Context
The ADOS-2 (Autism Diagnostic Observation Schedule) serves as a comprehensive diagnostic tool for Autism Spectrum Disorder (ASD). While highly effective, its administration incurs significant financial costs and delays due to a variety of factors. Moreover, the complexity of the assessment requires specialized training for administration and scoring, leading to limited availability in some states. Furthermore, the escalating prevalence of ASD in the United States has exacerbated this issue within the healthcare sector, primarily due to the significant financial burdens associated with diagnosis and treatment. Consequently, many young children with ASD encounter significant obstacles in obtaining timely diagnosis, crucial for accessing essential services, often due to factors such as geographic proximity, prolonged wait times, and financial constraints.
Focus
My research focuses on streamlining the current gold-standard assessment tool for ASD, the ADOS-2. The lengthy and expensive ature of the ADOS-2, coupled with limited availability of trained professionals, creates a significant barrier for many families.
My project aims to leverage machine learning to identify a more efficient and accessible version of the ADOS-2. Through analysis of existing data, I will identify key dimensions within the assessment that predict overall severity. This could lead to a shorter, more cost-effective tool, ultimately reducing wait times and financial burdens for families seeking diagnoses and crucial interventions.
Responsibilities
I'll be shouldering the responsibility of analyzing a pre-existing dataset on individuals with ASD. My mission is to utilize machine learning models, such as the XGBoost regressor and K-means clustering, to uncover hidden patterns within the data. Additionally, I'll be employing statistical analysis to assess the significance of my findings.
By wielding these tools, I aim to pinpoint the most crucial dimensions of the ADOS-2 assessment that hold the key to diagnosing ASD. Ultimately, the project seeks to determine if a smaller set of highly predictive items can be identified, allowing for a streamlined version of the ADOS-2. This, in turn, could revolutionize access to diagnoses, removing barriers for families and paving the way for timely interventions.
All SVGs and background images fall under fairuse under the unDraw and unSplash license