Working Memory Task Induced Hemodynamic Abnormalities in Positive and Negative Activations in Patients with Mild Cognitive Impairment

Zachariah Hoell, Medical College of Wisconsin, Milwaukee, United States
Kelly Ristow, Medical College of Wisconsin, Milwaukee, United States
Laura Umfleet, Medical College of Wisconsin, Milwaukee, United States
Alexander Cohen, Medical College of Wisconsin, Milwaukee, United States
Jessica Pommy, Medical College of Wisconsin, Milwaukee, United States
Melissa Lancaster, Medical College of Wisconsin, Milwaukee, United States
Malgorzata Franczak, Medical College of Wisconsin, Milwaukee, United States
Lilly Mason, Medical College of Wisconsin, Milwaukee, United States
Shawn Obarski, Medical College of Wisconsin, Milwaukee, United States
Yang Wang, Medical College of Wisconsin, Milwaukee, United States


Functional magnetic resonance imaging (fMRI) is a technique that utilizes the blood-oxygen-level-dependent (BOLD) effect to indirectly measure neuronal activity. During task performance, specific brain regions exhibit positive responses of the hemodynamic response function (HRF). Recent research has found that aging reduces the amplitude and area-under-the-curve (AUC) of HRF. Moreover, negative activations of task fMRI are frequently observed during cognitive tasks. Nevertheless, changes in HRF and negative activation patterns remain inconclusive in mild cognitive impairment (MCI). This study aims to evaluate the HRF patterns of both positive and negative responses during a working memory (WM) task in patients with MCI.

Participants and Methods:

This study was conducted on 22 clinically diagnosed MCI patients and 32 older adult cognitively unimpaired controls. Task fMRI data were acquired on a research-dedicated 3T Premier using a novel multiband multi-echo BOLD sequence during an n-back WM task (0-, 1-, and 2-back conditions). Preprocessing was performed using our established MEICA pipeline. Using AFNI 3dDeconvolve with CSPLIN and 3dClustSim functions, HRF curves were generated for each significant cluster (p < 0.05). For each HRF, the AUC was determined in the positive and negative regions along with the full-width-half-max, time-to-peak, and peak amplitude. The HRF time-points and summarizing parameters were compared between control and MCI groups via independent t-test. Using linear regression and canonical correlation analysis, the HRF parameters were compared to clinical cognitive assessments including the Trail Making Test (TMT) parts A and B, letter fluency, semantic fluency, and word list delayed recall measures.


Both significant positive and negative activation clusters were detected across 0-, 1-, and 2-back conditions of the WM task for each group. Although the control and MCI showed similar positive activations for the 1-back or 2-back conditions (relative to the 0-back condition), the MCI group showed significantly more negative activation in anterior and posterior default mode network (DMN) regions in the 2-back condition (p < 0.05, corrected). Independent t-tests between the control and MCI displayed significant differences in the peak amplitudes of HRF for 2-back negative anterior DMN (p = 0.019, d = 0.64) and in the negative AUC for the 2-back negative HRF posterior DMN (p = 0.017, d = 0.64). There were no significant differences in the 2-back positive DMN for the peak amplitude (p = 0.498) or the positive AUC (p = 0.809), highlighting the differences in 2-back negative activation. Finally, these results were compared with cognitive measures. Using multiple linear regression, the 2-back posterior DMN HRF positive AUC and negative AUC significantly predicted TMT-A results (R2 = 0.244, p = <0.001); no significant results emerged with other cognitive measures.


While MCI and control groups showed similar positive activation, patients with MCI exhibited significantly more deactivation during the 2-back condition in the anterior and posterior DMN regions. Furthermore, the HRF of negative activation illustrated distinct patterns in MCI compared to controls, which was associated with lower scores on a measure of psychomotor processing speed. The unexpected correlation with TMT-A warrants further research with larger sample sizes to ascertain the repeatability of this result.

Category: MCI (Mild Cognitive Impairment)

Keyword 1: neuroimaging: functional
Keyword 2: mild cognitive impairment
Keyword 3: cerebral blood flow