Sophia Mueller MD, Presenter: Nothing to Disclose

Danhong Wang, Abstract Co-Author: Nothing to Disclose

Michael D. Fox, Abstract Co-Author: Nothing to Disclose

Ruiqi Pan, Abstract Co-Author: Nothing to Disclose

Jie Lu MD, PhD, Abstract Co-Author: Nothing to Disclose

Kucheng Li, Abstract Co-Author: Nothing to Disclose

Randy L. Buckner, Abstract Co-Author: Consultant, Pfizer Inc Consultant, Johnson & Johnson

Hesheng Liu, Abstract Co-Author: Nothing to Disclose

Disease-related alterations in neuronal networks can be estimated using functional connectivity MRI (fcMRI). However, signal to noise ratio (SNR) of fcMRI is distributed non-uniformly across the cortex leading to systematic biases in network estimates including attenuated correlation in low SNR regions. This bias can be accounted for using test-retest reliability-based attenuation correction. To expand the applicability of this approach in clinical studies, where repeated sessions might not be available, we tested if reliability can be estimated from split data of a single session. Next, we applied this “split session approach” to publically available clinical data to demonstrate that reliability is influenced by age, neuropsychiatric state and scanner type.

Reliability estimated in a longitudinal dataset (n=25, 5 repeated scanning sessions) was compared to a reliability map based on a single split session of this dataset. Next, reliability maps were derived from 6 data sets (26 children and 26 matched adults scanned at a 3T Siemens scanner, 25 subjects with autism and their 25 matched controls scanned at a 3T Siemens scanner, 17 children scanned at a 3T GE scanner and 17 matched children scanned at a 3T Siemens scanner).

The reliability map derived from the “split-session” approach showed a high degree of similarity to the reliability map obtained from 5 sessions (r =0.81, Fig. A). Reliability distribution significantly differed between children and young adults (Fig. B1), autism subjects and normal controls (Fig. B2), and between data collected using a Siemens versus a GE scanner (Fig. B3). The medial temporal lobe, a brain area prone to susceptibility artifacts, consistently showed low reliability in all study groups.

fcMRI reliability distribution is influenced by age, neuropsychiatric state, and scanner type, suggesting the necessity to account for its impact on fcMRI estimates when comparing subject cohorts. The fact that reliability can be estimated based on split data of a single session makes reliability-based attenuation correction a readily available tool to improve the neurophysiological validity of fcMRI data.

Our analyses revealed that reliability distribution is influenced by age, neuropsychiatric status and scanner type, suggesting the necessity to account for its impact on fcMRI estimates in patient control studies.

Mueller, S, Wang, D, Fox, M, Pan, R, Lu, J, Li, K, Buckner, R, Liu, H, Reliability Correction for Functional Connectivity: Clinical Relevance and Implementation.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14012401.html