![Brain Connectivity]
![ADHD]
![Sex Differences]
This repository contains my dissertation project for the Women in Data Science (WiDS) Datathon 2025 challenge. The project focuses on developing machine learning models to predict ADHD diagnosis and biological sex using functional brain imaging data of children and adolescents, along with their socio-demographic, emotional, and parenting information.
"What brain activity patterns are associated with ADHD; are they different between males and females, and, if so, how?"
The analysis provides insights into sex-specific differences in ADHD manifestation, which could improve diagnosis accuracy, especially in females who are often underdiagnosed due to presenting more inattentive symptoms.
This project uses data from the Healthy Brain Network (HBN), the signature scientific initiative of the Child Mind Institute. The dataset includes:
- Functional MRI Connectome Matrices: Brain connectivity data showing correlations between different brain regions
- Categorical Metadata: Socio-demographic information (e.g., handedness, parent's education)
- Quantitative Metadata: Scores from assessments including:
- Strength and Difficulties Questionnaire (SDQ)
- Alabama Parenting Questionnaire
The models predict two target variables:
- ADHD_Outcome: Type of Diagnosis (0=Other/None, 1=ADHD)
- Sex_F: Sex of participant (0=Male, 1=Female)
- **ML_Powered_ADHD.ipynb**: Main analysis notebook containing all code, analysis, and results for the ADHD prediction project
- Multi-outcome prediction of both ADHD diagnosis and biological sex
- Enhanced model comparison with statistical significance testing
- Sex-specific analysis of brain connectivity patterns in ADHD
- Interactive visualizations of brain connectivity networks
- SHAP-based interpretation of model predictions
- Feature importance analysis highlighting sex differences