Machine Learning Applied to Brain SPECT Scans Distinguishes Schizophrenia in New Study

In a press release, researchers from the TReNDS Center and Amen Clinics described a study showing that whole-brain Single Photon Emission Computed Tomography (SPECT) imaging combined with machine learning can distinguish individuals with schizophrenia from healthy subjects.

The study, published in NeuroImage: Reports, analyzed brain scans from 213 participants, including 137 with schizophrenia and 76 healthy controls. Researchers used spatially constrained Independent Component Analysis guided by the NeuroMark functional brain network template to identify patterns linked to schizophrenia.

Among several tested algorithms, logistic regression achieved 87 percent sensitivity and 68 percent specificity, while random forest reached 88 percent sensitivity and 61 percent specificity. These models outperformed Support Vector Machine approaches.

Key predictive brain regions included the middle occipital gyrus, subthalamus, and putamen. The findings suggest that network-level abnormalities in visual processing and cognitive control play a significant role in distinguishing schizophrenia patients. Future studies are planned to validate these results with larger and more balanced populations.