Related Experiment Videos
Neural network simulations, cortical connectivity, and schizophrenic psychosis
1Yale Psychiatric Institute, New Haven, CT 06520-8038, USA. hoffman@biomed.med.yale.edu
M.D. Computing : Computers in Medical Practice
|May 1, 1997
Summary
Neural network simulations reveal how reduced brain connectivity in schizophrenia may cause "schizophrenic-like" symptoms, including hallucinations and delusions. This research models positive and negative symptoms and potential drug mechanisms.
Area of Science:
- Neuroscience
- Computational Psychiatry
- Artificial Intelligence
Background:
- Schizophrenia is increasingly linked to reduced corticocortical connectivity.
- Understanding the functional consequences of this reduced connectivity is crucial.
Purpose of the Study:
- To model schizophrenia-like behaviors using neural network simulations with reduced connectivity.
- To investigate the mechanisms underlying positive and negative symptoms of schizophrenia.
Main Methods:
- Simulated parallel, distributed processing systems with varying degrees of connectivity pruning.
- Utilized attractor network models to explore functional fragmentation and intrusive representations.
- Employed backpropagation simulations of speech perception networks to model auditory hallucinations.
Main Results:
- Excessive pruning led to functionally fragmented networks, mimicking "loose associations."
- Recurrent, intrusive representations emerged, modeling delusions.
- Simulated speech perception networks generated spontaneous output, modeling auditory hallucinations ("voices").
- The models suggested mechanisms for dopamine-blocking drugs reducing positive symptoms and the emergence of negative symptoms.
Conclusions:
- Reduced corticocortical connectivity can computationally explain core symptoms of schizophrenia.
- Neural network models offer insights into the pathophysiology of schizophrenia and the effects of antipsychotic medications.