1. A practical review of common artifacts encountered in fMRI studies in the pediatric population and methods to avoid misinterpretation K Shekdar, R Golembski, D Zarnow, A Vossough, E Schwartz, T Roberts

2. Disclosures: Nothing to disclose

3. Purpose To describe some artifacts commonly encountered in clinical fMRI studies in the pediatric population and methods to avoid their misinterpretation.

4. Materials and Methods We retrospectively reviewed clinical fMRI studies performed at our institute from 2008 to 2014 and evaluated some of the commonly encountered artifacts. All fMRI studies were performed on a 3T Siemens scanner. Primary fMRI post processing was done on an FDA- approved Siemens Leonardo post-processing workstation. A commercially available post processing software (Brain Voyager) with augmented motion correction capabilities was also used in conjunction.

5. Materials and Methods The fMRI assessment was mostly done for either 1- Language localization and/or 2- Motor mapping Initial sequence : Axial 3D MP-RAGE fMRI gradient echo BOLD (Blood Oxygen Level Dependent) sequence Activation determined with the general linear model (GLM) Task performance was monitored using button box recording and direct observation

6. Materials and Methods Language function tasks 1- Picture naming 2- Passive Listening 3- Verb generation 4- Auditory discrimination 5- Semantic categories Motor mapping (MM) MM of the hands (right and left index finger D2) Performed with 30 second alternating rest and active blocks with a self-paced (1-2 second) button press using each hand sequentially in separate runs, to evaluate activity- dependent signal intensity variations. In a few patients with neurological deficits affecting hands, an attempt was made to achieve MM with foot pressing if feasible.

7. Results Total 30 cases 16 cases: mass lesions 4 cases: vascular lesions 10 cases: medically refractive epilepsy being considered for surgery

8. Results Most artifacts were related to motion or stimulus correlated subtle motion Other artifacts due to presence of vascular structures or pathology The general linear model (GLM) described the time course of areas of apparent activation including translational and rotational parameters of movement, components of which we proceeded to regress out, yielding maps less contaminated by motion

9. 6 yr/F, with h/o neonatal left MCA stroke being evaluated for epilepsy surgery. Fig. A: LD2 task showed activity in the vicinity of the right premotor cortex(yellow arrow). Also note the lowered threshold (red arrow) used to visualize the activity. B:Correlation of the BOLD signal with respect to translational and rotational motion A B Case 1 T2 axial LD2: Brain voyager analysis with motion correction showed no activity

10. Left foot GLM The general linear model (GLM) described the time course of areas of apparent activation including translational and rotational parameters of movement, components of which we proceeded to regress out, yielding maps less contaminated by motion.

11. False positive BOLD signal due to motion How to Improve Improved motion correction post- processing software. Use of real-time motion detection capability would allow repeat scans and potentially salvage studies.

12. left foot: Accidental auditory response Case 1(cont.): Fig. A: Left foot task showed activity in the right auditory cortex(yellow arrow). B:Correlation of the BOLD signal with respect to translational and rotational motion AB

13. False positive: BOLD signal due to ‘accidental secondary paradigm’ Verbal instructions(start/stop)of the motor task can elicit auditory response in addition to intended response.

14. 16-year-male with focal epilepsy, right temporal mass lesion, needing preoperative fMRI for language localization. Activity noted within the right temporal lobe (yellow arrows)correlating to the location of the known lesion(white arrow) is presumed to be stimulus correlated motion (green arrows) masquerading as activity. Case 2:

15. Artifact from Stimulus correlated motion(SCM) Tasks requiring facial movement i.e., tongue tapping, nose & mouth movement, loud or overt speech are prone to motion Even small movements < 1mm can have profound effects on functional connectivity MRI Careful analysis of the BOLD signal with respect to the timing of stimulus application should differentiate SCM from true activation.

16. ADT: No activity, task too difficult Medically refractory epilepsy in 11 year/F, with lesion in left temporal lobe (white arrow), needing preoperative fMRI for language localization prior to surgery. Case 3:

17. ADT2- Changing to easier task elicited BOLD signal (yellow arrow) Case 3:

18. No signal due to difficult tasks How to Improve Need to use maintain an inventory of different language tasks of varying difficulty levels to use for children of different ages or those with mental delays If no visual response is detected in real time, may consider using an easier task Consider assessing the patient’s ability to perform or respond to the particular level of a language task before the fMRI to judge the patient’s capability

19. False negative due to maximal vasodilation Functional MRI examination for motor activation could not successfully identify the left primary motor cortex despite confirmed task compliance, likely due to shunting of the known arteriovenous malformation in this region. 13 yr/ M, with ruptured left temporal/posterior frontal arteriovenous malformation (red arrows), for consideration of resective surgery, needing preoperative fMRI for primary motor cortex localization

20. False negative possible due to reorganisation Functional MRI examination consistently, across multiple tasks) demonstrated activity (yellow arrow) within the right frontal lobe in keeping with Broca area, suggesting a component of reorganization, as the prior fMRI study had shown left frontal lobe activity. Given lack of knowledge of the vascular reactivity of peri-surgical tissue, the possibility of this being a false negative finding cannot be excluded.

21. 12 yo Right handed female with a diffuse left temporal lobe seizure, to assess language for surgical planning. fMRI examination consistently demonstrated clusters of activity(yellow arrows) within the left inferior lateral frontal region in keeping with Broca area and in the left posterior temporal region within the patient’s known lesion (white arrow). Case:

22. Recognizing True activity The BOLD activity was shown to be consistently within the patient’s known lesion and contrary to expectation, it was not shifted. The patient did not go to surgery. Multiple repetitions may be needed to confirm activity especially within a lesion, reliably.

23. Susceptibility artifact False negative BOLD signal can occur in regions obscured by artifact Prior craniotomy (white arrow) overlying recurrent left temporal tumor (red arrow) results in susceptibility signal loss (yellow arrow)

24. Coregistration Failure Misregistration can result in poor spatial accuracy on activation maps. Useful to perform manual inspection of registration before overlaying activation maps

25. Patient Stimulus Neuro Vascular & coupling Hemodynamic Response BOLD signal Post- processing The different components of obtaining fMRI study data Every component involved in the process of obtaining information from a fMRI study is prone to artifacts.

26. Conclusions Bold-fMRI is prone to different artifacts and sources of error Performing fMRI studies in the pediatric population is challenging. It is important to be familiar with commonly encountered artifacts and methods to avoid misinterpretation.