MF
Maarten A. Frens
info
Please Note
<p>This page displays the records of the person named above and is not linked to a unique person identifier. This record may need to be merged to a profile.</p>
3 records found
1
Visual Cues to Autism
How Gaze Behaviour Reflects Autistic Traits in Children
Autism Spectrum Disorder (ASD) is a neurodevelopmental condition characterised by deficits in social interaction and restrictive and repetitive behaviours. ASD is often accompanied by anxiety, elevated rates of depression, and reduced quality of life. Current diagnostic tools are time-consuming, subjective, and less effective in certain groups, particularly young children. Eye tracking has emerged as a promising biomarker, holding the potential to improve diagnostic accuracy and developmental outcomes, thereby reducing the lifetime socio-economic costs of autism. This thesis aims to advance the understanding of gaze behaviour as a biomarker for ASD, thereby driving the development of more accurate, accessible, and scalable diagnostic tools.
A literature review identified fifteen eye tracking features that are associated with ASD diagnosis or Social Responsiveness Scale (SRS) scores. These features were extracted from raw eye tracking data in the Generation R database. Statistical analyses were performed using Generalised Linear Models to evaluate their relationship with ASD diagnosis and SRS scores. Additional analyses addressed feature distributions, gender differences, and the effects of video content and participant age and gender. Features with significant relationships with ASD diagnosis or SRS scores were used in a nested cross-validation framework with predictive machine learning models. The Area Under the Curve (AUC) was the primary performance metric, supported by F1 score, precision, and recall.
Classification proved challenging due to limited discriminative power of individual features. The CatBoost Gradient Boosting Decision Tree achieved the highest performance for ASD prediction with an AUC of 0.71, indicating that gaze-derived features hold promise when used in complex non-linear models. In contrast, the model predicting SRS scores performed worse, with an AUC of 0.57, suggesting that social responsiveness is a part of ASD that may be more effectively masked in adolescents. An alternative explanation is that the features reflect aspects of the autism phenotype that are unrelated to social responsiveness.
Key limitations include the small number of 33 participants with an ASD diagnosis, which reduced statistical power. The reliance on pre-existing clinical diagnoses and SRS scores, which does not capture the full complexity of ASD, led to a lack of ground truth. The adolescent age of the participants posed limitations, as masking behaviours can obscure gaze-based markers. Co-occurring factors such as attention difficulties or cognitive ability were not controlled for. Finally, choices in feature engineering, such as gaze data aggregation, reduced temporal detail; and exclusion of participants with very low screentime may have caused individuals with low social engagement to be underrepresented.
In summary, gaze analysis is a promising tool for understanding and identifying ASD. The findings in this thesis suggest that while individual gaze features offer limited diagnostic power on their own, their integration within more advanced models holds potential to improve the diagnostic process and provide deeper insights into the mechanisms underlying ASD. As datasets grow in size and quality, and stimuli continue to evolve, there are substantial opportunities to uncover more nuanced relationships between gaze behaviour and ASD traits. These developments ultimately support the creation of low-cost, inclusive and scalable ASD assessment tools, thereby enhancing both individual quality of life and broader socio-economic outcomes. ...
A literature review identified fifteen eye tracking features that are associated with ASD diagnosis or Social Responsiveness Scale (SRS) scores. These features were extracted from raw eye tracking data in the Generation R database. Statistical analyses were performed using Generalised Linear Models to evaluate their relationship with ASD diagnosis and SRS scores. Additional analyses addressed feature distributions, gender differences, and the effects of video content and participant age and gender. Features with significant relationships with ASD diagnosis or SRS scores were used in a nested cross-validation framework with predictive machine learning models. The Area Under the Curve (AUC) was the primary performance metric, supported by F1 score, precision, and recall.
Classification proved challenging due to limited discriminative power of individual features. The CatBoost Gradient Boosting Decision Tree achieved the highest performance for ASD prediction with an AUC of 0.71, indicating that gaze-derived features hold promise when used in complex non-linear models. In contrast, the model predicting SRS scores performed worse, with an AUC of 0.57, suggesting that social responsiveness is a part of ASD that may be more effectively masked in adolescents. An alternative explanation is that the features reflect aspects of the autism phenotype that are unrelated to social responsiveness.
Key limitations include the small number of 33 participants with an ASD diagnosis, which reduced statistical power. The reliance on pre-existing clinical diagnoses and SRS scores, which does not capture the full complexity of ASD, led to a lack of ground truth. The adolescent age of the participants posed limitations, as masking behaviours can obscure gaze-based markers. Co-occurring factors such as attention difficulties or cognitive ability were not controlled for. Finally, choices in feature engineering, such as gaze data aggregation, reduced temporal detail; and exclusion of participants with very low screentime may have caused individuals with low social engagement to be underrepresented.
In summary, gaze analysis is a promising tool for understanding and identifying ASD. The findings in this thesis suggest that while individual gaze features offer limited diagnostic power on their own, their integration within more advanced models holds potential to improve the diagnostic process and provide deeper insights into the mechanisms underlying ASD. As datasets grow in size and quality, and stimuli continue to evolve, there are substantial opportunities to uncover more nuanced relationships between gaze behaviour and ASD traits. These developments ultimately support the creation of low-cost, inclusive and scalable ASD assessment tools, thereby enhancing both individual quality of life and broader socio-economic outcomes. ...
Autism Spectrum Disorder (ASD) is a neurodevelopmental condition characterised by deficits in social interaction and restrictive and repetitive behaviours. ASD is often accompanied by anxiety, elevated rates of depression, and reduced quality of life. Current diagnostic tools are time-consuming, subjective, and less effective in certain groups, particularly young children. Eye tracking has emerged as a promising biomarker, holding the potential to improve diagnostic accuracy and developmental outcomes, thereby reducing the lifetime socio-economic costs of autism. This thesis aims to advance the understanding of gaze behaviour as a biomarker for ASD, thereby driving the development of more accurate, accessible, and scalable diagnostic tools.
A literature review identified fifteen eye tracking features that are associated with ASD diagnosis or Social Responsiveness Scale (SRS) scores. These features were extracted from raw eye tracking data in the Generation R database. Statistical analyses were performed using Generalised Linear Models to evaluate their relationship with ASD diagnosis and SRS scores. Additional analyses addressed feature distributions, gender differences, and the effects of video content and participant age and gender. Features with significant relationships with ASD diagnosis or SRS scores were used in a nested cross-validation framework with predictive machine learning models. The Area Under the Curve (AUC) was the primary performance metric, supported by F1 score, precision, and recall.
Classification proved challenging due to limited discriminative power of individual features. The CatBoost Gradient Boosting Decision Tree achieved the highest performance for ASD prediction with an AUC of 0.71, indicating that gaze-derived features hold promise when used in complex non-linear models. In contrast, the model predicting SRS scores performed worse, with an AUC of 0.57, suggesting that social responsiveness is a part of ASD that may be more effectively masked in adolescents. An alternative explanation is that the features reflect aspects of the autism phenotype that are unrelated to social responsiveness.
Key limitations include the small number of 33 participants with an ASD diagnosis, which reduced statistical power. The reliance on pre-existing clinical diagnoses and SRS scores, which does not capture the full complexity of ASD, led to a lack of ground truth. The adolescent age of the participants posed limitations, as masking behaviours can obscure gaze-based markers. Co-occurring factors such as attention difficulties or cognitive ability were not controlled for. Finally, choices in feature engineering, such as gaze data aggregation, reduced temporal detail; and exclusion of participants with very low screentime may have caused individuals with low social engagement to be underrepresented.
In summary, gaze analysis is a promising tool for understanding and identifying ASD. The findings in this thesis suggest that while individual gaze features offer limited diagnostic power on their own, their integration within more advanced models holds potential to improve the diagnostic process and provide deeper insights into the mechanisms underlying ASD. As datasets grow in size and quality, and stimuli continue to evolve, there are substantial opportunities to uncover more nuanced relationships between gaze behaviour and ASD traits. These developments ultimately support the creation of low-cost, inclusive and scalable ASD assessment tools, thereby enhancing both individual quality of life and broader socio-economic outcomes.
A literature review identified fifteen eye tracking features that are associated with ASD diagnosis or Social Responsiveness Scale (SRS) scores. These features were extracted from raw eye tracking data in the Generation R database. Statistical analyses were performed using Generalised Linear Models to evaluate their relationship with ASD diagnosis and SRS scores. Additional analyses addressed feature distributions, gender differences, and the effects of video content and participant age and gender. Features with significant relationships with ASD diagnosis or SRS scores were used in a nested cross-validation framework with predictive machine learning models. The Area Under the Curve (AUC) was the primary performance metric, supported by F1 score, precision, and recall.
Classification proved challenging due to limited discriminative power of individual features. The CatBoost Gradient Boosting Decision Tree achieved the highest performance for ASD prediction with an AUC of 0.71, indicating that gaze-derived features hold promise when used in complex non-linear models. In contrast, the model predicting SRS scores performed worse, with an AUC of 0.57, suggesting that social responsiveness is a part of ASD that may be more effectively masked in adolescents. An alternative explanation is that the features reflect aspects of the autism phenotype that are unrelated to social responsiveness.
Key limitations include the small number of 33 participants with an ASD diagnosis, which reduced statistical power. The reliance on pre-existing clinical diagnoses and SRS scores, which does not capture the full complexity of ASD, led to a lack of ground truth. The adolescent age of the participants posed limitations, as masking behaviours can obscure gaze-based markers. Co-occurring factors such as attention difficulties or cognitive ability were not controlled for. Finally, choices in feature engineering, such as gaze data aggregation, reduced temporal detail; and exclusion of participants with very low screentime may have caused individuals with low social engagement to be underrepresented.
In summary, gaze analysis is a promising tool for understanding and identifying ASD. The findings in this thesis suggest that while individual gaze features offer limited diagnostic power on their own, their integration within more advanced models holds potential to improve the diagnostic process and provide deeper insights into the mechanisms underlying ASD. As datasets grow in size and quality, and stimuli continue to evolve, there are substantial opportunities to uncover more nuanced relationships between gaze behaviour and ASD traits. These developments ultimately support the creation of low-cost, inclusive and scalable ASD assessment tools, thereby enhancing both individual quality of life and broader socio-economic outcomes.
Introduction: Motor adaptation is the process of adapting a movement plan to unexpected results due to a changing environment or changes in physical performance. Using the theory of optimal forward control this process can be described using using a forward control model with two learning parameters and two noise factors. This study will use Bayesian inference and a No-U-Turn sampler to estimate these learning and noise parameters from movement data from a reaching task in a small (N=60) and a large dataset (N=2226).
Methods: In total, six models were created and tested following a state-space model for adaptation. Three models used the same hierarchical design for the learning parameters and compared different hierarchical approaches for priors for planning noise (ση and execution noise (σϵ). The other three models used the same non-hierarchical design for the noise parameters and compared different hierarchical hyperpriors and a non-hierarchical design for the learning rate (A) and adaptation rate (B).
Results: For the smaller dataset, the same issue was seen for all model designs, where the posterior distributions of ση are heavily skewed toward 0, with the HMC unable to converge (^R
> 1.1). For the large dataset a non-hierarchical approach for both the learning parameters and noise parameters was able to converge for the majority of subjects for all four parameters.
Discussion: While none of the models performed perfectly, in this paper a model is created that is able to quantify motor adaptation and motor noise from visuomotor task data in a large dataset of 2226 subjects. In the future this model can be used for further research into neurological and genetic factors that influence motor adaptation. ...
Methods: In total, six models were created and tested following a state-space model for adaptation. Three models used the same hierarchical design for the learning parameters and compared different hierarchical approaches for priors for planning noise (ση and execution noise (σϵ). The other three models used the same non-hierarchical design for the noise parameters and compared different hierarchical hyperpriors and a non-hierarchical design for the learning rate (A) and adaptation rate (B).
Results: For the smaller dataset, the same issue was seen for all model designs, where the posterior distributions of ση are heavily skewed toward 0, with the HMC unable to converge (^R
Discussion: While none of the models performed perfectly, in this paper a model is created that is able to quantify motor adaptation and motor noise from visuomotor task data in a large dataset of 2226 subjects. In the future this model can be used for further research into neurological and genetic factors that influence motor adaptation. ...
Introduction: Motor adaptation is the process of adapting a movement plan to unexpected results due to a changing environment or changes in physical performance. Using the theory of optimal forward control this process can be described using using a forward control model with two learning parameters and two noise factors. This study will use Bayesian inference and a No-U-Turn sampler to estimate these learning and noise parameters from movement data from a reaching task in a small (N=60) and a large dataset (N=2226).
Methods: In total, six models were created and tested following a state-space model for adaptation. Three models used the same hierarchical design for the learning parameters and compared different hierarchical approaches for priors for planning noise (ση and execution noise (σϵ). The other three models used the same non-hierarchical design for the noise parameters and compared different hierarchical hyperpriors and a non-hierarchical design for the learning rate (A) and adaptation rate (B).
Results: For the smaller dataset, the same issue was seen for all model designs, where the posterior distributions of ση are heavily skewed toward 0, with the HMC unable to converge (^R
> 1.1). For the large dataset a non-hierarchical approach for both the learning parameters and noise parameters was able to converge for the majority of subjects for all four parameters.
Discussion: While none of the models performed perfectly, in this paper a model is created that is able to quantify motor adaptation and motor noise from visuomotor task data in a large dataset of 2226 subjects. In the future this model can be used for further research into neurological and genetic factors that influence motor adaptation.
Methods: In total, six models were created and tested following a state-space model for adaptation. Three models used the same hierarchical design for the learning parameters and compared different hierarchical approaches for priors for planning noise (ση and execution noise (σϵ). The other three models used the same non-hierarchical design for the noise parameters and compared different hierarchical hyperpriors and a non-hierarchical design for the learning rate (A) and adaptation rate (B).
Results: For the smaller dataset, the same issue was seen for all model designs, where the posterior distributions of ση are heavily skewed toward 0, with the HMC unable to converge (^R
Discussion: While none of the models performed perfectly, in this paper a model is created that is able to quantify motor adaptation and motor noise from visuomotor task data in a large dataset of 2226 subjects. In the future this model can be used for further research into neurological and genetic factors that influence motor adaptation.
Robotic assistance for rehabilitation has benefited from the use of models for motor adaptation. The assist-as-needed paradigm for rehabilitation robotics was based on a single-state model of human adaptation to a neurological handicap. Recent studies have shown that human motor adaptation consists of two or more parallel adaptation processes. A two-state model of adaptation based on the presence of a fast process and a slow process has been widely adopted. The fast process adapts faster than the slow process but has a lower retention than the slow process. Designing training methods that can influence the individual adaptation processes could help make sure that patients retain what is desired (how to adapt to a neurological injury) and forget what is detrimental to rehabilitation (dynamics of the robotic assistance for example). The goal of this work is to design an optimal control paradigm for selectively influencing the slow and fast processes.
A feedforward discrete-time linear-quadratic tracking controller was designed for a 2-state linear time-invariant model of sensorimotor adaptation to increase the contribution of the slow process to the net adaptation at the end of training. This control signal was implemented as the sequence of visuomotor rotations in an upper-limb reaching task. This sequence of visuomotor rotations were dubbed the Adaptation-State-Tracking (AST) perturbation. The retention behaviour after this AST perturbation was compared with that after a non-adaptive (constant-level) perturbation. A between-subject comparison of the retention behaviour showed that the AST perturbation exhibited better retention than the constant-level perturbation (p=0.0415). As far as the author is aware, this is first time the 2-state Linear Time-Invariant (LTI) model has been used to design a perturbation and to predict the subsequent behaviour of the participants. The sufficiency of the control based on the 2-state LTI model and the possibility of improving retention with optimal control could positively impact the domain of robot-assisted rehabilitation.
...
A feedforward discrete-time linear-quadratic tracking controller was designed for a 2-state linear time-invariant model of sensorimotor adaptation to increase the contribution of the slow process to the net adaptation at the end of training. This control signal was implemented as the sequence of visuomotor rotations in an upper-limb reaching task. This sequence of visuomotor rotations were dubbed the Adaptation-State-Tracking (AST) perturbation. The retention behaviour after this AST perturbation was compared with that after a non-adaptive (constant-level) perturbation. A between-subject comparison of the retention behaviour showed that the AST perturbation exhibited better retention than the constant-level perturbation (p=0.0415). As far as the author is aware, this is first time the 2-state Linear Time-Invariant (LTI) model has been used to design a perturbation and to predict the subsequent behaviour of the participants. The sufficiency of the control based on the 2-state LTI model and the possibility of improving retention with optimal control could positively impact the domain of robot-assisted rehabilitation.
...
Robotic assistance for rehabilitation has benefited from the use of models for motor adaptation. The assist-as-needed paradigm for rehabilitation robotics was based on a single-state model of human adaptation to a neurological handicap. Recent studies have shown that human motor adaptation consists of two or more parallel adaptation processes. A two-state model of adaptation based on the presence of a fast process and a slow process has been widely adopted. The fast process adapts faster than the slow process but has a lower retention than the slow process. Designing training methods that can influence the individual adaptation processes could help make sure that patients retain what is desired (how to adapt to a neurological injury) and forget what is detrimental to rehabilitation (dynamics of the robotic assistance for example). The goal of this work is to design an optimal control paradigm for selectively influencing the slow and fast processes.
A feedforward discrete-time linear-quadratic tracking controller was designed for a 2-state linear time-invariant model of sensorimotor adaptation to increase the contribution of the slow process to the net adaptation at the end of training. This control signal was implemented as the sequence of visuomotor rotations in an upper-limb reaching task. This sequence of visuomotor rotations were dubbed the Adaptation-State-Tracking (AST) perturbation. The retention behaviour after this AST perturbation was compared with that after a non-adaptive (constant-level) perturbation. A between-subject comparison of the retention behaviour showed that the AST perturbation exhibited better retention than the constant-level perturbation (p=0.0415). As far as the author is aware, this is first time the 2-state Linear Time-Invariant (LTI) model has been used to design a perturbation and to predict the subsequent behaviour of the participants. The sufficiency of the control based on the 2-state LTI model and the possibility of improving retention with optimal control could positively impact the domain of robot-assisted rehabilitation.
A feedforward discrete-time linear-quadratic tracking controller was designed for a 2-state linear time-invariant model of sensorimotor adaptation to increase the contribution of the slow process to the net adaptation at the end of training. This control signal was implemented as the sequence of visuomotor rotations in an upper-limb reaching task. This sequence of visuomotor rotations were dubbed the Adaptation-State-Tracking (AST) perturbation. The retention behaviour after this AST perturbation was compared with that after a non-adaptive (constant-level) perturbation. A between-subject comparison of the retention behaviour showed that the AST perturbation exhibited better retention than the constant-level perturbation (p=0.0415). As far as the author is aware, this is first time the 2-state Linear Time-Invariant (LTI) model has been used to design a perturbation and to predict the subsequent behaviour of the participants. The sufficiency of the control based on the 2-state LTI model and the possibility of improving retention with optimal control could positively impact the domain of robot-assisted rehabilitation.