Neuroimage. 2015 Mar;108:410-22. doi: 10.1016/j.neuroimage.2014.12.074.

Investigation of negative BOLD responses in human brain through NIRS technique. A visual stimulation study.

Maggioni E1, Molteni E2, Zucca C2, Reni G2, Cerutti S3, Triulzi FM4, Arrigoni F2, Bianchi AM3.
  • 1Department of Electronics, Information and Bioengineering, Politecnico di Milano, Milano, Italy; Scientific Institute IRCCS E.Medea, Bosisio Parini, Lecco, Italy. Electronic address:
  • 2Scientific Institute IRCCS E.Medea, Bosisio Parini, Lecco, Italy.
  • 3Department of Electronics, Information and Bioengineering, Politecnico di Milano, Milano, Italy.
  • 4Scientific Institute IRCCS E.Medea, Bosisio Parini, Lecco, Italy; Neuroradiology Unit, Fondazione IRCCS Cà Granda, Ospedale Maggiore Policlinico, Milano, Italy.


Despite negative blood oxygenation level dependent (BOLD) responses to visual stimuli have recently gained considerable interest, the explanation for their underlying neuronal and vascular mechanisms is still controversial. In the present study, a multimodal experimental approach is presented to shed light on the negative BOLD phenomenon in the human brain. In particular, information from functional magnetic resonance imaging (fMRI) and near infrared spectroscopy (NIRS) was integrated to confirm and gain insight into the phenomenon of negative BOLD responses (NBRs) to unpatterned intermittent photic stimulation (IPS) in healthy subjects. Eight healthy subjects participated in the study. Consistent findings emerged from the activation analysis of fMRI and NIRS data and the comparison of BOLD and hemoglobin responses at the single channel level showed that NBRs are related to a decrease in oxyhemoglobin (HbO) combined with a lower increase in deoxyhemoglobin (HHb), corresponding to a decrease in total hemoglobin (THb) and estimated cerebral blood volume (CBV). The HbO and HHb variations were significant in at least one channel in six subjects out of eight (p<0.05). The NIRS technique allowed obtaining valuable information on the vascular determinants of the NBRs, since the discrimination between HbO, HHb and THb information provided a more comprehensive view of the negative BOLD phenomenon. The within and between subject heterogeneous BOLD-Hb temporal relations pave the way to further investigations into the neurovascular properties of NBRs.




Functional magnetic resonance imaging (fMRI) has been largely employed to investigate the human brain processing of visual stimuli. The majority of fMRI studies focused on positive BOLD responses (PBR) to visual tasks, which are hypothesized to reflect evoked neuronal activity (Logothethis et al., 2001). However, it has been shown that visual stimuli can provoke negative BOLD responses (NBRs), i.e. BOLD amplitude decreases from rest to task, whose origin is still debated.

The NBRs have been explained in terms of either “vascular stealing” (Schmuel et al., 2002) or “neuronal inhibition” (Smith et al., 2004), but the most recent literature supports the hypothesis of neuronal origin for NBRs (Mullinger et al., 2014). However, the contribution of cerebral blood flow (CBF), blood volume (CBV) and metabolic rate of oxygen consumption (CMRO2) to the genesis of NBR needs further investigation.

Multimodal imaging is a powerful tool to this purpose, as it can provide a comprehensive view of the negative BOLD phenomenon. Relevant information on the hemodynamic and metabolic aspects of NBRs can be provided by near infrared spectroscopy (NIRS). Indeed, NIRS measures the changes in oxy-hemoglobin (HbO) and deoxy-hemoglobin (HHb) concentrations, whose mutual balance contributes to BOLD changes. The complementary temporal and spatial resolutions of NIRS and fMRI make their integration even more promising. In this study, the NIRS and fMRI techniques were integrated to investigate the intense NBRs elicited by intermittent photic stimulation (IPS) in healthy subjects. 



Eight healthy volunteers underwent separate NIRS and fMRI recordings with the same block-designed IPS protocol (14 s blocks at 6, 8, 10 and 12 Hz IPS alternated with 14 s resting blocks. Repetitions: 5 in fMRI study, 8 in NIRS study). The magnetic resonance (MR) images were acquired using a 3T scanner (Philips Achieva, the Netherlands). A continuous wave NIRS device (DYNOT Compact, NIRx, Berlin) equipped with a bilateral parieto-occipital 16-channels montage cap was used for the NIRS study. Two different wavelengths (760 nm and 830 nm) allowed to selectively probe HbO and HHb in the brain. Structural MR images provided morphological reference for both NIRS and fMRI data.

The fMRI activation analysis was performed using SPM8 ( software. After standard pre-processing, the effects of IPS compared to rest were evaluated in each subject and in the group using General Linear Model (GLM) analyses. The group significant BOLD responses were then extracted and quantitatively compared. Similarly, in the NIRS study the effects of IPS on HbO and HHb dynamics were assessed using GLM designs, but only at the single subject level.

For each subject, the fMRI statistical map resulting from the first level GLM analysis was projected on the subject’s cortical surface (reconstructed using the Freesurfer software,, on which the subject’s NIRS channels were projected. This procedure allowed to 1) identify the NIRS channels that were in proximity of the subject’s significant fMRI responses, 2) check whether the NBRs corresponded to negative NIRS responses.

To perform a quantitative comparison between fMRI and NIRS time series, we extracted the mean BOLD signal corresponding to the fMRI voxels underlying each NIRS channel. The NIRS sensitivity region was approximated as a 1 cm radius semi-sphere centered on the projection of the NIRS channel over the pial surface. In this way, we obtained one BOLD fMRI time series relative to each NIRS time series. The IPS effect size of BOLD and Hb signals was estimated, then the channels with an inverted hemodynamic response in both fMRI and NIRS (NBNR channels) were selected and used for a quantitative analysis. First, the correlation coefficient between BOLD and each Hb species was calculated. Finally, we selected a subset of NBNR channels characterized by sustained BOLD and NIRS responses to IPS, in which two peaks (after IPS onset and offset) were clearly identifiable. In the double peak response (DPR) channels, the times of occurrence of the BOLD, HbO and HHb peaks were used to assess any time delays between the signals.



The results of the fixed-effects group GLM analysis (p<10-4, False Discovery Rate (FDR) corrected) are shown in Figure 1. The regions with positive and negative BOLD response to IPS are shown in yellow/red and light blue/blue colors respectively. An extended activation was elicited by IPS in the primary visual cortex of both hemispheres. The PBR was in calcarine cortex, lingual gyrus and cuneus. Two significant NBRs were detected in symmetrical areas belonging to the extra-striate visual cortex, mainly in the lateral occipital inferior cortex. The left and right NBRs showed a very similar temporal trend. No significant differences emerged between PBR and NBRs in terms of onset time, peak amplitude and time and falling edge; however, compared to PBR, the NBRs were characterized by a later onset and a faster return to baseline.



Figure 1 Results of the fixed-effects GLM fMRI analysis on the eight subjects. Regions with significant BOLD response to IPS (FDR corrected, p<10-4) are projected on the reconstructed pial surface of a representative subject. Positive and negative response are yellow/red and light blue/blue respectively.


Although the effect size analysis suggested a higher IPS-related SNR in fMRI data compared to NIRS data, the two techniques led to results that were in overall agreement.

In every subject, the NIRS activation analysis showed a negative NIRS response to IPS (inverted compared to the canonical one) in one or more channels above the NBR regions. Such pattern was bilateral in all subjects except from two, who were characterized by low t-statistics modules (and low effect size of NIRS response) due to signal artifacts.

The NIRS-fMRI coregistration of one exemplar subject is shown in Figure 2. The NIRS channels located over the NBR regions of both hemispheres, n. 3-4 and 4-8, showed a significant negative NIRS response to IPS. In Figure 3, the HbO and HHb responses are plotted together with the underlying BOLD response: the NBR to IPS was associated with a decrease in HbO and an increase in HHb. The HbO response amplitude was greater than the HHb one, thereby indicating a decrease in total Hb (THb) concentration.

In the group, among the 39 channels with NBR, 30 channels had a negative NIRS response as well (NBNR channels). The median values of effect size across the NBNR channels and subjects were higher in BOLD (2.55) compared to HbO (0.4) and HHb (0.45); however, the Hb species showed higher effect size in the NBNR channels compared to the rest of occipital channels. The correlation analysis showed no significant differences in the correlation between BOLD and the three Hb species. The time-lag analysis performed on the DPR channels showed that the BOLD peaks usually preceded the HbO and HHb ones, as shown in Figure 3.



Figure 2 NIRS and fMRI results in one exemplar subject. The significant ROIs of the single-subject GLM fMRI analysis in native space (FDR corrected, p<0.001) are represented on the reconstructed pial surface of the subject (PBR in red, NBRs in blue). The NIRS channels are projected on the same surface. The GLM results of the HbO and HHb signals of the NIRS channels are represented in the two panels below.



Figure 3. Plot of BOLD and hemoglobin responses to IPS in channels 3-4 (left) and 4-8 (right) of the exemplar subject represented in Figure 2.



In this work, two different neuroimaging techniques, NIRS and fMRI, were employed to investigate the physiological hemodynamic response to IPS in a group of healthy subjects. The NIRS technique was introduced to extend the results of the fMRI study, which revealed the presence of two symmetric regions in extra-striate visual cortex characterized by significant NBR to IPS. For the first time, the NBR to visual stimuli were investigated using NIRS. The advantage of using NIRS relies in its ability to separate the dynamics of HbO and Hbb, thus providing a better insight into the physiological mechanisms involved in the genesis of the negative BOLD phenomenon.

Indeed, a full hemodynamic characterization of this phenomenon is still lacking. The NIRS-fMRI study confirmed that the NBRs to IPS in lateral occipital cortex resulted from a negative HbO response and a positive HHb response. The HHb response was smaller in amplitude than the HbO one, indicating a THb decrease. Since the THb changes provide a measure of CBV changes, our results suggest that the negative BOLD phenomenon is associated with a vasoconstriction, in line with (Huber et al. 2014). Moreover, although we did not provide CBF measures, our results are compatible with the hypothesis that NBRs correspond to decreases of both CBF and CMRO2 and are neuronal in origin (e.g., Mullinger et al., 2014).

The temporal relations that emerged from the NIRS-fMRI comparison have to be interpreted with caution, due to the fact that the acquisitions were made in different sessions. The time delay between the two signals may be related to the different sensitivity of NIRS and fMRI measures, to pial contamination in the NIRS study, or alternatively it may represent a peculiarity of the negative BOLD phenomenon. Further investigations are needed.

To verify the hypothesis of neuronal origin for NBRs, our group has recently provided direct knowledge of neuronal activity during IPS using simultaneous EEG-fMRI (Maggioni et al., 2016). The results of the EEG-informed fMRI analysis suggest a link between NBRs to IPS and occipital EEG power changes at 10 Hz and 12 Hz. Specifically, the BOLD fluctuations in inferior occipital cortex were 1) positively correlated to 10 Hz EEG dynamics, mainly representative of the resting-state alpha rhythm, 2) negatively correlated to 12 Hz EEG dynamics, showing a power increase during IPS compared to rest. Our EEG-fMRI results confirmed the relationship between neuronal activity and negative BOLD, but without providing evidence of neuronal inhibition itself.

Overall, our EEG-fMRI and NIRS studies provided useful information on the vascular characteristics of negative BOLD, as well as on its relationship with neuronal activity. However, our findings leave open questions on the ratio behind such relationship, which have to be addressed in the next future. Future measurements of CBF, CBV and CMRO2 changes are planned for this purpose.



  1. Huber, Laurentius, et al. “Investigation of the neurovascular coupling in positive and negative BOLD responses in human brain at 7T.” Neuroimage 97 (2014): 349-362.
  2. Logothetis, Nikos K., et al. “Neurophysiological investigation of the basis of the fMRI signal.” Nature 412.6843 (2001): 150-157.
  3. Maggioni, Eleonora, et al. “Investigation of the electrophysiological correlates of negative BOLD response during intermittent photic stimulation: An EEG‐fMRI study.” Human brain mapping 37.6 (2016): 2247-2262.
  4. Mullinger, Karen J., et al. “Evidence that the negative BOLD response is neuronal in origin: a simultaneous EEG–BOLD–CBF study in humans.” Neuroimage 94 (2014): 263-274.
  5. Shmuel, Amir, et al. “Sustained negative BOLD, blood flow and oxygen consumption response and its coupling to the positive response in the human brain.” Neuron 36.6 (2002): 1195-1210.
  6. Smith, Andrew T., Adrian L. Williams, and Krishna D. Singh. “Negative BOLD in the visual cortex: evidence against blood stealing.” Human brain mapping 21.4 (2004): 213-220.



Multiselect Ultimate Query Plugin by InoPlugs Web Design Vienna | Webdesign Wien and Juwelier SchönmannMultiselect Ultimate Query Plugin by InoPlugs Web Design Vienna | Webdesign Wien and Juwelier Schönmann