Frontoparietal Connectivity Neurofeedback Training for Promotion of Working Memory: An fNIRS Study in Healthy Male Participants
Meiyun Xia, Pengfei Xu, Yuanbin Yang, Wenyu Jiang, Zehua Wang, Xiaolei Gu, Mingxi Yang, Deyu Li, Shuyu Li, Guijun Dong, Ling Wang, Daifa Wang
A functional near-infrared spectroscopy-based frontoparietal connectivity neurofeedback training method for cognitive functions promotion P ENGFEI X U , Z EHUA W ANG , M EIYUN X IA , X IAOLEI G U , M INGXI Y ANG , D EYU L I , S HUYU L I , , G UIJUN D ONG , L ING W ANG , AND D AIFA W ANG School of Biological Science and Medical Engineering, Beihang University, Beijing, 100083, P. R. China College of Computer Science, Sichuan Normal University, Chengdu, 610101, P. R. China Beijing Advanced Innovation Center for Biomedical Engineering, Beihang University, Beijing, 100083, P. R. China State Key Laboratory of Virtual Reality Technology and System, Beihang University, Beijing, 100083, P. R. China Shandong Sport University, Jinan, 250102, P. R. China *[email protected]; [email protected]
Pengfei Xu, Zehua Wang and Meiyun Xia contributed equally to this paper.
Abstract:
An innovative approach which can promote cognitive functions effectively and efficiently is an urgent need for healthy elderly and patients with cognitive impairment. In this study, we proposed a novel functional near-infrared spectroscopy (fNIRS)-based frontoparietal functional connectivity (FC) neurofeedback training paradigm related to working memory. Compared with conventional cognitive training studies, we chose the frontoparietal network, a key brain region for cognitive function modulation as neurofeedback, resulting in strong targeting effect. In the experiment, ten participants received three 20 min cognitive training sessions with fNIRS-based frontoparietal FC as neurofeedback, and the other ten participants served as the normal control (NC) group. Frontoparietal FC was significantly increased in the tested group ( p = 0.005) and the cognitive functions (memory and attention) were significantly promoted compared to the NC group. Follow-up evaluations indicated that the training effect can last for over half a month. The proposed method shows great potential to be developed as a fast, effective and widespread training approach for cognitive functions enhancement and rehabilitation applications. © 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
Cognitive training has become an attracting approach to promote the cognitive functions for healthy people or patients with various neurodevelopmental and neurodegenerative diseases [1-4]. However, conventional cognitive training is usually lengthy, taking several months. Benefit from its ability of modulating the targeted brain regions, neurofeedback techniques emerge and provide a novel way of promoting the effectiveness and applicability of cognitive training. Based on the real-time visualization of the neurophysiological status, neurofeedback techniques enable the participant to directly attempt to regulate his/her own brain activity. Furthermore, the participant may upregulate or downregulate the neural activation of targeted brain regions, and may further improve the outcome of cognitive training [5, 6]. Prior publication findings have indicated that providing individuals with real-time feedback of their own brain activities can help them learn how to control the activation of specific brain regions [2, 7-9]. Currently, the primary neuroimaging modalities used to provide neurofeedback include electroencephalography (EEG), functional magnetic resonance imaging (fMRI) and functional near-infrared spectroscopy (fNIRS), etc. EEG has an inherently high-temporal resolution and a low cost. It can measure real-time brain activation information and has been used as the neurofeedback signal [10]. However, EEG is difficult to locate the specific brain regions during regulation owing to the relatively low-spatial resolution. In turn, fMRI is able to capture noninvasively the blood oxygen level-dependent (BOLD) signal changes in deep brain regions and provide near-real-time hemodynamic neurofeedback with high-spatial resolution over the whole brain. Due to this advantage, fMRI has been more and more utilized in the neurofeedback studies. For example, patients with Parkinson’s disease and attention deficit hyperactivity disorder (ADHD) successfully learned how to increase the brain activities in their motor-related or attention-related cortices [11, 12]. Nevertheless, there is one major limitation. The fMRI measurement is expensive and associated with a stringent imaging environment that limits the designs of neurofeedback training experiments. As a result, the fNIRS, a brain imaging modality that provides relatively good spatial resolution, low requirement for application environment and relatively low cost, has attracted attention [8, 9, 13-16]. fNIRS is able to locate specific cortical regions and provide simultaneous BOLD signals over the whole brain cortex. Additionally, fNIRS is portable and facilitates the experimental design, and it is more convenient to be used to evaluate patients with motion instabilities (e.g., adults and children with ADHD) [14] compared to fMRI or EEG which are more sensitive to motion artifacts. Owing to its simplicity and low cost [15, 16], fNIRS is fit for repetitive measurements and is thus a practical and convenient tool for neurofeedback applications in the practical clinical and rehabilitation environment. Previous studies have shown that fNIRS-based neurofeedback training was able to manipulate the activation of the lateral orbitofrontal cortex and prefrontal cortex in healthy participants [13, 17], and the motor cortex in healthy participants and in patients after stroke [8, 9]. Recently, connectivity neurofeedback is developing. It is well known that the brain functions of human beings are coherently controlled by multiple brain regions called brain networks [18-20], and many cognitive functions as well as psychiatric [21-24] and neurodegenerative [25, 26] diseases are closely related to brain networks. For example, memory is closely related to the frontoparietal and the default mode networks, while attention is related to the dorsal attention network [20, 27-29]. Fukuda et al. [30] and Yamashita et al. [5] revealed that fMRI-based connectivity neurofeedback training can induce the aimed directional changes in functional connectivity (FC) between the left primary motor cortex and the left lateral parietal cortex. Kim et al. also reported that the brain activity-plus-connectivity neurofeedback based on fMRI helped the heavy smokers to more effectively regulate their psychological functions [31]. In the present study, we proposed a novel fNIRS-based frontoparietal FC neurofeedback training paradigm related to working memory (WM), and investigated whether the proposed paradigm can manipulate the frontoparietal FC and further effectively promote cognitive functions by using fNIRS signals. If the answer provided by this study is “yes”, benefit from the advantages of fNIRS mentioned before, a low-cost, easy to use, potentially portable and robust to motion connectivity neurofeedback training strategy can probably be developed. In the proposed paradigm, a Sternberg working memory task [32-34] was executed at first and fNIRS signals of frontal and parietal regions were acquired simultaneously. After that, a score computed from the strength of frontoparietal FC measured by fNIRS was displayed as feedback. During the neurofeedback tasks, the participants were asked to make the score as high as possible. In the paradigm, WM was adopted due to its critical involvement in the execution of cognitive tasks by the brain, and the frontoparietal FC was chosen as the modulation target in our neurofeedback strategy is due to its close relationship with WM. Previous studies have revealed that WM-related cognitive functions are closely related to the frontoparietal brain network [20, 27, 35-40]. For example, patients with schizophrenia, Alzheimer’s disease, or other related diseases encountered cognitive declines in attention and WM that were found to be closely related to the frontoparietal FC [27, 35, 36]. In this study, 20 young, healthy participants were enrolled in the experiment in which ten participants received and the other ten participants did not receive frontoparietal FC as neurofeedback during three 20 min cognitive training sessions. Oxyhemoglobin (HbO) signals were recorded by an fNIRS device from the prefrontal and parietal cortices and used to calculate the frontoparietal FC. The primary behavioral outcome of the proposed cognitive training was evaluated based on behavioral testing. The results of our study showed that the proposed fNIRS-based connectivity neurofeedback training paradigm can significantly upregulate the frontoparietal FC and further promote related cognitive functions. This study for the first time validates the feasibility and effectiveness of fNIRS brain connectivity neurofeedback training to improve the cognitive functions. The results have important potential applications for cognitive enhancement in healthy people and cognitive improvement in patients.
2. Materials and methods
Twenty adults (all male, aged 23.5 ± Fig. 1. Neurofeedback setting. (a)
A 42-channel fNIRS probe set covering the prefrontal, parietal, and temporal regions. The red circles indicate fNIRS sources, blue rectangles indicate fNIRS detectors, and black lines indicate fNIRS channels. The sources in green color are placed on Fz and Pz positions according to the international 10–20 system. (b)
Frontoparietal FC.
The fNIRS signals were measured by the NirScan system (Huichuang, China) at the wavelengths of 740 nm and 850 nm at a sampling rate of 13 Hz. As shown in Fig. 1, 16 light sources and 16 detectors were distributed on the scalp (separating distance = 3 cm) on positions according to the international 10–20 system [41, 42]. In total, 42 fNIRS channels were used which covered the prefrontal (1–22 channels), parietal (23–40 channels), and temporal (41, 42 channels) cortices (see Fig. 1(a)).
As shown in Fig. 2(a), the experiment was composed of six training sessions, one baseline estimation session (referred to as Baseline), three fNIRS-based connectivity neurofeedback training sessions (referred to as CNF training, including three sessions referred to as T1, T2, and T3), and two follow-up evaluation training sessions (referred to as Evaluation, including two sessions referred to as WEEK1 and WEEK3). Baseline and CNF training was conducted within the first week, and WEEK1 and WEEK3 were conducted at one and three weeks after the CNF training, respectively. There were 25 trials (14–15 min) in each training session. After each training session, the participant received a 3 min resting-state fNIRS measurement during which the participant was asked to stay relaxed and to watch the center of the screen. During Baseline and Evaluation, all the participants received cognitive training without connectivity neurofeedback. During the CNF training, fNIRS-based connectivity neurofeedback was provided to the participants in the tested group in the form of a feedback score appearing on the screen. The feedback score represented the strength of the frontoparietal FC. The participants needed to try their best to increase the feedback score during the experiment, and they were
Fig. 2. Experimental protocol. (a)
Experimental workflow, including one baseline estimation session (Baseline) and three fNIRS-based connectivity neurofeedback training sessions (CNF training, including three sessions referred to as T1, T2, and T3) within one week, followed by two follow-up evaluation sessions (Evaluation) after one (WEEK1) and three weeks (WEEK3). Resting-state brain activity (rsfNIRS) was measured after each training session. Participants performed behavioral testing after Baseline, T3, WEEK1, and WEEK3. (b)
A trial based on the Sternberg task for the tested group (with fNIRS-based connectivity neurofeedback) during the CNF training. informed before the experiment that their monetary reward was positively related to the feedback scores. The entire experimental protocol of the NC group was the same as that of the tested group with the exception that there was no connectivity neurofeedback during the CNF training. The cognitive training paradigm was based on the verbal WM task (i.e., the Sternberg task). As shown in Fig. 2(b), a trial of the verbal WM task is composed of three phases: a memory phase (2 s), a retention phase (10 s), and an inquiry phase (2 s). In the memory phase, six nonrepetitive letters appeared on the screen and the participant was asked to memorize them within 2 s. The letters then disappeared and the retention phase began. A “+” sign appeared in the center of the screen during which participant needed to keep the six nonrepetitive letters in memory for 10 s. The inquiry phase followed. A random letter appeared in the center of the screen and the participant was asked to a) judge whether this letter was within the six nonrepetitive letters in the memory phase, and b) choose “yes/no” (“y/n” on the keyboard in front of the participant) according to his/her judgment. For the participants in the tested group, the feedback score of the participants appeared on the screen within 2 s after the inquiry phase in each trial during the CNF training. A resting period of 20 s followed before the onset of the subsequent trial.
The feedback score in each trial was calculated from the HbO signals measured from the prefrontal and parietal cortices during the memory and retention phases (time window: 0–12 s). Regarding the hemodynamic delay, we delayed the calculation window by 2 s, i.e., the time-shifted window was 2–14 s, which is consistent with previous neurofeedback studies [17, 41, 28]. Specifically, in order to reduce the physiological noise caused by heart beating, respiration, and other physiological processes, the recorded HbO signals were bandpass filtered with cut-off frequencies of 0.01 and 0.2 Hz [43]. Region-averaged HbO signals corresponding to the prefrontal cortex ( (cid:1876) (cid:3002) , averaged HbO signal from channel 1 to 22) and parietal cortex ( (cid:1876) (cid:3003) , averaged HbO signal from channel 23 to 40) were then calculated, and the Pearson’s correlation ( r ) between (cid:1876) (cid:3002) and (cid:1876) (cid:3003) was calculated as follows, ( ) ( )( ) ( ) r K A A B BkK A A B Bk x k x x k xx k x x k x == − − = − − , (where K is the number of values in (cid:1876) (cid:3002) and (cid:1876) (cid:3003) , (cid:1876) (cid:3002) (cid:4666)(cid:1863)(cid:4667) and (cid:1876) (cid:3003) (cid:4666)(cid:1863)(cid:4667) are the k th values in (cid:1876) (cid:3002) and (cid:1876) (cid:3003) , and (cid:1876) (cid:3002) and (cid:1876) (cid:3003) are the mean values of (cid:1876) (cid:3002) and (cid:1876) (cid:3003) , respectively. To observe the changes of the frontoparietal FC, a Fisher r - z transform was performed on the calculated Pearson’s correlation, and the z -value between (cid:1876) (cid:3002) and (cid:1876) (cid:3003) was calculated based on, rz r += − . (Using the z -value calculated above, the feedback score of the i th trial was calculated as,
50( 3 ) (0 100)3 i basei i z SD zScore ScoreSD + −= ≤ ≤ . (Herein, (cid:1878) (cid:3036) is the z -value of the frontoparietal FC in the i th trial, and (cid:1878) (cid:3029)(cid:3028)(cid:3046)(cid:3032) and (cid:1845)(cid:1830) are respectively the mean and standard deviation of the z -values of the frontoparietal FC during Baseline. The purpose for the calculation of the feedback score was to provide the participant with the following information: the baseline performance corresponded to 50 scores, and a monetary reward was provided if the frontoparietal FC of the current trial was higher than the baseline level (i.e., when the feedback score exceeded 50). Scores which dropped below 0 or exceeded 100 were kept at 0 or 100, respectively. The online signal processing and all visual presentations in the experimental protocol were performed using MATLAB (R2016a, MathWorks, Natick, MA, USA), and the Psychtoolbox was used for visual presentations. The effects of the proposed neurofeedback training on cognitive functions were evaluated by behavioral testing. The classical n -back testis designed to evaluate the changes in working memory [44, 45]. As a result, the n -back test was carried out in our experiment before and after the CNF training and after WEEK1 and WEEK3 (see Fig. 2(a)). Given that the participants were healthy young adults, the 3-back test ( n = 3) was used. The accuracy and reaction time of the participant in the 3-back test were used as the primary outcomes of behavioral testing. Additionally, to investigate whether the proposed neurofeedback training affected other cognitive abilities beyond WM, the psychomotor vigilance test (PVT) and the color-word stroop test (CWST) were also conducted. PVT evaluates the ability of fixing the attention and CWST is a response inhibition test that evaluates the ability of the participant to inhibit inappropriate responses under certain conditions [5]. Similar to WM, the above two cognitive abilities are also modulated by the frontoparietal brain network. The reaction times of PVT and CWST were used to evaluate the possible migratory aptitude of the proposed neurofeedback training on other cognitive abilities. The experimental procedures of behavioral testing were as follows. During the test, a series of letters appeared on the screen in sequence. Each letter lasted 2 s and the interval between the appearances of two sequential letters was 1 s (a “+” sign appeared on the screen during this interval). From the instant the ( n+1 ) th letter appeared and subsequently, the participant was asked to judge whether the current letter was the same as the forward n th letter which appeared on the screen. In the 3-back task, n = 3, and the participant began to judge from the instant the th (( n+1) th ) letter appeared. Specifically, the participant assessed whether the th letter was the same as the st letter (4 - n = 1, where n = 3), and whether the th letter was the same as the nd letter (5 - n = 2, where n = 3), and so on. If two letters were the same, the participant pressed “1” on the keyboard; otherwise he/she pressed “2”. Twenty five trials were executed, and the mean accuracy and mean reaction time were calculated as the outcomes of the n -back test. When a trial started, there was a white “+” sign at the center of the black background. The participant was asked to keep watching the screen to wait for the stimulus. When the stimulus appeared, the white “+” sign suddenly changed to red, and the participant needed to press button “1” promptly. If the participant successfully pressed the button “1” within 3 s after the appearance of the stimulus, the red “+” sign changed to its original white color immediately after the action. Otherwise, the red “+” sign automatically changed to a white color after 3 s. In our experiment, each participant was asked to attend to 10 PVT trials, and the between-trial intervals ranged from 5 to 15 s. The time interval from the appearance of the stimulus to the instant the button “1” was pressed was the reaction time of each trial. The final PVT outcomes were the average values of the 10 trials.
One of the three words, namely, “Red”, “Yellow”, and “Green” appeared on the screen randomly and in sequence for 2 s. If the color of the word matched the meaning of the word (e.g., the color of the word “Red” was red), the participant pressed button “1”, else (e.g., when the color of the word “Red” was green) he/she pressed button “2”. In our experiment, each participant took part in 30 trials, and the averaged reaction time (the time from the appearance of the word to the instant the button being pressed) over 30 trials was used as the result of CWST.
3. Statistical analyses
Paired-sample t -tests (one-tailed) were used to evaluate whether the behavioral performance (behavioral indices) of the tested group changed significantly after the CNF training, and how the effects were maintained after WEEK1 and WEEK3 compared with Baseline. Independent-sample t -tests (one-tailed) were adopted to evaluate whether there were significant intergroup (between the tested and NC groups) differences on the behavioral performance before and after the CNF training, and how these differences changed or remained after WEEK1 and WEEK3. To evaluate the modulation effect of the CNF training on the frontoparietal FC, paired-sample t -tests (one-tailed) were applied to evaluate whether the frontoparietal FC (z-value) of the tested group significantly changed in T3, WEEK1 and WEEK3 compared with Baseline. Independent-sample t- tests (one-tailed) were utilized to evaluate whether there were significant intergroup differences on the frontoparietal FC (z-value) measured during Baseline, T3, WEEK1 and WEEK3. To further investigate the modulating effect of the CNF training on bilateral frontoparietal FCs, the changes of the left and right frontoparietal FCs were analyzed. Additionally, changes in the resting-state frontoparietal FC—that reflects more accurately the intrinsic changes of the cognitive ability—were analyzed using the same statistical methods as those applied on the task-state data mentioned above. Additionally, to investigate whether the CNF training affected the FCs between the untargeted brain regions, HbO signals measured from the temporal cortex were analyzed. The changes of the intragroup (within the tested group) and intergroup bilateral frontotemporal and temporal–parietal FCs were calculated and analyzed using the same statistical methods adopted in the analysis of the frontoparietal FC mentioned above. In all statistical analyses, a p value less than 0.05 was considered as significant. Results
For the participants in the tested group, the n -back accuracies (mean±SD) significantly increased from 65±11% to 81±10% ( p = 0.00003), and the n -back reaction time (mean±SD) significantly decreased from 1.35±0.20 s to 1.14±0.18 s ( p = 0.00001) from Baseline to post-CNF training. After WEEK1, the n -back accuracies decreased to 75±9% and the n -back reaction time increased to 1.23±0.19 s. Both values were significantly different from the values measured in Baseline (accuracy: p = 0.0001, reaction time: p = 0.0001). After WEEK3, both the n -back accuracies and reaction time showed no significant changes from the baseline level (accuracy: p = 0.07, reaction time: p = 0.23). There were no significant differences between the tested and NC groups on either the n -back accuracy ( p = 0.15) or n -back reaction time ( p = 0.49) before the CNF training. Fig. 3(a) and (b) show the relative post-CNF training changes with respect to Baseline (relative changes = post-training value-baseline value) for n -back accuracy and reaction time. After CNF training, the relative changes of n -back accuracy increased significantly ( p = 0.0003), and those of the n -back reaction time decreased significantly ( p = 0.0005) in the tested group compared to the NC group. After WEEK1, the intergroup differences became marginally significant for n -back accuracy ( p = 0.06) and insignificant for n -back reaction time ( p = n -back testing results ( n -back accuracy: p = 0.098, n -back reaction time: p = 0.41). Within the tested group, the PVT reaction time decreased significantly ( p = 0.0009) from 0.40±0.06 s to 0.38±0.05 s from Baseline to the post-CNF training level. This decrease became insignificant after WEEK1 ( p = 0.07) and WEEK3 ( p = 0.39). There was no significant intergroup difference ( p = 0.33) on the PVT reaction time before the CNF training. After the CNF training, the PVT reaction time decreased at a significantly faster rate ( p = 0.039) in the tested group compared to the NC group (see Fig. 3(c)). The significance of these intergroup differences disappeared after WEEK1 ( p = 0.16) and WEEK3 ( p = 0.44). The CWST reaction time of the tested group did not change significantly from Baseline to the post-CNF training level. As shown in Fig. 3(d), there were no significant differences between the tested and NC groups on the relative changes of CWST reaction time after CNF training. Fig. 3. Behavioral testing results after CNF training. Results on relative changes of (a) n -back accuracy, (b) n -back reaction time, (c) PVT reaction time, and (d)
CWST reaction time post-CNF training. **: p < 0.01, *: p < 0.05, n.s. : not significant. Fig. 4 shows the changes in z -value of frontoparietal FC in the tested and NC groups throughout the experiment. The frontoparietal FC ( z -value) of the tested group increased significantly from 0.98±0.31 to 1.27±0.34 from Baseline to T3 ( p = 0.005). The significance of such increase dropped on WEEK1 ( p = 0.017), and disappeared on WEEK3 ( p = 0.267). There was no significant intergroup difference on the frontoparietal FC ( z -value) measured at Baseline ( p = 0.13). On T3, the frontoparietal FC (z-value) in the tested group were significantly higher than those in the NC group ( p = 0.015). The significance of these intergroup differences decreased ( p = 0.039) on WEEK1 and disappeared ( p = 0.474) on WEEK3. We compared separately the changes in the bilateral frontoparietal FCs ( z -value). It was found that within the tested group, the left and right frontoparietal FCs ( z -value) both significantly increased from Baseline to T3 and the increase was more significant on the left side (left: p = 0.002, right: p = 0.018). At Baseline, there were no significant intergroup differences on the z -values of bilateral frontoparietal FCs (left: p = 0.321, right: p = 0.438). On T3, the bilateral frontoparietal FCs ( z -value) of the tested group were both significantly higher than those in the NC group. Furthermore, the significance of the intergroup differences was higher on the left side (left: p = 0.008, right: p = 0.011). Fig. 4. Z -values of frontoparietal FC before and after CNF training. **: p < 0.01, *: p < 0.05, n.s. : not significant. As shown in Fig. 5, the changes of the n -back accuracy and those of the frontoparietal FC ( z -value) within the tested group before and after the CNF training are compared. As shown, the changing trends of these two variables are fairly consistent. Fig. 5. Changing trends of n -back accuracy and the z -value of the frontoparietal FC within the tested group before and after CNF training. The frontoparietal FC (z-value) at the resting-state within the tested group did not significantly increase after the CNF training. Fig. 6 shows the z -values of the bilateral frontoparietal FCs at the resting-state before and after the CNF training. It is found that the post-training resting-state z -values of the left frontoparietal FC were (marginally) significantly higher in the tested group compared to those of the NC group ( p = 0.076). Fig. 6. Z -values of the resting-state bilateral frontoparietal FCs before and after the CNF training. The regulation effect of CNF training on the frontoparietal FC may radiate to other related FCs through the brain networks. Therefore, HbO signals were recorded from the temporal cortex in all experimental sessions to investigate the possible diversion effect of the CNF training regulation on the bilateral frontotemporal and temporal–parietal FCs. Results showed that within the tested group, neither the frontotemporal nor the temporal–parietal FCs ( z -value) significantly changed, while the right temporal–parietal FC ( z -value) were (marginally) significantly increased ( p = 0.069) after the CNF training (on T3). Furthermore, there were no significant intergroup differences on either the frontotemporal or temporal–parietal FCs ( z -value) after the CNF training, while the right temporal–parietal FC ( z -value) was (marginally) significantly higher ( p = 0.091) in the tested group compared to the NC group during T3.
5. Discussion
In this study, we proposed a novel fNIRS-based frontoparietal FC neurofeedback training paradigm related to WM. We investigated whether the proposed method was able to effectively regulate the frontoparietal FC and promote the cognitive functions. Twenty healthy participants took part in the experiment. Ten participants received fNIRS-based connectivity neurofeedback in the cognitive training, while the other 10 participants did not. Results showed that the post-training frontoparietal FC was significantly upregulated and the related cognitive performance was significantly promoted with short training time in the case of the participants who received fNIRS-based connectivity neurofeedback. Moreover, there were significant post-training differences among participants who received/did not receive fNIRS-based connectivity neurofeedback not only regarding the frontoparietal FC but also cognitive performance. These results demonstrated that the proposed fNIRS-based connectivity neurofeedback training is a promising approach for the upregulation of the frontoparietal brain network of healthy people and for the improvement of their cognitive performance.
Comparison of the bilateral frontoparietal FCs showed that although both sides of the frontoparietal FC in the tested group were significantly increased after the CNF training, the increase on the left side was more than that on the right side. Several studies have reported significant bilateral neural activation in both prefrontal and parietal cortices during the encoding, maintenance, and retrieval of the WM information [46, 47]. A recent study conducted by Baker et al. further revealed that compared with the visuospatial WM tasks that relied mainly on the neural activation of right prefrontal cortex, the verbal WM tasks mainly activated the left prefrontal cortex [48]. In another early study, D'Esposito et al. analyzed the results from 20 working memory-related fMRI/positron emission tomography (PET) studies and found that spatial WM tasks (similar to those in Baker’s study) exhibited greater activations in the right prefrontal cortex, while many nonspatial WM tasks (similar to those in this study) exhibited greater activation in the left prefrontal cortex [49]. Our results showed that the left frontoparietal FC was more enhanced by the proposed verbal WM task-related-paradigm, and the outcomes were consistent with previous findings. We were excited to observe that after the CNF training, the left frontoparietal FC at the resting-state was (marginally) significantly higher ( p = 0.076) in the tested group than in the NC group. It was also observed in some previous studies that long-term cognitive training may have changed the resting-state brain networks. For example, it was reported that after a period of WM-task-based cognitive training, the resting-state frontoparietal FC of the participants was altered [50, 20]. Compared with the changes of task-state brain connectivity, the changes of the intrinsic resting-state brain connectivity would reflect more adequately the effectiveness of cognitive training in the improvement of cognitive functions. Our results suggested that the proposed fNIRS-based connectivity neurofeedback training may also change the resting-state frontoparietal brain network. Simple cognitive training has been argued for the lack of an effective diversion effect from targeted to untargeted cognitive regions [3]. To investigate whether the proposed fNIRS-based connectivity neurofeedback training yielded a diversion effect, two other cognitive abilities regulated by the frontoparietal FC, i.e., the attention focusing ability and the inappropriate response inhabitation ability were tested before and after the CNF training. Results showed that the attention focusing ability (PVT reaction time) of the tested group was significantly improved after the CNF training. This suggested that the proposed paradigm yield a good diversion effect on the untargeted cognitive regions. The prefrontal and parietal cortices are known to be related to many other functional regions in the brain. As a result, we inferred that many related regions and their FCs may also be affected by the proposed neurofeedback training. Previous studies have reported that besides the prefrontal and parietal cortices, the temporal cortex also played an important role in the regulation of the process of WM [51, 52]. Therefore, in our study, the changes of the frontotemporal and temporal–parietal FCs were analyzed before and after the CNF training. Results showed that there were no significant changes on the frontotemporal FC within the tested group after the CNF training. The right temporal–parietal FC exhibited an increasing trend and was (marginally) significantly ( p = 0.069) increased post-CNF training. The reason for these results may be attributed to the relatively short duration (three 20 min training sessions) of the proposed CNF training which was not long enough to induce significant changes on the indirectly-regulating (untargeted) brain regions. Similar to our results, Fukuda et al. conducted an fMRI-based connectivity neurofeedback study and found that their neurofeedback training (4 days) resulted in significant changes in the FC between targeted regions but led to insignificant changes in the brain networks beyond the targeted regions [30]. To-this-date, it is still uncertain whether the connectivity neurofeedback training affects the FCs beyond the targeted regions. We consider that it does, and infer that the main reason why Fukuda’s and our study did not observe significant and positive changes is likely attributed to the relatively short durations of training. To investigate the retainability of the proposed paradigm, two follow-up evaluations, WEEK1 and WEEK3 were conducted on all participants at one and three weeks after the CNF training. Results showed that through only three 20 minutes fNIRS-based frontoparietal connectivity neurofeedback training sessions, the training effects can be maintained for over half a month. In the future, if the proposed paradigm is applied as routine training of cognitive enhancement, we optimistically believe that with the increase of training times (frequency), the training effect can be able to last for several months or even longer. This hypothesis would be studied in our proceeding research.
6. Conclusions
In this study, we proposed a novel WM task-related, fNIRS-based frontoparietal connectivity neurofeedback training paradigm, and verified its capacity in manipulating the frontoparietal FC and improving the cognitive abilities within limited training time. The results showed that the proposed method could effectively upregulate the frontoparietal brain network and promote memory cognitive abilities through only three 20-minutes training. The results further indicated that the regulation and promotion on cognitive performance induced by the proposed paradigm not only affect the targeted cognitive abilities but also exert diversion effects on other cognitive abilities such as attention. With further validations on different populations and brain networks, the proposed method is promising to be developed as a fast, effective and widely used training tool for cognitive enhancement and rehabilitation in the future.
Funding
National Natural Science Foundation of China (NSFC) (61675013, 61101008, 31771071, 81972160, 81622025); National Key Research and Development Plan (2018YFC2001700); Chinese Ministry of Justice Key Research Project on the Detoxification Theory in Justice Administration (19ZD07).
Acknowledgments
Disclosures
The authors declare no conflicts of interest.
References S. L. Willis, S.L. Tennstedt, M. Marsiske, K. Ball, J. Elias, K. M. Koepke, J. N. Morris, G. W. Rebok, F. W. Unverzagt, A. M. Stoddard and E. Wright, “Long-term effects of cognitive training on everyday functional outcomes in older adults,” J. Am. Med. Assoc. (23):2805 (2006). 2.
B. M. Hampstead, A. Y. Stringer, R. F. Stilla, M. Giddens, and K. Sathian, “Mnemonic strategy training partially restores hippocampal activity in patients with mild cognitive impairment,” Hippocampus (8):1652–1658 (2012). 3. D. C. Park and G. N. Bischof, “The aging mind: neuroplasticity in response to cognitive training,” Dialogues Clin. Neurosci. (1):109-119 (2013). 4. F. D. Wolinsky, W. M. W. Vander, H. M. Bryant, M. P. Jones, M. M. Dotson, and L. Jerson, “A randomized controlled trial of cognitive training using a visual speed of processing intervention in middle aged and older adults,” PLoS ONE (5):e61624 (2013). 5. A. Yamashita, S. Hayasaka, M. Kawato, and H. Imamizu, “Connectivity neurofeedback training can differentially change functional connectivity and cognitive performance,” Cereb. Cortex (10):4960–4970 (2017). 6. A. C. Ehlis, B. Barth, J. Hudak, H. Storchak, L. Weber, A. C. S. Kimmig, B. Kreifelts, T. Dresler, and A. J. Fallgatter, “Near ‐ infrared spectroscopy as a new tool for neurofeedback training: applications in psychiatry and methodological considerations,” Jpn. Psychol. Res. (4): 225–241 (2018). F. Scharnowski, C. Hutton, O. Josephs, N. Weiskopf, and G. Rees, “Improving visual perception through neurofeedback,” J. Neurosci. (49):17830–17841 (2012). 8. M. Mihara, N. Hattori, M. Hatakenaka, H. Yagura, T. Kawano, T. Hino, and I. Miyai, “Near-infrared spectroscopy-mediated neurofeedback enhances efficacy of motor imagery-based training in poststroke victims: a pilot study,” Stroke (4):1091–1098 (2013). 9. S. E. Kober, G. Wood, J. Kurzmann, E. V. C. Friedrichab, M. Stangla, T. Wippela, A. Väljamäe, and C. Neuperab, “Near-infrared spectroscopy based neurofeedback training increases specific motor imagery related cortical activation compared to sham feedback,” Biol. Psychol. :21–30 (2014). 10. U. Leins, G. Goth, T. Hinterberger, C. Klinger, N. Rumpf and U. Strehl, “Neurofeedback for children with ADHD: a comparison of SCP and theta/beta protocols,” Appl. Psychophysiol. Biofeedback :73–88 (2007). 11. L. Subramanian, J. V. Hindle, S. Johnston, M. V. Roberts, M. Husain, R. Goebel, and D. Linden, “Real-time functional magnetic resonance imaging neurofeedback for treatment of Parkinson's disease,” J. Neurosci. (45):16309–16317 (2011). 12. J. Levesque, M. Beauregard, and B. Mensour, “Effect of neurofeedback training on the neural substrates of selective attention in children with attention-deficit/hyperactivity disorder: a functional magnetic resonance imaging study,” Neurosci. Lett. (3):216–221 (2006). 13.
K. Li, Y. Jiang, Y. Gong, W. Zhao, Z. Zhao, X. Liu, K. M. Kendrick, C. Zhu, and B. Becker, “Functional near-infrared spectroscopy-informed neurofeedback: regional-specific modulation of lateral orbitofrontal activation and cognitive flexibility,” Neurophotonics (2):025011 (2019). 14. F. Irani, S. M. Platek, S. Bunce, A. C. Ruocco, and D. Chute, “Functional near infrared spectroscopy (fNIRS): an emerging neuroimaging technology with important applications for the study of brain disorders,” Clin. Neuropsychol. (1):9–37 (2007). 15. H. Atsumori, M. Kiguchi, A. Obata, H. Sato, T. Katura, K. Utsugi, T. Funane, and A. Maki, “Development of a multi-channel, portable optical topography system,” Int. Conf. IEEE Eng. Med. Biol. Soc. (2007). 16.
K. Sagara, K. Kido, and K. Ozawa, “Portable single-channel NIRS-based BMI system for motor disabilities' communication tools,” Conf. Proc.: Ann. Int. Conf. IEEE Eng. Med. Biol. Soc. IEEE Eng. Med. Biol. Soc. Conf. :602–605 (2009). 17.
S. M. H. Hosseini, M. Pritchard-Berman, N. Sosa, A. Ceja, and S. R. Kesler, “Task-based neurofeedback training: A novel approach toward training executive functions,” NeuroImage :153–159 (2016). 18.
M. Mesulam, “From sensation to cognition,” Brain 121: 1013–1052 (1998). 19.
S. M. Smith, P. T. Fox, K. L. Miller, D. C. Glahn, P. M. Fox, C. E. Mackay, N. Filippini, K. E. Watkins, R. Toro, A. R. Laird, and C. F. Beckmann, “Correspondence of the brain’s functional architecture during activation and rest,” PNAS USA (31): 13040–13045 (2009). 20.
D. D. Jolles, M. A. Van Buchem, E. A. Crone, and S. A. R. B. Rombouts, “Functional brain connectivity at rest changes after working memory training,” Hum. Brain Mapp. (2):396–406 (2013). 21. S. J. Broyd, C. Demanuele, S. Debener, S. K. Helps, C. J. James, and E. J. S. Sonuga-Barke, “Default-mode brain dysfunction in mental disorders: A systematic review,” Neurosci. Biobehav. Rev. (3):279–296 (2009). 22. A. Fornito, A. Zalesky, and M. Breakspear, “The connectomics of brain disorders,” Nat. Rev. Neurosci. (3):159–172 (2015). 23. K. Lu, G. Xu, W. Li, C. Huo, Q. Liu, Z. Lv, Y. Wang, Z. Li, and Y. Fan, “Frequency-specific functional connectivity related to the rehabilitation task of stroke patients,” Medical Physics. 46:1545-1560 (2019). 24.
X. Wang, Y. Yu, W. Zhao, Q. Li, X. Li, S. Li, C Yin, and Y. Han, “Altered Whole-Brain Structural Covariance of the Hippocampal Subfields in Subcortical Vascular Mild Cognitive Impairment and Amnestic Mild Cognitive Impairment Patients,” Frontiers in Neurology. , 342 (2018). 25. X. Li, Z. Zhu, W. Zhao, Y. Sun, D. Wen, Y. Xie, X. Liu, H. Niu, and Y. Han, “Decreased resting-state brain signal complexity in patients with mild cognitive impairment and Alzheimer's disease: a multiscale entropy analysis,” Biomed. Opt. Express. (4):1916-1929 (2018). 26. H. Niu, Z. Zhu, M. Wang, X. Li, Z. Yuan, Y. Sun, Y. Han, “Abnormal dynamic functional connectivity and brain states in Alzheimer’s diseases: functional near-infrared spectroscopy study,” Neurophoton. (2), 025010 (2019). 27. B. J. He, A. Z. Snyder, J. L. Vincent, A. Epstein, G. L.Shulman, and M. Corbetta, “Breakdown of functional connectivity in frontoparietal networks underlies behavioral deficits in spatial neglect,” Neuron (6):905–918 (2007). 28. G. J. Thompson, M. E. Magnuson, M. D. Merritt, H. Schwarb, W. Pan, A. McKinley, L. D. Tripp, E. H. Schumacher, and S. D. Keilholz, “Short-time windows of correlation between large-scale functional brain networks predict vigilance intraindividually and interindividually,” Hum. Brain Mapp. (12):3280–3298 (2013). 29. C. Liu, Z. Chen, T. Wang, D. Tang, G. Hitchman, J. Sun, X. Zhao, L. Wang, and A. Chen, “Predicting stroop effect from spontaneous neuronal activity: a study of regional homogeneity,” PLoS ONE (5):e0124405 (2015). 30. M. Fukuda, Y. Ayumu, K. Mitsuo, and I. Hiroshi, “Functional MRI neurofeedback training on connectivity between two regions induces long-lasting changes in intrinsic functional network,” Frontiers Hum. Neurosci. , 160 (2015). D. Y. Kim, S. S. Yoo, M. Tegethoff, G. Meinlschmidt and J. H. Lee, “The inclusion of functional connectivity information into fMRI-based neurofeedback improves its efficacy in the reduction of cigarette cravings,” J. Cognit. Neurosci. (8):1552–1572 (2015). 32. S. Sternberg, “The discovery of processing stages: extensions of Donders’ method,” Acta Psychol. :276–315 (1969). 33. D. J. Veltman, S. A. R. B. Rombouts, and R. J. Dolan, “Maintenance versus manipulation in verbal working memory revisited: an fMRI study,” NeuroImage (2):247–256 (2003). 34. T. S. Woodward, T. A. Cairo, C. C. Ruff, Y. Takane, M. A. Hunter, and E. T. C. Nganet, “Functional connectivity reveals load dependent neural systems underlying encoding and maintenance in verbal working memory,” Neuroscience (1):317–325 (2006). 35.
A. S. Fleisher, A. Sherzai, C. Taylor, J. B. S. Langbaum, K. Chen, and R. B. Buxtonc, “Resting-state BOLD networks versus task-associated functional MRI for distinguishing Alzheimer’s disease risk groups,” NeuroImage (4):1678–1690 (2009). 36. G. Repovs and D. M. Barch, “Working memory related brain network connectivity in individuals with schizophrenia and their siblings,” Frontiers in Hum. Neurosci. , 137 (2012). 37. S. L. Roser, P. G. Cleofé, M. A. U. Eider, V. P. Dídac, B. Nuria, J. Carme, and B. F. David, “Brain connectivity during resting state and subsequent working memory task predicts behavioural performance,” Cortex (9):1187–1196 (2012). 38. A. T. Newton, V. L. Morgan, B. P. Rogers, and J. C. Gore, “Modulation of steady state functional connectivity in the default mode and working memory networks by cognitive load,” Hum. Brain Mapp. (10):1649–1659 (2011). 39. M. W. Cole, T. Yarkoni, G. Repovs, A. Anticevic, and T. S. Braver, “Global connectivity of prefrontal cortex predicts cognitive control and intelligence,” J. Neurosci. (26):8988–8999 (2012). 40. F. Tian, A. Yennu, A. Smith-Osborne, F. Gonzalez-Lima, C. S. North, H. Liu, “Prefrontal responses to digit span memory phases in patients with post-traumatic stress disorder (ptsd): a functional near infrared spectroscopy study,” NeuroImage: Clinical. :808-819 (2014). 41. M. Okamoto, H. Dan, K. Sakamoto, K. Takeo, K. Shimizu, S. Kohno, I. Oda, S. Isobe, T. Suzuki, K. Kohyama, and I. Dan, “Three-dimensional probabilistic anatomical cranio-cerebral correlation via the international 10-20 system oriented for transcranial functional brain mapping,” NeuroImage (1):99–111 (2004). 42. S. M. H. Hosseini, Y. Mano, M. Rostami, M. Takahashi, M. Sugiura, and R. Kawashima, “Decoding what one likes or dislikes from single-trial fNIRS measurements,” NeuroReport (6):269–273 (2011). 43. Y. Zheng, D. Zhang, L. Wang, Y. Wang, H. Deng, S. Zhang, D. Li, and D. Wang, “Resting-state-based spatial filtering for an fNIRS-based motor imagery brain-computer interface,” IEEE Access :120603–120615 (2019). 44. J. Jonides, E. H. Schumacher, E. E. Smith, E. J. Lauber, E. Awh, S. Minoshima, and R. A. Koeppe, “Verbal working memory load affects regional brain activation as measured by PET,” J. Cognit. Neurosci. (4):462–475 (1997). 45. A . Berglund-Barraza, F. Tian , C. Basak, and J.L. Evans JL, “ Word Frequency Is Associated With Cognitive Effort During Verbal Working Memory: A Functional Near Infrared Spectroscopy (fNIRS) Study,” Front. Hum. Neurosci. :433 (2019). 46. C. Habeck, B. C. Rakitin, J. Moeller, N. Scarmeas, E. Zarahn, T. Brown, and Y. Stern, “An event-related fMRI study of the neural networks underlying the encoding, maintenance, and retrieval phase in a delayed-match-to-sample task,” Cognit. Brain Res. (2-3):207–220 (2005). 47. J. Spaniol, P. S. R. Davidson, A.S. N. Kim, H. Han, M. Moscovitch, and C. L. Grady, “Event-related fMRI studies of episodic encoding and retrieval: Meta-analyses using activation likelihood estimation,” Neuropsychologia (8–9):1765–1779 (2009). 48. J. M. Baker, J. L. Bruno, A. Gundran, S. M. H. Hosseini, and A. L. Reiss, “fNIRS measurement of cortical activation and functional connectivity during a visuospatial working memory task,” PLoS ONE (8):e0201486 (2018). 49. M. D'Esposito, G. K. Aguirre, E. Zarahn, D. Ballard, R. K. Shin, and J. Lease, “Functional MRI studies of spatial and nonspatial working memory,” Cognit. Brain Res. (1):1–13 (1998). 50. D. A. Fair, N. U. F. Dosenbach, J. A. Church, A. L. Cohen, S. Brahmbhatt, F. M. Miezin, D. M. Barch, M. E. Raichle, S. E. Petersen, and B. L. Schlaggar, “Development of distinct control networks through segregation and integration,” PNAS USA (33):13507–13512 (2007). 51.
A. M. Owen, R. G. Morris, B. J. Sahakian, C. E. Polkey, and T. W. Robbins, “Double dissociations of memory and executive functions in working memory tasks following frontal lobe excisions, temporal lobe excisions or amygdalo-hippocampectomy in man,” Brain (5):1597–1615 (1996). 52.
I. R. Olson, “Working memory for conjunctions relies on the medial temporal lobe,” J. Neurosci.26