Benhamou, Elia; (2021) Predictive cognition in dementia: the case of music. Doctoral thesis (Ph.D), UCL (University College London).

Preview

Text Thesis_postViva.pdf - Accepted Version Download (3MB) | Preview

Abstract

The clinical complexity and pathological diversity of neurodegenerative diseases impose immense challenges for diagnosis and the design of rational interventions. To address these challenges, there is a need to identify new paradigms and biomarkers that capture shared pathophysiological processes and can be applied across a range of diseases. One core paradigm of brain function is predictive coding: the processes by which the brain establishes predictions and uses them to minimise prediction errors represented as the difference between predictions and actual sensory inputs. The processes involved in processing unexpected events and responding appropriately are vulnerable in common dementias but difficult to characterise. In my PhD work, I have exploited key properties of music – its universality, ecological relevance and structural regularity – to model and assess predictive cognition in patients representing major syndromes of frontotemporal dementia – non-fluent variant PPA (nfvPPA), semantic-variant PPA (svPPA) and behavioural-variant FTD (bvFTD) - and Alzheimer’s disease relative to healthy older individuals. In my first experiment, I presented patients with well-known melodies containing no deviants or one of three types of deviant - acoustic (white-noise burst), syntactic (key-violating pitch change) or semantic (key-preserving pitch change). I assessed accuracy detecting melodic deviants and simultaneously-recorded pupillary responses to these deviants. I used voxel-based morphometry to define neuroanatomical substrates for the behavioural and autonomic processing of these different types of deviants, and identified a posterior temporo-parietal network for detection of basic acoustic deviants and a more anterior fronto-temporo-striatal network for detection of syntactic pitch deviants. In my second chapter, I investigated the ability of patients to track the statistical structure of the same musical stimuli, using a computational model of the information dynamics of music to calculate the information-content of deviants (unexpectedness) and entropy of melodies (uncertainty). I related these information-theoretic metrics to performance for detection of deviants and to ‘evoked’ and ‘integrative’ pupil reactivity to deviants and melodies respectively and found neuroanatomical correlates in bilateral dorsal and ventral striatum, hippocampus, superior temporal gyri, right temporal pole and left inferior frontal gyrus. Together, chapters 3 and 4 revealed new hypotheses about the way FTD and AD pathologies disrupt the integration of predictive errors with predictions: a retained ability of AD patients to detect deviants at all levels of the hierarchy with a preserved autonomic sensitivity to information-theoretic properties of musical stimuli; a generalized impairment of surprise detection and statistical tracking of musical information at both a cognitive and autonomic levels for svPPA patients underlying a diminished precision of predictions; the exact mirror profile of svPPA patients in nfvPPA patients with an abnormally high rate of false-alarms with up-regulated pupillary reactivity to deviants, interpreted as over-precise or inflexible predictions accompanied with normal cognitive and autonomic probabilistic tracking of information; an impaired behavioural and autonomic reactivity to unexpected events with a retained reactivity to environmental uncertainty in bvFTD patients. Chapters 5 and 6 assessed the status of reward prediction error processing and updating via actions in bvFTD. I created pleasant and aversive musical stimuli by manipulating chord progressions and used a classic reinforcement-learning paradigm which asked participants to choose the visual cue with the highest probability of obtaining a musical ‘reward’. bvFTD patients showed reduced sensitivity to the consequence of an action and lower learning rate in response to aversive stimuli compared to reward. These results correlated with neuroanatomical substrates in ventral and dorsal attention networks, dorsal striatum, parahippocampal gyrus and temporo-parietal junction. Deficits were governed by the level of environmental uncertainty with normal learning dynamics in a structured and binarized environment but exacerbated deficits in noisier environments. Impaired choice accuracy in noisy environments correlated with measures of ritualistic and compulsive behavioural changes and abnormally reduced learning dynamics correlated with behavioural changes related to empathy and theory-of-mind. Together, these experiments represent the most comprehensive attempt to date to define the way neurodegenerative pathologies disrupts the perceptual, behavioural and physiological encoding of unexpected events in predictive coding terms.

Download activity - last month

Export as JSONXMLCSV

Download activity - last 12 months

Export as XMLCSVJSON

Downloads by country - last 12 months

Export as JSONXMLCSV

Archive Staff Only

View Item