TL;DR
This study provides a crucial link between the macroscopic BOLD signal measured by fMRI and the microscopic neural firing rates in the human brain. By simultaneously recording both in patients with epilepsy, researchers found that fMRI "deactivations" (BOLD signal decreases) in the fusiform gyrus are directly associated with a decrease, and sometimes even a complete suppression, of single neuron spiking activity when viewing faces. This work clarifies the physiological basis of fMRI signals, particularly negative BOLD responses, in human association cortex.
What Problem Does This Paper Address?
Functional Magnetic Resonance Imaging (fMRI) has been a revolutionary tool for understanding brain function by measuring the Blood-Oxygen-Level-Dependent (BOLD) signal. However, the precise relationship between this BOLD signal and the underlying neural activity, specifically single neuron firing rates, has remained a significant question, especially in human association cortices. While animal studies suggest BOLD activation correlates with neural activity, the cellular basis of BOLD deactivations (negative BOLD signals) is particularly unclear. These deactivations have been attributed to various factors, including relative decreases in neural activity, but direct evidence in humans, particularly in complex cognitive areas like face processing, has been lacking. This paper aims to bridge this gap by directly comparing fMRI BOLD signals with single-unit (SU) recordings in neighboring regions of the human fusiform gyrus (FG), a key area for face recognition.
Methods
The study employed a unique combination of neuroimaging and electrophysiological techniques in two patients with refractory epilepsy who were undergoing clinical evaluation with depth electrodes.
- fMRI (Functional Magnetic Resonance Imaging): A well-established Fast Periodic Visual Stimulation (FPVS) paradigm was used. This involved presenting sequences of face images interspersed with object images at a specific frequency (0.111 Hz for fMRI). This method allows for the identification of brain regions that respond selectively to faces by analyzing the BOLD signal at the tagged frequency.
- Intracerebral Electrophysiology: Hybrid macro-microelectrodes were implanted in the fusiform gyrus. These electrodes allowed for the simultaneous recording of both macroscopic local field potentials (LFPs) and microscopic single-unit (SU) spiking activity. The same FPVS paradigm used in fMRI was adapted for these recordings, with faces presented at a higher frequency (1.2 Hz) to allow for precise temporal analysis of neural responses.
- Frequency-Tagging Paradigm: This core technique was applied to both fMRI and electrophysiology. By presenting stimuli at specific frequencies, researchers could isolate neural responses at those frequencies, effectively separating face-selective activity from general visual stimulation and low-level visual cues.
- Spike Sorting and Analysis: Raw electrophysiological data were processed to identify and classify individual neuronal spikes (SU) and multi-unit activity (MU). Time-domain and frequency-domain analyses were performed on these spike trains to quantify responses to faces versus objects.
Datasets
The study utilized data from two human participants who underwent both fMRI and intracerebral recordings. The data generated by this study, including intracerebral EEG and fMRI data, are made accessible on OSF (Open Science Framework) under the title "A tight relationship between BOLD fMRI activation/deactivation and increase/decrease in single neuron responses in human association cortex." No new benchmarks were introduced; rather, existing paradigms were applied to a novel experimental setup.
Key Insights and Ideas
- BOLD Deactivations Reflect Neural Suppression: The most significant insight is the direct demonstration that BOLD signal decreases in the human fusiform gyrus, particularly in the medial FG, are associated with a decrease in single neuron firing rates. This provides strong evidence that negative BOLD signals can reflect genuine neural suppression, not just a relative reduction compared to a more active condition.
- Dual Nature of Face Representation: The study reveals that face-selective neural activity in the human brain is not monolithic. While the lateral FG shows increased neural firing to faces (consistent with BOLD increases), the medial FG exhibits decreased firing, and in some cases, complete suppression of baseline activity, to faces. This highlights a more complex neural coding scheme for faces.
- Physiological Basis of fMRI: This work offers a critical piece of the puzzle in understanding the physiological underpinnings of fMRI signals in humans. It validates that BOLD signals, both positive and negative, can be directly linked to specific patterns of neuronal firing in higher-order association cortices.
- Importance of Inhibitory Activity: The finding of genuine suppression of baseline spiking activity suggests that inhibitory neural processes play a crucial role in shaping category-selective responses in the human brain, a concept that has been theorized but is harder to demonstrate directly with non-invasive methods.
Key Results
- Face-Selective Neurons: Approximately 70% of recorded neurons in both participants showed face-selective responses.
- Lateral FG (P1): Neurons in the lateral FG (associated with fMRI increases) showed a significant increase in spiking activity to faces (88.9% of face-selective neurons).
- Medial FG (P2): Neurons in the medial FG (associated with fMRI decreases) showed a significant decrease in spiking activity to faces (95.9% of face-selective neurons).
- Genuine Suppression: Among the neurons showing reduced firing to faces in the medial FG, about one-third (34 out of 95) exhibited genuine suppression of their baseline spiking activity.
- Response Latency: The onset latency for both increases and decreases in spiking activity was similar (around 118-127 ms).
- Average Firing Rate: Decreased responses in the medial FG were associated with a significantly lower absolute average firing rate compared to increased responses in the lateral FG.
Limitations and Future Work
- Individual Differences: The study involved only two participants, and while the findings were consistent between them and across hemispheres (based on previous literature), the small sample size limits generalizability. The authors acknowledge that differential recordings in the same hemisphere would have been optimal.
- Clinical Context: The participants were patients with epilepsy, and the presence of electrodes and their clinical condition could potentially influence neural activity. However, the electrodes were placed to delineate seizure zones, and the study focused on regions outside these zones.
- Functional Relevance: While the study clarifies the physiological basis of the fMRI signal, further work is needed to directly link these cellular responses to specific visual recognition behaviors.
- Cellular Mechanisms: The precise cellular and circuit mechanisms underlying both the excitatory and inhibitory neuronal activities in these face-selective regions remain to be fully elucidated. Future research could explore these mechanisms through more detailed cellular recordings and anatomical studies.
Summary updated