Machine learning based classification of excessive smartphone users via neuronal cue reactivity

Area of Science:

• Neuroscience
• Psychiatry
• Computer Science

Background:

• Excessive Smartphone Use (ESU) is a growing societal concern with unclear diagnostic criteria.
• Understanding the neural underpinnings of ESU is crucial for developing effective interventions.
• Existing research lacks definitive neural markers for classifying ESU.

Purpose of the Study:

• To classify individuals with Excessive Smartphone Use (ESU) versus non-excessive users (n-ESU) using neuroimaging.
• To identify neural Cue-Reactivity (CR) signatures differentiating ESU from n-ESU.
• To explore similarities between ESU neural patterns and those of established addictive disorders.

Main Methods:

• Functional magnetic resonance imaging (fMRI) data was collected during a Cue-Reactivity (CR) task.
• Machine learning algorithms, including Support Vector Machines (SVM), were employed for classification.
• Feature selection (RFE, Model-based) and dimensionality reduction (PCA) were used to optimize model performance with high-dimensional fMRI data.

Main Results:

• The classification model achieved up to 79.9% accuracy in distinguishing ESU from n-ESU.
• Specific brain regions showed significant activation, potentially serving as biomarkers for ESU.
• Spatial similarities were found between ESU neural patterns and those of other addictive disorders.

Conclusions:

• Machine learning effectively identifies neural correlates of Excessive Smartphone Use (ESU).
• The findings highlight potential neurobiological markers for ESU, informing diagnostic and therapeutic strategies.
• Further research with larger datasets is needed to enhance model generalizability and stability for clinical applications.