Predicting the conversion time from normal cognition to mild cognitive impairment based on dual attention...

Area of Science:

• Neurology
• Artificial Intelligence
• Medical Diagnostics

Background:

• Dementia is a progressive neurological disorder and a leading global cause of death.
• Early diagnosis, prevention, and treatment are critical for managing dementia's impact.
• Dementia is often underdiagnosed in its initial stages.

Purpose of the Study:

• To precisely predict the conversion time from normal cognition (NC) to mild cognitive impairment (MCI).
• To offer insights for the early diagnosis and treatment of dementia.
• To develop a model for predicting neurodegenerative disease progression.

Main Methods:

• A novel dual attention convolutional network was developed.
• The model integrates feature and temporal attention modules for dependency capture.
• A custom loss function was employed to improve clinical interpretability.

Main Results:

• The model significantly improved prediction accuracy, reducing MSE by 9.67% and MAE by 26.24%.
• The model demonstrated a 16.71% increase in R-squared compared to a basic convolutional model.
• The model successfully predicted NC to MCI conversion, guiding early intervention strategies.

Conclusions:

• The dual attention convolutional network is an effective tool for predicting NC to MCI conversion.
• This model provides valuable support for early dementia diagnosis.
• The findings contribute to advancing the management of neurodegenerative conditions.