DTI Association with Working Memory and Speed in Cognitive Network Pathways
Junyu Guo1, John O. Glass1, JungWon Hyun1, Yimei Li1, Conklin Heather1, Lisa Jacola1, Ching-Hon Pui1, Sima Jeha1, and Wilbrun E. Reddick1

1St Jude Children's Research Hospital, Memphis, TN, United States


We investigate the relationship of structural diffusion tensor imaging (DTI) metrics with working memory and decision speed performance in children treated for acute lymphoblastic leukemia (ALL). We built a core neurocognitive network including a central executive network, a salience network, and subcortical cortex based on previous fMRI findings. We generated structural connectivity pathways based on high-resolution DTI data from the human connectome project, and applied those in ALL patients to quantify DTI measures in each pathway. We found that DTI measures in most pathways were significantly associated with working memory and decision speed performance suggesting an essential structural neurocognitive network.


Acute lymphoblastic leukemia (ALL) is the most common malignancy of childhood and adolescence. Even though 5-year survival is above 90% with contemporary treatment, survivors may be at risk for neurocognitive deficits in domains including working memory and decision speed. FMRI studies of executive function have identified central executive and salience networks (1, 2, 3). The central executive network includes dorsolateral prefrontal cortex (DLPFC) and posterior parietal cortex (PPC). The salience network includes anterior cingulate cortex (ACC) and insula (3). Working memory may involve prefrontal cortex (PFC), PPC and subcortical cortex (4, 5, 6). Here we provided a possible structural core cognitive network including a central executive network, a salience network, and subcortical network based on previous fMRI findings. We created structural connectivity pathways between these regions using DTI data from the Human Connectome Project to explore the association of DTI parameters with working memory and decision speed in children treated for ALL.


We evaluated 142 survivors of childhood ALL (age at MR / Cognitive Assessment 9.86±4.48 years; 85 male, 57 female) treated on a chemotherapy-only protocol, Total Therapy Study XVI (NCT00549848), at end of treatment. DTI was acquired with 12 directions and four averages and voxel-wise tensor calculations were performed with SPM8 (fil.ion.ucl.ac.uk/spm/). Working memory was assessed using the Auditory Working Memory subtest from the Woodcock-Johnson Tests of Academic Achievement, 3rd Edition (WJ-III) and the Digit Span Backward subtest from the age-appropriate Wechsler scales. The WJ-III Auditory Working Memory task included listening to a series of digits and words and then reordering the information to repeat first the objects and then the digits in sequential order. The Wechsler Digit Span Backward task required children to repeat a series of numbers in reverse order of presentation. Decision speed was assessed using the WJ-III Decision Speed subtest where subjects were required to make semantic decisions under timed conditions.

Due to the limitation on acquisition time for the clinical examinations of sedated children, low spatial and angular resolution DTI data were acquired, which cannot be used to extract certain pathways such as central executive pathway (link between DLPFC and PPC). We built a core cognitive network including 14 pathways (7 on each side of brain as shown in Fig. 1) using high-resolution DTI data from 81 subjects in the Human Connectome Project. The pathways in each subject were extracted and were registered to a standard MNI space to create pathway regions of interest based on their probabilistic maps using the FMRIB Software Library (FSL). These pathways in MNI space were transformed into each patient space to be used to quantify DTI metrics in each pathway. For higher sensitivity, the third quartile of fractional anisotropy (FA) values and the first quartiles of axial diffusivity (AD) and radial diffusivity (RD) values were used for further analyses instead of mean.

A multiple linear regression model was used to fit the data. The model can be expressed as

$$Y_i = \beta_0 + \beta_1 DTI_i +\beta_2 Age_i +\epsilon_i$$

where $$$Y_i$$$ is a neurocognitive measure for the $$$i^{th}$$$ subject. We included age as a covariate to adjust for the age effect on the neurocognitive measures as well as DTI measures. We applied the false discovery rate correction procedure of Benjamini and Hochberg to correct the p-values for multiple testing. Results were considered significant at the P < 0.05 level.


Table 1 shows FA Q3 and RD Q1 were significantly associated with WJIII decision speed in most of pathways; however, AD Q1 was not significantly associated with WJIII decision speed in any pathway. Table2-3 shows both AD Q1 and RD Q1 were significantly associated with WJIII auditory working memory and digit span backward measures. FA was positively associated, and RD and AD were negatively associated with psychological test scores in which lower scores indicated worse performance.

Discussion / Conclusion

DTI measures were significantly associated with working memory and decision speed scores in most of the pathways. FA was positively associated only with decision speed. These results are consistent with the understanding that signals transmit faster in axons with thicker myelin sheaths, where FA is larger. AD was negatively associated only with working memory, and RD was negatively associated with both decision speed and working memory. Brain injury leads to increases in AD and RD, which is associated with decreased working memory. Our findings provide potential evidence for a structural core neurocognitive network, which could be further used for evaluating and classifying survivors at greatest risk for reduced neurocognitive outcomes following therapy, most notably decreased decision speed and poorer working memory.


This work was supported in part by RO1 CA90246 and Cancer Center Support Grant P30 CA-21765 from the National Cancer Institute at the National institutes of Health, Bethesda, MD, and by the American Lebanese Syrian Associated Charities (ALSAC), Memphis, TN.


1. Bressler S, Memon V. Large-scale brain networks in cognition: emerging methods and principles. Trends in Cognitive Sciences 14 (2010), 277-290.

2. Memon V. Developmental pathways to functional brain networks: emerging principles. Trends in Cognitive Sciences 17 (2013), 627-640.

3. Seeley, W. et al. Dissociable intrinsic connectivity networks for salience processing and executive control. J. Neuosci. 27 (2007), 2349-2356.

4. Frank M. et al. Interactions between frontal cortex and basal ganglia in working memory: a computational model. Cogn Affect Behav Neurosci. 1 (2001), 137-160.

5. Collette F. et al. Exploration of the neural substrates of executive functioning by functional neuroimaging. Neuroscience. 139 (2006), 209-221.

6. O’Reilly R. et al. Making working memory work: a computational model of learning in the prefrontal cortex and basal ganglia. Neural Comput. 18 (2006), 283-328.


Figure 1. Diagram for the structural cognitive networks including 14 pathways.

Table 1. Significance (P values) for DTI measures associated with WJ III decision speed.

Table 2. Significance (P values) for DTI measures associated with WJ III working memory.

Table 3. Significance (P values) for DTI measures associated with digit span backward scores.

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)