Identifying Methamphetamine Abstainers With Convolutional Neural Networks and Short-Time Fourier Transform

Area of Science:

• Neuroscience
• Neuroimaging
• Machine Learning

Background:

• Understanding the neurobiological mechanisms in methamphetamine abstainers is crucial for explaining abnormal behaviors.
• Limited research exists on the functional brain patterns of individuals in long-term methamphetamine abstinence.

Purpose of the Study:

• To identify and differentiate functional brain patterns between male methamphetamine abstainers and healthy controls.
• To elucidate the pathological mechanisms underlying methamphetamine abstinence using novel analytical techniques.

Main Methods:

• Functional magnetic resonance imaging (fMRI) was employed on 42 male methamphetamine abstainers (abstinence ≥ 14 months) and 32 male healthy controls.
• A combination of convolutional neural network (CNN) and short-time Fourier transform (STFT) was utilized to analyze time-frequency spectrograms of brain activity.
• The STFT provided time-localized frequency information, while the CNN extracted structural features from these spectrograms.

Main Results:

• The developed classifier achieved high accuracy (98.9%) in distinguishing between methamphetamine abstainers and controls.
• Highly discriminative brain voxels were identified, primarily located in the left inferior orbital frontal gyrus, bilateral postcentral gyri, and bilateral paracentral lobules.
• The analysis successfully extracted robust brain voxel information, highlighting specific regional differences.

Conclusions:

• This study offers novel insights into the distinct functional brain patterns differentiating methamphetamine abstainers from healthy individuals.
• The findings suggest that time-frequency spectrogram analysis can elucidate the pathological mechanisms associated with methamphetamine abstinence.
• Identifying specific brain regions involved provides a foundation for understanding behavioral abnormalities in this population.