Neural network models of schizophrenia

Area of Science:

• Neuroscience
• Computational Psychiatry
• Cognitive Science

Background:

• Schizophrenia is linked to reduced corticocortical connectivity in the brain.
• Understanding the functional impact of this connectivity reduction is crucial.

Purpose of the Study:

• To simulate the functional consequences of reduced corticocortical connectivity in schizophrenia using neural network models.
• To explore how these simulated abnormalities relate to specific symptoms like delusions and hallucinations.

Main Methods:

• Two neural network simulations were employed: an attractor network for memory and a speech perception network.
• A pruning rule simulating reduced connectivity was applied to these networks.
• Human patient studies using transcranial magnetic stimulation (TMS) were conducted in parallel.

Main Results:

• The attractor network model showed functional fragmentation and "parasitic" attractors, modeling delusions and cognitive disorganization under ambiguous input.
• The speech perception network simulation produced spontaneous percepts (hallucinations) and subtle narrative speech impairments with excessive pruning.
• Human patients with hallucinations showed similar speech perception impairments, and TMS confirmed its efficacy in reducing auditory hallucinations.

Conclusions:

• Neural network simulations offer testable hypotheses linking cognitive, phenomenological, and neurobiological aspects of schizophrenia.
• These models provide a framework for understanding how reduced brain connectivity contributes to core schizophrenic symptoms.
• The findings support the use of targeted interventions like TMS for symptom management.