Deep Learning Analyses of Brain MRI to Identify Sustained Attention Deficit in Treated Obstructive Sleep Apnea: A...

Area of Science:

• Neuroimaging
• Artificial Intelligence
• Sleep Medicine

Background:

• Persistent sustained attention deficit (SAD) significantly impacts quality of life and occupational function in obstructive sleep apnea (OSA) patients despite continuous positive airway pressure (CPAP) treatment.
• Predicting SAD in OSA patients undergoing CPAP therapy remains challenging, hindering timely intervention.
• Brain magnetic resonance (MR) imaging offers potential biomarkers for predicting treatment outcomes.

Purpose of the Study:

• To investigate the efficacy of deep learning models in predicting persistent SAD from brain MR images in CPAP-treated OSA patients.
• To identify potential neuroanatomical correlates of persistent SAD using advanced imaging analysis.
• To establish a predictive tool for early identification of high-risk individuals.

Main Methods:

• Secondary analysis of brain MR images from 26 middle-aged men with OSA using CPAP for >6 hours daily.
• Application of a Convolutional Neural Network (CNN) model for classifying MR images into SAD (+SAD) and no SAD (-SAD) categories.
• SAD was defined by >2 lapses on the psychomotor vigilance task.

Main Results:

• The CNN model achieved high accuracy in classifying MR images: 97.02±0.80% at the image level.
• Participant-level accuracy reached 99.11±0.55% with a 90% probability threshold.
• Stable image-level accuracy of 97.45±0.63% was demonstrated.

Conclusions:

• Deep learning models, specifically CNNs, can accurately predict persistent SAD from brain MR images in OSA patients.
• This predictive capability may facilitate early initiation of adjunctive treatments to improve patient outcomes.
• Future research incorporating explainable AI could reveal neuroanatomical insights and novel therapeutic targets for SAD.