Conference Agenda

Hallucinations, considered manifestations of altered consciousness, occur when normal sensory processing is disrupted, as seen in Parkinson’s disease (PD). Minor hallucinations (MH), such as presence hallucinations and misperceptions, affect 50-60% of PD patients early on, and are a major risk factor for cognitive decline and dementia (Aarsland et al., 2021). However, the neuropsychological correlates of MH in PD remain unclear. Many studies fail to link MH to cognitive deficits, as patients with MH show similar cognitive performance to those without hallucinations (Bejr‐kasem et al., 2019). Using a neurotechnological approach (Bernasconi et al., 2022), we investigated whether specific cognitive deficits are associated with robot-induced MH, namely presence hallucinations (PH), the convincing sensation that a human presence is nearby when no one is here. Twenty-five PD patients with MH (PD-MH) and 25 without hallucinations (PD-nH) underwent neuropsychological assessments and a robotic procedure that induces PH. Our results show that PD-MH patients are more sensitive to the robotic procedure than PD-nH patients, supporting previous findings. While both groups performed similarly in neuropsychological tests, PD-MH patients reported significantly more subjective cognitive decline. Critically, we found a significant association between increased sensitivity to the robotic procedure (inducing hallucinations) and poorer performance in frontal-subcortical cognitive functions (associated with altered aspects of consciousness, mainly in attention, executive control), particularly in PD-MH patients. These findings suggest that when sensitivity to robot-induced hallucinations is considered, a more detailed link between cognitive decline and hallucinations emerges, providing early detection of PD progression not captured by current methods in clinical practice.

Introduction: Many different brain regions have been implicated in auditory hallucinations leading to divergent theories regarding their cause. Lesion-induced cases of auditory hallucinations can help identify brain structures that are causal rather than correlated.

Methods: We completed a systematic literature review of lesions associated with new onset pure auditory hallucinations without other psychotic features. Lesion locations were mapped for their functional connectivity from a large, resting-state connectome (n=1000). Lesion network maps associated with auditory hallucinations were compared on voxelwise two-sample T-test to controls and to each other based on phenotype.

Results: Most lesional causes of pure auditory hallucinations were of music (54%) and associated with hearing loss. We did not identify any cases of verbal, lesion-induced hallucinations without other psychotic features. Peak positive connections in our musical hallucination network were in early auditory centers (cochlear, superior olivary and trapezoid nuclei) with anti-correlations in higher auditory processing regions (superior temporal lobe, medial geniculate body). When contrasted, musical hallucinations had peaks in rhythm generation centers (supplementary motor area) while vocal hallucinations mapped to memory locations (hippocampi).

Conclusions: Musical hallucinations were the most common semiology of lesional auditory hallucinations despite the fact that verbal hallucinations are far more common in psychiatric disorders. Lesions causing pure auditory hallucinations demonstrate connections to lower and higher order auditory processing regions which fits hallucination theories focused on disconnection of bottom-up and top-down processing. Auditory hallucinations with delusions better fit memory intrusion models. Phenomenology could clarify how competing theories of hallucinations differentially explain semiologically distinct hallucinatory disorders.

Simple visual hallucinations are characterised by colourful geometric patterns that are reported across distinct aetiologies, including stroboscopic stimulation, suggesting shared underlying neural mechanisms. To characterise the content of these hallucinations, we conducted an analysis using a unsupervised machine learning image clustering approach on a large dataset comprising over 10,000 drawings that were created by participants in response to an invitation to visually describe their experiences in the Dreamachine — an immersive multisensory experience (developed by Collective Act) that combined strobe lighting and 360-degree spatial sound to induce hallucination-like experiences (and other altered states of consciousness) in nearly 40,000 participants (https://dreamachine.world/).

Using DINOv2—a pre-trained, vision transformer—to extract visual features, and k‑means clustering to partition the drawings, we found that 38 clusters best captured the geometries in the dataset, as validated by the average silhouette score, a measure of the clustering quality. The clusters confirmed and extended previous findings, notably identifying all four Klüver form constants as separate clusters.

We identified four dimensions of visual content describing the geometries within the clusters: Global Geometry, Curvature, Polar Geometry, and Rotational Polar Form Symmetry. Our analysis revealed a subset of clusters dominated by a 4‑fold rotational symmetry (accounting for over 5% of drawings) as well as clusters defined by non‑Klüver forms (e.g., amorphous, uniform/mix of colours, miscellaneous geometries, and semantic content). Together, our findings expand the description of stroboscopically induced simple visual hallucinations, and encourage the development of more sophisticated computational models that can elucidate their neural mechanisms.

Divergent perception is the atypical experience of sensory information, which includes simple (e.g., geometric patterns, colors) and complex visual hallucinations (meaningful objects and environments). Complex hallucinations are a hallmark symptom of psychiatric disorder, but recent research has found they can also be induced in the healthy population using a technique called Ganzflicker – a rhythmic flickering of colors on a computer screen. This work suggests there may be common cognitive mechanisms of complex hallucinations that can be explored in the healthy population. In the current study, we explored the predictive role of three evidence-based factors in complex hallucination proneness: previous anomalous experiences (e.g., seeing ghosts), schizotypal traits (i.e., schizophrenia-like traits found in the general population), and mental imagery vividness (i.e., the ability to mentally simulate visual sensory information in the “mind’s eye”). 142 participants completed the Vividness of Visual Imagery Questionnaire (VVIQ; mental imagery measure), the Creative Experiences Questionnaire (CEQ; schizotypy measure), and reported any prior anomalous experiences. They then underwent 10 minutes of Ganzflicker stimulation, and afterwards reported the contents of hallucinations experienced in that time, both simple and complex. VVIQ, CEQ, and prior anomalous experience (yes/no) were added as predictors in a logistic regression with hallucination complexity (simple, complex) as the outcome variable. Results showed that more vivid imagery, higher schizotypal traits, and prior anomalous experience all contributed to a higher likelihood of complex Ganzflicker hallucinations. This research aligns with a recent model of divergent perception and has implications for understanding hallucination susceptibility in diverse populations.

Aberrant perceptions, such as hallucinations, are a predominant non-motor symptom in Parkinson’s disease (PD) and have often been attributed to dopaminergic treatments that are systematically used to improve the motor symptoms of the disease. While clinical wisdom suggests a strong association between dopaminergic medication and hallucinations, empirical data do not clearly support this regard, and the cortical and subcortical mechanisms by which dopamine putatively lead to hallucinations remain unresolved. This is because hallucinations are challenging to study under controlled experimental settings. To circumvent those limitations we developed a method and procedure inducing clinically relevant hallucinations in PD. Here, to provide evidence on whether dopaminergic treatments are associated with hallucinations we tested patients with PD under On (i.e. usual intake) and Off (i.e. medication withdraw) dopaminergic medications, while performing our sensorimotor robotic procedure inducing presence hallucinations, and recorded resting-state fMRI under both medications conditions. Our results show that i) presence hallucinations in PD are induced through conflicting sensorimotor stimulation, and ii) dopaminergic medications facilitate the occurrence of those hallucinations. Moreover, iii) partial least squares fMRI analysis associated the increased sensitivity to experience those hallucinations in the On conditions with a higher functional connectivity within the dorsal striatum and between the dorsal striatum and prefrontal-premotor regions. Collectively, these results present conclusive experimental-clinical evidence that dopamine favors (sensorimotor) hallucinations occurrence, and provide direct evidence of the involvement of subcortical and cortical structures in hallucinations, thereby suggesting a subcortical-cortical view of the neural correlates of (aberrant) perception.

Hallucinations in Parkinson’s Disease (PD) have been linked to dopaminergic medication, yet this association is still debated (Ravina et al, 2007). Our group developed a robotic setup to induce presence hallucinations (robot-induced presence hallucinations, riPH), enabling the controlled investigation of hallucination susceptibility. Recent work has showed increased riPH sensitivity in PD patients during dopaminergic treatment (Bernasconi & Stampacchia, in prep.). The present study investigates how dopaminergic medication modulates functional connectivity (FC) fingerprints (Stampacchia et al., 2024) and their relationship with riPH sensitivity.

Twenty PD patients underwent resting-state fMRI and the robot experiment in ON- and OFF-medication state. riPH sensitivity was quantified for each individual using a delay-dependent response probability. We estimated individual’s identifiability based on FC matrices within and across medication states (ON vs. OFF). Edgewise intra-class correlation (ICC) assessed topographic features of FC fingerprints within and across medication states. A linear mixed model tested whether high-ICC FC edges and medication state predicted riPH sensitivity, adjusting for age, sex, and disease duration.

Individual FC profiles were highly identifiable within sessions (IDiff ON=0.36, OFF=0.35; 100% success), but cross-medication comparisons reduced subjects’ identifiability (IDiff=0.17-0.21; 98-99% success). ON-medication increased ICC in visual, somatomotor, salience, subcortical, and cerebellar networks (FDR-corrected). High-ICC FC (β=16.41, p=0.041) and medication state (β=67.52, p=0.017) predicted riPH sensitivity, with an interaction indicating altered effects OFF-medication (β=-28.32, p=0.018).

Dopaminergic medication reconfigures FC fingerprints in PD, stabilizing sensorimotor and salience networks and modulating hallucination sensitivity. These findings advance understanding of individual connectomes in PD, and the association between dopaminergic treatment and hallucinations.