical populations and ages. While children with ASD would require the most additional participants (26–34% more), all clinical groups below 13-years-old have a failure rate of at least 15%. In contrast, the TYP controls have a failure rate of 10–15%, with the exception of 4–6-year-olds who have a failure rate of 22%.
Age plays an important role in completing an entire fMRI battery. TYP controls completed an entire battery more readily (69%); when 4–6-year-olds are excluded their complete fMRI battery success rate increases to 79%. Moreover, within 7–12-year-olds the Epilepsy group (51%) completed an entire fMRI battery at the same rate as ADHD (off MPH: 50%; on MPH: 50%), and ASD (59%). The significant age related findings within the Epilepsy and TYP control groups suggest that age may play a larger role in completing fMRI batteries than diagnosis. This age finding is strengthened when considering that 4–6-year-olds received a shortened, developmentally appropriate battery designed to improve their success rates. Of note, the TYP controls in the Epilepsy studies had the lowest success rates of all TYP control groups. While all TYP children scanned within 1 hour time blocks, the Epilepsy studies had the most fMRI runs (4 to 6), and the higher failure rate may have occurred in the later runs.
On the surface, fMRI task type may appear to play a role in success rate (e.g., language tasks may appear more difficult based on data above); however, two alternatives may explain the lower success rate for completing a battery in the TYP group in the Epilepsy studies better than task difficulty. Children in the Epilepsy group were asked to complete anywhere from four to six runs, whereas children in the other groups were asked to complete three to four runs. Thus, the lengthier battery may be driving the group differences rather than the language tasks being more difficult than the executive control and learning tasks. Another potential explanation is that an “other” category existed for tallying undocumented reasons for failed scans in research participants, and this category accounted for 33% of failed runs in the 7–12-year-old TYP group in the Epilepsy studies. Therefore, it is possible that some portion of these incomplete batteries may have resulted from a technical issue (e.g., E-prime did not trigger in synch with scan), or a non-task related reason for stopping the scanner (e.g., scan was stopped so child could use the restroom). As a result, this group difference may result from an overly conservative estimate of failures in the Epilepsy study that do not reflect task differences.
Success rate differences in the TYP children in the ASD studies may also be driven by factors other than task type. Although this sub-group of TYP children was also less likely to complete a single task, an examination of the data reveals a potential ceiling effect. The TYP children in the ASD studies had a success rate of 91%, in comparison to 100% from the other groups. Moreover, TYP children in the ASD group completed fewer runs than TYP children in the other studies, and this may have afforded them fewer opportunities to complete at least one run. Taken together, these data suggest that while there may have been some statistically significant differences for success rates across different fMRI task types, it is more likely that other factors such as battery length and ceiling effects may be driving the observed group differences.
As the failure data are descriptive, these findings should be considered preliminary. Excessive head motion was the leading cause for failed scans across all groups. While the use of MPH in children with ADHD did not affect overall fMRI success rate, our failure data suggest that children taking their prescribed dose of MPH had the lowest failed fMRI runs due to excessive head motion. After movement, the second most common reason for failure differed by diagnostic group. Null activation was the reason for the Epilepsy group, undocumented reason for the ADHD group, and refusal to complete a run for the ASD group. These secondary reasons may or may not be specific to the diagnostic group. For example, the higher refusal rate inside the scanner for the ASD group may result from sensitivity to auditory stimuli (Tomchek & Dunn, 2007
) or anxiety common in this population (Leyfer et al., 2007). On the other hand, the null activation in the clinical pre-surgical Epilepsy group may be related to the scan time that the clinical study runs (4–8pm on a weeknight) resulting in a higher likelihood that a child falls asleep in one or more runs.
When using a similar definition of scan success as a previous report (Byars et al., 2002
), all groups with the exception of the ASD group completed at least one fMRI run over 90% of the time. Twenty-one of 22 (95%) of the TYP 4–6-year-olds completed at least one run, which exceeded the previous report of 47% of TYP 5–6-year olds completing at least one run (Byars et al., 2002
). One notable difference between the current Epilepsy study that recruited 4–6-year-old TYP children and the previous study is the number of fMRI runs in the batteries – two in the Epilepsy study and four in the previous study. The Epilepsy study also reduced cognitive demands by not presenting instructions for multiple runs all at one time. Thus, these small changes in the protocol execution may have contributed to the difference in single run success rate for our 4–6-year-olds.
There are potential factors that limit the generalization of our findings. The pre-surgical Epilepsy study had a different aim than the other studies, and therefore employed a different definition for success. The pre-surgical study deemed “null activation” maps (individual maps with no significant clusters of activation) as failed runs because of the need to derive pre-surgical language maps; whereas a null activation map in the other studies was considered successful because sub-threshold maps at the individual level contribute meaningfully to the random effects group maps. Another factor is that our studies varied in how stringently the one voxel rule for head movement was applied. In the pre-surgical Epilepsy study, knowledge of expected and variant patterns of activation at the individual level allowed clinicians to read through motion artifact, and therefore the one voxel rule was violated for clinical gain. This approach likely contributed to the different findings within the Epilepsy group for completing an entire fMRI battery. The ADHD study strictly adhered to the one voxel rule and children were excluded if head movement exceeded one voxel. The ASD study adhered to the one voxel rule except in the case of four runs (from a total of 176) where children moved greater than one voxel once during the run; children who moved more than one voxel repeatedly were excluded. A future study could examine head movement across diagnosis and age as an additional measure of fMRI compliance.
Another limiting factor to discuss is the usefulness of the mock scanner protocol employed by the Epilepsy study. The Epilepsy study only employed the mock scanner protocol for children 4–10-years-old, and this did not appear to be an important factor in improving fMRI success rates. No analysis showed that the Epilepsy group or the TYP group in the Epilepsy study, completed fMRI runs or batteries at a greater percentage relative to ASD and ADHD groups and their respective matched control TYP groups. Although speculative, one explanation may be that our research group makes a concerted effort for participants to interact with the same staff members at multiple research appointments (neuroimaging and psychological testing). Maintaining the same staff across settings may reduce parent and child anxiety, and increase compliance in task completion. Another potential explanation is that the lengths of the fMRI batteries are adjusted based on the child’s age. Children under the age of seven completed only two runs, and their success rates were comparable to the closest age group (7–9-year-olds). The youngest children may have been less fatigued and more likely to successfully complete this shorter fMRI battery, whereas children over 10 years of age are able to complete up to six runs successfully.
There are recent improvements in mock scanner protocols, such as using head motion feedback to inform the child if movement is within acceptable limits. One pilot study showed that an extensive mock scanner protocol that included eight mock scans lasting seven minutes each in conjunction with operant conditioning techniques (e.g., inform child of baseline motion and then positively reinforce small increments of improvement) reduced head motion from a baseline of 31.3 mm/min to 0.98 mm/min. in 7–10-year-olds (Slifer, Koontz, & Cataldo, 2002
). Other recent feedback techniques include computer monitoring of head motion. During this technique, children watch a video or engage in the task of interest and the computer pauses the video or task if head motion exceeds a given threshold (e.g., 1 mm in any direction). The current study did not use such extensive mock scanning protocols, and therefore it is unknown whether they would substantially improve fMRI success rates. Therefore, a future study on pediatric fMRI success rate could compare the use of such mock scanner protocols to older protocols (or no protocols).
In addition to mock scanner procedures, other strategies that may prove useful for increasing scan success include explaining task directions and the MRI environment at developmentally appropriate levels, as well as limiting the length of fMRI batteries. For children under the age of seven, our research group only provides instructions for one fMRI task prior to the child entering the MRI environment; instructions for the second fMRI task are given while the child is already in the MRI environment. Avoiding overloading very young children with multiple task instructions may aid in limiting performance anxiety or anxiety about the MRI environment. While this study did not explicitly manipulate fMRI battery length, using a shorter battery thought to be more developmentally appropriate based on the child’s age helped maintain the individual run success rates for 4–7-year-old children similar to 7–9-year-old children.
Taken as a whole, our data indicate that research and clinical application fMRI studies need to establish appropriate recruitment goals in order to reach their desired sample size if patient and/or younger populations are involved. Furthermore, strategies that increase success rate in these populations need to be applied (i.e. maintaining the the same staff across appointments, presenting instructions one run at a time for young children, limiting time in scanner), and investigators need to plan for the possibility of only analyzing partial data sets.