fNIRS Classification of Adults with ADHD Enhanced by Feature Selection

Area of Science:

• Neuroscience
• Psychiatry
• Machine Learning Applications

Background:

• Adult attention deficit hyperactivity disorder (ADHD) is a common condition with significant functional impairments.
• Adult ADHD is often under-prioritized compared to childhood ADHD, necessitating improved diagnostic approaches.
• Functional near-infrared spectroscopy (fNIRS) offers a non-invasive method for brain activity assessment.

Purpose of the Study:

• To develop and validate a machine learning model for differentiating adults with ADHD from healthy controls using fNIRS data.
• To enhance the accuracy of fNIRS-based ADHD diagnosis through efficient feature selection in high-dimensional datasets.
• To introduce a novel hybrid feature selection method, Bayesian-Tuned Ridge RFECV (BTR-RFECV), for improved diagnostic performance.

Main Methods:

• Utilized functional near-infrared spectroscopy (fNIRS) to record brain activity during verbal fluency tasks in 120 adults with ADHD and 75 healthy controls.
• Implemented a hybrid feature selection approach, Bayesian-Tuned Ridge RFECV (BTR-RFECV), combining wrapper and embedded methods for high-dimensional fNIRS data.
• Trained and evaluated machine learning models using selected HbO features from frontal and temporal brain regions.

Main Results:

• The BTR-RFECV method effectively reduced feature dimensionality and optimized hyperparameter tuning for fNIRS data.
• The developed models achieved high diagnostic performance, including precision (89.89%), recall (89.74%), and accuracy (89.74%).
• Key predictive features were identified within HbO signals from the combined frontal and temporal regions.

Conclusions:

• The combination of fNIRS and machine learning, particularly with the BTR-RFECV feature selection, shows significant promise for diagnosing adult ADHD.
• This approach offers a potential pathway to reduce manual intervention in clinical diagnostic settings.
• Further validation in diverse clinical populations is warranted to establish widespread utility.