Adolfo M. García
Cognitive Neuroscience Center (Universidad de San Andrés, Argentina)
Global Brain Health Institute (GBHI), University of California, San Francisco
Departamento de Lingüística y Literatura, Facultad de Humanidades, Universidad de Santiago de Chile
Title: “Digital speech markers: The verbal blueprint of brain disorders”
Abstract:
The increasing global incidence and diagnostic complexity of neurodegenerative conditions underscore the urgent need for assessment tools that are objective, accessible, and scalable. Digital speech markers—leveraging artificial intelligence to analyze natural language—offer a promising solution that aligns with these demands. In this presentation, I will outline our team’s approach to developing and applying such markers across diverse populations. I will begin by introducing the TELL platform, a web-based application designed to evaluate speech. Next, I will review recent empirical evidence supporting specific speech features as potential indicators of various disorders. These include disruptions in speech timing in the nonfluent/agrammatic variant of primary progressive aphasia, lexical and semantic irregularities in Alzheimer’s disease, motor speech and action-language impairments in Parkinson’s disease, and atypical patterns of self-reference in behavioral variant frontotemporal dementia. I will then address the global disparities in access to these innovations and highlight ongoing multicenter efforts aimed at promoting equity in this field. Lastly, I will discuss the key scientific and translational challenges—and corresponding opportunities—that will shape the evolution of this research over the coming decade. Collectively, this talk will showcase the utility of digital speech markers as a pathway toward more equitable and efficient clinical assessments across neurodegenerative disorders.
Slides: The speaker's slides will be shared upon request.
Shantala Hegde | Additional Professor, Consultant Neuropsychologist
National Institute of Mental Health & Neuro Sciences, Bengaluru, India
Title: “Mapping brain-body-behaviour connection through understanding disconnection syndromes”
Abstract:
An impaired communication between different brain areas and pathways often causes a range of syndromes in neurological conditions. These syndromes are due to impaired communication between brain areas involved in processing various cognitive and motor functions. Agnosia, apraxia, callosal syndromes are a few examples which underscore the intricate connections between localized and specialized brain areas. In the early part of 20th century, in the field of neurology and neuropsychology, there was lowered interest in disconnection syndromes owing lesser importance on localizations view on brain functions. Deeper interest in understanding these syndromes were rekindled by the works of Norman Geschwind and it not only enhanced our understanding of the complex interconnections and distributed networks responsible for processing key brain functions but development of newer methods of restoring these connections via neurorehabilitation methods. Advances in neuroimaging methods have aided in providing the hypothesized connections between areas earlier supported by lesion studies and now moving towards understanding disorders of hyperconnectivity. An indepth understanding is crucial in advancing newer methods of neurorehabilitation as these syndromes can be devastating and debilitating for individuals suffering from various neurological conditions. Advances in newer brain imaging methods have played an important role in providing required evidence for promotion of these new rehabilitation techniques and methods. In my lecture I will focus on understanding the key theme of this workshop from a clinical perspective of disconnection syndrome and interconnections between brain areas.
Shalini Narayana | Professor
The University of Tennessee Health Science Center
Title: “Neural Correlates of Speech and Language in Neurological Disorders: Lessons in Brain (Re)Organization”
Abstract:
Speech and language functions are key to successful communication in humans. Several neurological disorders adversely impact the speech and language networks, compromising effective communication and overall quality of life. The talk will discuss the importance of localizing speech and language behaviors in the brain in various neurological disorders. The utility of non-invasive and invasive mapping modalities to localize speech and language in the brain in the clinical population will be described. In particular, the use of transcranial magnetic stimulation (TMS) as a diagnostic and a therapeutic tool will be highlighted. The talk will present examples of altered organization of the speech and language systems resulting from the underlying neurological disorders as well as appropriate interventions attesting to the importance of neural plasticity. Finally, the need for integrating augmented/artificial intelligence into the clinical evaluation of speech and language systems will be addressed.
Slides: The speaker's slides are available at this link.
Tanja Schultz | Professor
University of Bremen, Germany
Title: “Talk tells: Speech features-based prediction of cognitive change over the lifespan”
Abstract:
Speech is the most natural form of communication and demands high cognitive ability. It undergoes significant changes when cognitive resources decline, as seen for example dementia. These changes appear acoustically, with more pauses and hesitations, and linguistically, with reduced syntactic complexity, vocabulary, and semantic content. Machine learning (ML) methods can automatically and in real time extract, analyze, and interpret these acoustic and linguistic markers from everyday speech communication, and thus provide a cost-effective, simple, and casual screening method that can be performed at home without prior training or the need for specialized equipment.
In my talk, I will present ML methods that were developed at the Cognitive Systems Lab and evaluated in collaboration with the Institute for Gerontology at the University of Heidelberg using interview recordings from the Interdisciplinary Longitudinal Study on Adult Development and Aging (ILSE), which span a lifespan of over 20 years. Our results reveal that acoustic and linguistic markers differ between participants with dementia and healthy individuals. These methods can be applied to past recordings or used in real-time during conversations, identifying current cognitive states and detecting cognitive changes up to 12 years before the onset of dementia. In collaboration with Fraunhofer MEVIS we compare the performance of speech-based markers with markers derived from Magnetic Resonance Imaging (MRI) using ILSE and structural MRI data. Classifiers combining both modalities outperform single modalities for both current and predictive screening. In our dataset, speech markers proved as effective as MRI markers for predictive screening, indicating that speech markers can effectively complement traditional methods and operate independently of specialists' time. Further research with KIT demonstrates that speech-based markers are also effective in detecting depression, and in collaboration with INESC-ID and University of Lisbon we advocate speech as a biomarker for detecting speech-affecting disorders and diseases, such as Parkinson's disease and obstructive sleep apnea.
Finally, I will talk about new research training group in the field of hearables, which form a versatile health center on the ear, and about the Lifespan AI research group, in which we develop AI methods that lead from longitudinal predictions to lifespan inference in health.