Schizophrenia II: Computational models Flashcards

1
Q

What characterizes the Data Driven Approach to Computational Psychiatry? Give 2 examples from the field of schizophrenia research.

A
  • Agnostic to theory
  • Machine learning
    Diagnostic (data driven): 75% success if sMRI and fMRI are combined
    Treatment success (data driven): 80% percent success
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

What characterizes the Theory Driven Approach to Computational Psychiatry?

A
  • Based on conceptual models + prior evidence
  • Formal math. Models of biological/mental processes
  • Synthesizing different levels of evidence and explanation
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

What are the 3 types of theory driven models?

A
  1. Synthetic (biophysical)
  2. Algorythmic
  3. Optimal (Bayesian)
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

Give an example of how Syntehtic Models are used in schiz. research.

A

Example: NMDA receptor Antagonists like PCP and Ketamine = psychotic symptoms –> simulate psychosis

Finding: damped lateral inhibition = mediating factor between psychotic symptoms and biophysical causes

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

Give an example of the use of algorythmic models in schizo. research.

A

Example: Schizophr. = hard to learn from positive outcomes (i.e. problem With reward based learning): could be: cant judge positive to be so (WM in OFC), or don’t learn from it (Dopamine).
Reinforcement learning used to model: actor- critic and Q learning

finding: negative systems: not related to reduced learning from positive outcomes as such; but with failure to represent positive value
High negative symptoms -> Actor critique model fit

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

What aspect of Schizophrenia can be explained using optimal (Bayesian) models? Explain the Framework.

A

abnormality in brains (Bayesian) inference mechanism leading to failure to integrate new evidence leading to false predictions.
Predictive Coding = decreased precision of prior beliefs, increased precision of sensory data =increase in prediction error

Hallucinations = shift away from prior beliefs and toward sensory evidence = Reduced prior-to-likelihood ratio.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly