Named One of London's Top 10 Scaleups for 2025!

  • Named One of London's Top 10 Scaleups for 2025!

  • Named One of London's Top 10 Scaleups for 2025!

Publications

With over 200 publications between the team, thymia has proven scientific excellence.

270

publications

270

publications

270

publications

Paper

1 September 2024

Revealing Confounding Biases: A Novel Benchmarking Approach for Aggregate-Level Performance Metrics in Health Assessments

This research examines the overestimation of performance metrics in speech-based health assessments that use machine learning or artificial intelligence, particularly for conditions like depression and Alzheimer’s disease. Confounding factors, such as age, sex, and accent can significantly bias these analyses, often leading to misleading results.

The paper proposes a novel benchmarking approach that helps highlight and quantify these biases in a dataset in the presence of confounders. This underscores the importance of considering demographic factors such as age, sex, and accent in developing and evaluating speech-based diagnostic tools to ensure more accurate and fair health assessments.

Paper

1 September 2024

Revealing Confounding Biases: A Novel Benchmarking Approach for Aggregate-Level Performance Metrics in Health Assessments

This research examines the overestimation of performance metrics in speech-based health assessments that use machine learning or artificial intelligence, particularly for conditions like depression and Alzheimer’s disease. Confounding factors, such as age, sex, and accent can significantly bias these analyses, often leading to misleading results.

The paper proposes a novel benchmarking approach that helps highlight and quantify these biases in a dataset in the presence of confounders. This underscores the importance of considering demographic factors such as age, sex, and accent in developing and evaluating speech-based diagnostic tools to ensure more accurate and fair health assessments.

Paper

17 June 2024

Predicting different dimensions of fatigue from speech data: a longitudinal study in shift workers

This research study examines how speech data can be used to predict fatigue levels in shift workers. Over two weeks, 1,197 participants from six countries across various shift work patterns and industries provided speech samples twice daily, making this one of the most comprehensive investigations of its kind.

The research found that speech characteristics could reliably indicate sleep deprivation and feelings of sleepiness from a single speech activity, in models which properly accounted for individual differences in age, sex, and language/accent. Notably, the ability to predict physical and mental exhaustion improved when analysing data from the same individuals over time.

Sponsored by Innovate UK, this study highlights the potential of speech analysis as a simple and effective tool for real-time sleepiness and fatigue monitoring, paving the way for early intervention strategies in populations at risk.

Paper

17 June 2024

Predicting different dimensions of fatigue from speech data: a longitudinal study in shift workers

This research study examines how speech data can be used to predict fatigue levels in shift workers. Over two weeks, 1,197 participants from six countries across various shift work patterns and industries provided speech samples twice daily, making this one of the most comprehensive investigations of its kind.

The research found that speech characteristics could reliably indicate sleep deprivation and feelings of sleepiness from a single speech activity, in models which properly accounted for individual differences in age, sex, and language/accent. Notably, the ability to predict physical and mental exhaustion improved when analysing data from the same individuals over time.

Sponsored by Innovate UK, this study highlights the potential of speech analysis as a simple and effective tool for real-time sleepiness and fatigue monitoring, paving the way for early intervention strategies in populations at risk.

Paper

1 February 2024

Assessing Depression and Anxiety in Elderly Brazilians: Transferability of AI Models Trained on a Non-Verbal Working Memory Task

The current study aims to illustrate the preliminary transferability of two established AI models designed for detecting depression and anxiety.

The models were initially trained on data from a non-verbal working memory game (1- and 2-back) in the thymia dataset, encompassing over 6,000 participants from the United Kingdom, United States, Mexico, Spain, and Indonesia.

This study successfully demonstrates the preliminary transferability of two AI models trained on a non-verbal working memory task, one for depression and the other for anxiety classification, to a novel demographic of elderly Brazilian adults.

Paper

1 February 2024

Assessing Depression and Anxiety in Elderly Brazilians: Transferability of AI Models Trained on a Non-Verbal Working Memory Task

The current study aims to illustrate the preliminary transferability of two established AI models designed for detecting depression and anxiety.

The models were initially trained on data from a non-verbal working memory game (1- and 2-back) in the thymia dataset, encompassing over 6,000 participants from the United Kingdom, United States, Mexico, Spain, and Indonesia.

This study successfully demonstrates the preliminary transferability of two AI models trained on a non-verbal working memory task, one for depression and the other for anxiety classification, to a novel demographic of elderly Brazilian adults.

Paper

31 January 2024

Path Signature Representation of Patient-Clinician Interactions as a Predictor for Neuropsychological Tests Outcomes in Children

This research report presents a proof-of-concept study on the application of machine learning techniques to video and speech data collected during diagnostic cognitive assessments of children with a neurodevelopmental disorder.

The study utilised a dataset of 39 video recordings, capturing extensive sessions where clinicians administered, among other things, four cognitive assessment tests. From the first 40 minutes of each clinical session, covering the administration of the Wechsler Intelligence Scale for Children (WISC-5), we extracted head positions and speech turns of both clinician and child.

Paper

31 January 2024

Path Signature Representation of Patient-Clinician Interactions as a Predictor for Neuropsychological Tests Outcomes in Children

This research report presents a proof-of-concept study on the application of machine learning techniques to video and speech data collected during diagnostic cognitive assessments of children with a neurodevelopmental disorder.

The study utilised a dataset of 39 video recordings, capturing extensive sessions where clinicians administered, among other things, four cognitive assessment tests. From the first 40 minutes of each clinical session, covering the administration of the Wechsler Intelligence Scale for Children (WISC-5), we extracted head positions and speech turns of both clinician and child.

Paper

22 June 2023

Bayesian Networks for the robust and unbiased prediction of depression and its symptoms utilizing speech and multimodal data

Our paper got accepted into interspeech 2023!

In this study, we apply a Bayesian framework to capture the relationships between depression, depression symptoms, and features derived from speech, facial expression and cognitive game data collected at thymia.

Paper

22 June 2023

Bayesian Networks for the robust and unbiased prediction of depression and its symptoms utilizing speech and multimodal data

Our paper got accepted into interspeech 2023!

In this study, we apply a Bayesian framework to capture the relationships between depression, depression symptoms, and features derived from speech, facial expression and cognitive game data collected at thymia.

Paper

30 March 2022

Speech and the n-Back task as a lens into depression

Our paper got accepted into interspeech 2022!

This paper presents strong evidence that embedded in any speech signal is a rich combination of cognitive, neuromuscular and physiological information. This richness makes speech a powerful signal in relation to a range of different health conditions, including major depressive disorders (MDD).

We then present a set of machine learning experiments that demonstrate that combining speech with features from an n-Back working memory assessment improves classifier performance when predicting the popular eight-item Patient Health Questionnaire depression scale (PHQ-8).

Paper

30 March 2022

Speech and the n-Back task as a lens into depression

Our paper got accepted into interspeech 2022!

This paper presents strong evidence that embedded in any speech signal is a rich combination of cognitive, neuromuscular and physiological information. This richness makes speech a powerful signal in relation to a range of different health conditions, including major depressive disorders (MDD).

We then present a set of machine learning experiments that demonstrate that combining speech with features from an n-Back working memory assessment improves classifier performance when predicting the popular eight-item Patient Health Questionnaire depression scale (PHQ-8).