Publicação
Identification of Subtypes of Autism Spectrum Disorder Patients using Machine Learning Methods
| Resumo: | Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder affecting brain structure and neuronal connectivity, characterized by a shortage of social interaction and communication, as well as repetitive patterns of behavior and restricted interests. Highly heterogeneous clinical features pose great challenges for ASD diagnosis, such that children who receive a diagnosis of ASD have a range of vastly different presentations, trajectories, and outcomes. Identifying ASD subgroups can be helpful for researchers and clinicians to gain insights into distinct characteristics and patterns within these groups, as well as identify specific factors that may influence long-term outcomes. Moreover, it can also contribute to advancing scientific knowledge about ASD. It allows researchers to explore the underlying mechanisms, genetic factors, environmental influences, and brain processes that may be specific to each subgroup. This deeper understanding can lead to more targeted research efforts, improved diagnostic tools, and the development of innovative therapies for different subgroups. This study used Hierarchical Clustering on Principal Component (HCPC) to analyze a sample of 661 children, aged 1-8 years, diagnosed with ASD following DSM-IV or DSM-V criteria and reaching the thresholds for ASD from the Autism Diagnostic Interview–Revised and the Autism Diagnostic Observation Schedule. In total, 11 variables were selected for cluster analysis, which included, apart from the diagnostic/screening ones, measures that can capture variations in children’s development, such as communication and social abilities assessed through Vineland Adaptive Behavior Scales and The Griffiths Mental Development Scales. Our analysis identified three distinct subgroups based on multiple developmental and behavioral domains. Cluster 1 exhibited lower levels of intellectual and adaptive abilities, accompanied by more severe social symptoms, repetitive behaviors, and developmental abnormalities. In comparison, Cluster 2 displayed similar levels of developmental abnormalities as Cluster 1, but demonstrated higher severity in social interactions, communication, and adaptive behavior than Cluster 3. On the other hand, Cluster 3 showcased the highest scores in language and adaptive abilities, and presented the lowest severity across social and developmental symptoms among all three clusters, indicating the least impairments. These findings emphasize the importance of considering multiple developmental and behavioral domains, as well as core symptoms of autism, in order to distinguish subgroups of young children with ASD and provide more comprehensive developmental profiles. |
|---|---|
| Autores principais: | Almeida, Ana Sofia Sousa Teixeira de |
| Assunto: | Perturbação do espectro do autismo Diagnóstico Análise estatística Aprendizagem automática Clusters Teses de mestrado - 2023 |
| Ano: | 2023 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso restrito |
| Instituição associada: | Universidade de Lisboa |
| Idioma: | inglês |
| Origem: | Repositório da Universidade de Lisboa |
| Resumo: | Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder affecting brain structure and neuronal connectivity, characterized by a shortage of social interaction and communication, as well as repetitive patterns of behavior and restricted interests. Highly heterogeneous clinical features pose great challenges for ASD diagnosis, such that children who receive a diagnosis of ASD have a range of vastly different presentations, trajectories, and outcomes. Identifying ASD subgroups can be helpful for researchers and clinicians to gain insights into distinct characteristics and patterns within these groups, as well as identify specific factors that may influence long-term outcomes. Moreover, it can also contribute to advancing scientific knowledge about ASD. It allows researchers to explore the underlying mechanisms, genetic factors, environmental influences, and brain processes that may be specific to each subgroup. This deeper understanding can lead to more targeted research efforts, improved diagnostic tools, and the development of innovative therapies for different subgroups. This study used Hierarchical Clustering on Principal Component (HCPC) to analyze a sample of 661 children, aged 1-8 years, diagnosed with ASD following DSM-IV or DSM-V criteria and reaching the thresholds for ASD from the Autism Diagnostic Interview–Revised and the Autism Diagnostic Observation Schedule. In total, 11 variables were selected for cluster analysis, which included, apart from the diagnostic/screening ones, measures that can capture variations in children’s development, such as communication and social abilities assessed through Vineland Adaptive Behavior Scales and The Griffiths Mental Development Scales. Our analysis identified three distinct subgroups based on multiple developmental and behavioral domains. Cluster 1 exhibited lower levels of intellectual and adaptive abilities, accompanied by more severe social symptoms, repetitive behaviors, and developmental abnormalities. In comparison, Cluster 2 displayed similar levels of developmental abnormalities as Cluster 1, but demonstrated higher severity in social interactions, communication, and adaptive behavior than Cluster 3. On the other hand, Cluster 3 showcased the highest scores in language and adaptive abilities, and presented the lowest severity across social and developmental symptoms among all three clusters, indicating the least impairments. These findings emphasize the importance of considering multiple developmental and behavioral domains, as well as core symptoms of autism, in order to distinguish subgroups of young children with ASD and provide more comprehensive developmental profiles. |
|---|