Interoception Conference Flashcards
Can you walk me through your poster? - (8)
- We recorded single units from the posterior insula (PI), anterior insula (AI), and Heschl’s gyrus (HG) in patients with medically intractable epilepsy who were implanted with intracranial electrodes. While the patients passively listened to tones (did 300 trials of this), we simultaneously recorded the ECG.
- Our study aimed to answer two questions using human single-unit recordings: 1) Does cardiac-related information reach the insula and auditory cortex? 2) Does the cardiac phase modulate auditory responses in the auditory cortex?
- Q1:We found that in the PI and AI, some units (6% in PI and 5% in AI) responded to heartbeats, either increasing (point to PI and AI) or decreasing firing rate (point to decrease in PI) after R-peak onset.
- Q1: More interestingly, in HG, 8% of the units analysed showed an inhibitory response to heartbeats, with this unit in HG having firing rate dipping around 300ms after the R-peak (point to graph). This suggests that cardiac information also reaches the auditory cortex, but with an inhibitory effect.
- For the second question, we divided 300 sound trials into systole (early after R-peak) and diastole (later in the RR interval) (point to figure).
- For results, second question there is two example of units in HG, 0 is tone onset and shows firing rate and rastor plots during systole and diastole – see much more cluster of spikes in rastor plot and higher firing rate in HG during diastole and systole in 7% of untis I analysed but no units showing the opposite.
- Which what we expected in the baroreceptor hypothesis which states during systole (cardiac contraction) it induces cortical inhibition than diastole (cardiac relaxation)
- In conclusion, cardiac-related information reaches both the insula and the auditory cortex, and auditory responses in HG are modulated by the cardiac phase, particularly with inhibition during systole – which is consistent with baroreceptor hypothesis.
Can you give me more information about your participants? - (3)
- These participants had medically intractable epilepsy, meaning their seizures are not adequately controlled with medication.
- Electrodes were placed on basis of clinical requirements to identify seizure foci and after implantation underwent 2-3 week monitoring period to localize their seizure focus where we did the tone task
- We identified participants who had depth electrodes in three regions of interest: AI, PI, HG which had not only marco contacts for LFP recordings but also microwires for unit activity.
What is the key message of your poster? (If someone is in a rush)
- The key message of my poster is that this is the first study to show, at the level of single neurons in humans, that cardiac-related signals reach not only the insula — a key region for interoception — but also the primary auditory cortex (Heschl’s gyrus), where they exert an inhibitory effect.
- Additionally, we show that auditory responses in Heschl’s gyrus are modulated by the phase of the cardiac cycle, with stronger responses during diastole than systole, supporting the baroreceptor hypothesis that systole (cardiac contraction) induces cortical inhibition than diastole (heart relaxes).
Can you give me in percentages of how many units in HG response during systole and diastole? - (2)
- HG No response: 93% (55 - removed some units not responding to sound)
- HG Higher D> S: 7% (4)
What is the limitations of your poster? - (3)
- The auditory task used simple tones; future studies could explore more naturalistic or ecologically valid sounds (e.g., speech sounds) to assess whether cardiac modulation generalizes across all auditory contexts.
- We defined systole and diastole offlin. A more dynamic approach would involve real-time (online) detection and alignment of stimulus presentation (tones) with cardiac phases — something we’re currently working with Hugo Critchley and his PhD student.
Pulsatility artifacts, which can arise due to mechanical displacement of electrodes, might affect signals in certain contexts, especially when analyzing cardiac-related responses. However, this is less likely to influence our second question on auditory ERP modulation, since pulsatility mainly affects responses to heartbeats. Regarding the pulsatility artifact, I spoke with an expert who works with single-unit data, and he believes pulsatility may not be a concern for single units. I still need to explore the literature more to rule this out
What does ‘single units were isolated offline via spike sorting followed by manual curation’ mean? - (2) - both Q1 and Q2
- Took raw signals from recorded from brain and separated activity of individual neurons (spike sorting)
- Then human expert reviewed and refined sorting (manual curation) to ensure only clean and well-isolated neurons are kept
What does ‘SDF computed with time resolution of 1 ms and SD of 5ms’ mean? both Q1 and Q2 - (5)
- We smoothed each neuron’s spike train using a spike density function (SDF), which turns discrete spikes into a continuous estimate of firing rate over time.
We did this at a 1 ms resolution, so we could track neural activity millisecond by millisecond.
To reduce noise but keep timing precise, we applied Gaussian smoothing with a standard deviation (SD) of 5 ms — that means each spike contributes to the firing rate in a small window around it, mainly within ±5 ms, but tapering off beyond that.
This helps reveal real trends in firing without losing fine timing.
We then averaged the SDF across trials to get the firing rate plots you see aligned to heartbeats or tones.
What does ‘Statistical analysis of firing rate for cardiac responses were performed by conducting t-tests at every time point followed by FDR correction. - Q1 - (2)
- To test whether neurons responded to heartbeats, we compared their firing rate to baseline at every millisecond, using t-tests – comparing it to baseline.
- Since we’re doing lots of tests across time, we used False Discovery Rate (FDR) (less stringent than Bonferroni) correction to reduce the chance of false positives
What does ‘For sys/dis comparison, permutation testing with cluster correction performed’? - (3)
- For the systole vs. diastole comparison, we used permutation testing with cluster correction
— this method randomly shuffles labels across trials to build a distribution of what results you’d expect by chance, and looks for clusters of significant differences that are unlikely to occur randomly.
It’s a robust way to test for differences while controlling for multiple comparisons
What about pulsatility artifacts? - (3)
- Pulsatility artifacts are distortions caused by the heartbeat physically moving the brain or electrodes, often affecting low-frequency signals.
They’re more relevant for our first question (heartbeat-locked responses), but likely don’t affect single-unit spikes much — an expert in single-unit data confirmed this might not be a concern, though I plan to explore it further.
- For our second question (auditory ERPs during systole vs. diastole), artifacts are less of an issue since the analysis is sound-locked, and systole/diastole trials include the same stimuli, helping cancel out any unrelated noise.
Introduction: Q1:Measurement of single unit activity in the human insula and auditory cortex in response to heartbeats. - (6)
- Insula is a key region involved in sensing and perception of interoceptive signals in general and heartbeats in particular.
- The evidence for insula involvement in processing heartbeats comes from measurement of heart-beat evoked potentials (HEP) using EEG/MEG.
- Given comparatively low spatial resolution of EEG/MEG, it is difficult to pinpoint the role of insula in generating HEPs.
- Although there are few studies recording local field potentials (LFPs) in humans in response to heartbeats (e.g. Park et al, 2018), the response of human insula to heartbeats at single-neuron level has not been studied before.
- There is no previous work of cardiac responses in the primary auditory cortex (Heschl’s gyrus;HG)
- We recorded single neuron activity from the posterior (PI) and anterior insula (AI) and HG in human epileptic patients undergoing intracranial-EEG monitoring.
Introduction Q2: Is the excitability of the auditory cortex modulated by cardiac phase? - (3)
- The baroreceptor hypothesis (Lacey & Lacey, 1978) states that the systole (cardiac contraction) phase of cardiac cycle induces cortical inhibition.
- To test this hypothesis, we measured single neuron activity in response to sounds in the HG and compared its activity in the systole and diastole phases of the cardiac cycle.
- We expected firing rate to be lower in systole compared to the diastole
Methods - Participants and Data Collection - (6)
Participants were 10 adult patients with intractable epilepsy who were implanted with intracranial electrodes and undergoing 2 to 3-weeks of monitoring for localization of their seizure focus.
- Depth electrodes (Ad-Tech Medical, Oak Creek, WI) were implanted using stereo encephalography (sEEG) in brain sites determined by clinical requirements.
- Single unit activity was measured using Behnke Fried depth electrodes, which are hybrid electrodes consisting of macro contacts for LFP recordings and microwires for unit activity.
- All data were collected using Nihon Kohden 256-channel amplifier and Neuralynx ATLAS Neurophysiology System.
- Electrocardiogram (ECG) was simultaneously recorded with the neural activity.
- Out of 10 participants, HG was implanted in 4, PI in 6 and AI in 7 participants.
Methods: Stimuli, Paradigm and Data Analysis - (9)
- Stimuli were pure tones (300ms) of six different frequencies (250 to 8000Hz in octave steps).
- Participants passively listened to these tones.
- A total of 300 trials (50 each for the 6 frequencies) were presented in random order.
- Heartbeats (R-peaks) from the ECG signals were detected and visually inspected using R-DECO toolbox (Moeyersons et al., 2019).
- Single neuorns were isolated offline via spike sorting, followed by manual curation.
- To detect if unit activity was different in the systole and diastole phases, trials were divided in two categories: trials in which tone was presented in the first-half (systole trials) and those in the later half of RR interval (diastole trials).
- Spike density functions (SDF) was computed at a time resolution of 1ms with a standard deviation of 5ms.Neural firing rate
was computed by averaging SDF across trials. - Statistical analysis of firing rate for cardiac response was performed by conducting t-tests at every time point followed by FDR (False discovery rate) correction.
- For Sys/Dias comparison, permutation testing with cluster correction was performed.
Results: % of Units Analysed - (3)
- PI: 128
- AI: 124
- HG: 58
Results: Latency of Response to Heartbeats (ms) - (3)
- PI: ~240ms
- AI: ~270ms
- HG: ~300ms
Results: Percentage of units activated by heartbeats - (3)
- PI: 2% ([2] increase), 4% ([5]decrease)
- AI: 1%( [1] increase), 3% ([4] decrease)
- HG: 8/9% decrease [5]
Discussion - (8)
- Units in the PI respond to heartbeats. This is consistent with PI being a recipient of afferent cardiac signals from within the body.
- Units in the AI, however, are also activated by the heartbeats. The latency measure suggests that these may be activated later in time compared to the activation in the PI.
- The surprising finding, however, is that the primary auditory cortex (HG) also responds to cardiac beats. Interestingly, the latency of activation for HG is longer than that found in both PI and AI, suggesting an indirect pathway for cardiac beats to reach the primary auditory cortex.
- There is emerging evidence (e.g. Park et al, 2014) that signals from the body modulate perception of signals from the external world (e.g. visual and auditory). Activation of HG in response to heartbeats could be a possible neural signal for modulation of auditory perception by the heart.
- How the cardiac signal reaches the primary auditory cortex needs further investigation.
- Analysis of response to sounds in the HG show that firing rate of neurons is stronger during diastole compared to systole. The results are in agreement with cortical inhibition during systole phase when baroreceptors sends a strong afferent signal (Lacey and Lacey, 1978).
- Interestingly, the response of HG to cardiac beats is always inhibitory. That is, cardiac beat lowers the unit’s firing rate ~350 ms after its occurence. If a sound occurs around this inhibition phase of HG, response to sounds may be weaker compared to when the sound occurs outside the inhibition window. This provides a mechanistic explanation of how cardiac signals could modulate perception of the auditory signals.
- Further confirmation is needed by systematically changing the sound onset with respect to the cardiac beats.
Conclusion - (3)
- To the best of our knowledge, this is the first work to show cardiac responses in the human insula and auditory cortex at the single neuron level.
- The data shows that information regarding heartbeats reaches to the primary auditory cortex and that its response to sounds is modulated by the cardiac phase.
- How the cardiac signals affect perception at the behavioral level and its correlation with HG inhibition needs further investigated
Pathways of cardiac signals to HG - (3)
It is known cardiac afferent signals during systole go to brainstem, thalamus and insula
Unknown for pathway of cardiac signals to reach HG
Could be after cardiac signals reaches posterior insula first it reaches HG since they are close together
