Machine learning techniques in a structural and functional MRI diagnostic approach in schizophrenia: a systematic...

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Diagnostics

Background:

• Schizophrenia (SCZ) diagnosis currently relies solely on clinical assessment due to the absence of specific biomarkers.
• Machine learning (ML) offers a promising avenue to assist clinicians in diagnosing mental health conditions.

Purpose of the Study:

• To systematically review the accuracy of ML techniques in distinguishing schizophrenia patients from healthy individuals.
• To evaluate the diagnostic performance of ML models based on neuroimaging data.

Main Methods:

• A systematic literature search was conducted across major databases (PubMed, Embase, MEDLINE, PsychINFO, Cochrane Library) up to December 2018.
• Thirty-five studies utilizing structural and/or functional neuroimaging were included, focusing on ML techniques for SCZ diagnosis.
• Extracted data included sensitivity, specificity, and accuracy of the ML models.

Main Results:

• Machine learning models, particularly Support Vector Machines, achieved near 100% accuracy when combined.
• Prefrontal and temporal cortices were identified as key brain regions for SCZ diagnosis.
• ML effectively detected altered brain connectivity in SCZ patients across various networks (e.g., DMN, salience network).

Conclusions:

• Integrated ML models demonstrate high accuracy, crucial for early SCZ diagnosis and treatment response evaluation.
• Future challenges involve integrating ML algorithms with clinical evaluation for improved patient outcomes.
• ML-based neuroimaging analysis holds significant potential for objective SCZ diagnosis and prognosis.