GARCIA adolfo Martin
congresos y reuniones científicas
Linguistic markers of behavioral variant frontotemporal dementia. Automated speech analysis and multimodal signatures of text comprehension
Conferencia; Alzheimer's Association 2021 International Conference; 2021
Institución organizadora:
Alzheimer's Association
Background: Behavioral variant frontotemporal dementia (bvFTD) is typified by insidious deficits in personality and interpersonal conduct. While classical cognitive studies have targeted relevant domains (e.g., empathy, emotion recognition), recent evidence suggests that early deficits can also be present in language functions, some of which may be affected in 75% of patients. However, the evidence is limited in its linguistic coverage and methodological sophistication, with most studies targeting English speakers via basic language tasks. Seeking to bridge both gaps, my team has examined Spanish-speaking bvFTD cohorts through automated speech analyses (Experiment-1) and a novel framework revealing multimodal markers of naturalistic text comprehension (Experiment-2).Methods: In Experiment-1, we audio-recorded 17 bvFTD patients and 19 controls as they described their daily routine. Their speech was analyzed with automated measures of semantic organization and phonemic identifiability. Features from both measures were fed into a machine learning pipeline to establish the accuracy with which patients can be distinguished from controls. In Experiment-2, we recruited 20 bvFTD patients, 41 controls, and persons with Alzheimer?s disease (n = 23) and Parkinson?s disease (n = 25). All groups completed a text comprehension task involving four conditions: high social content, low social content, high action content, and low action content. Behavioral performance in each condition was correlated with offline measures of brain volume, fMRI connectivity, and EEG connectivity.Results: In Experiment-1, we found that semantic organization and phonemic identifiability provided maximal classification between bvFTD patients and controls when framed together rather than separately (accuracy = 77.8). In Experiment-2, bvFTD patients showed selective deficits in comprehending texts with high social content, associated with the volume and fMRI connectivity of social cognition regions (e.g., medial orbitofrontal cortex, medial temporal gyrus, amygdala), as well as reduced EEG connectivity over frontotemporal electrodes. No other patient group showed this selective behavioral pattern, and their specific deficits were linked to different multimodal patterns.Conclusions: These findings suggest that specific linguistic alterations can robustly discriminate bvFTD patients from controls and persons with other neurodegenerative diseases. Moreover, they extend language research on bvFTD by revealing novel candidate markers in an under-researched language via cutting-edge, naturalistic tools.