Computational Psychiatry - a proposed new integrated approach
(partial excerpt from my Master's thesis (2019): "On the implementation of Computational Psychiatry within the framework of Cognitive Psychology and Neuroscience", written at the Ohio State University by Prof. Dr. Roger Ratcliff and Prof. M.D. M.Ed. Eugene L. Arnold)
In my thesis, I propose hidden ADHD prototypes that are beyond the scope of the existent literature because of the nature of administered test procedures, and the applied analysis methods, as well as the use of pre-categorized subjects. However, the various results of existing literature hint to a spectrum of ADHD and comorbidities that could be measured using computational modeling and a sensitive test procedure proposed in my thesis. Hence, in my thesis, I propose a new approach (figure 1) to investigate ADHD and comorbidities.
Figure 1: New integrated Approach
The following paragraphs discuss differences and similarities between the proposed approach and other frameworks pursued thus far.
There exists considerable research on the behavioral characteristics of ADHD and the neural substrates underlying ADHD, but little work has been done in linking together neuronal and cognitive measures.
The considerable heterogeneity observed among ADHD individuals may have different reasons. First, the two proposed hallmarks of ADHD, inattention and hyperactivity/impulsivity, may lead to distinct cognitive dysfunctions. Second, ADHD subtypes defined by DSM-IV do not provide testable predictions. Moreover, frequent comorbidities, accompanied with ADHD, can often not be considered as sensitive neurocognitive test procedures are missing and/or controversial. Third, studies suggest that two-third of ADHD individuals diagnosed during childhood experience ADHD symptomology throughout adulthood. It is important to discuss characteristics of ADHD and comorbidities in a representative population comprised of children as well as adults, and considering females as well as males.
Sensitive Test Procedures
Evidence suggests that ADHD deficits are not detectable in all types of neurocognitive tasks. Therefore, the design of diagnostic test procedures, sensitive to different prototypes of ADHD is important.
Ratcliff Diffusion Model (RDM)
Computational modeling together with machine learning and neuronal measures could provide a toolkit to measure ADHD and comorbidities on a spectrum and based on predefined dimensions. The difference to the discrete DSM subtypes is that these dimensions are testable, and individuals are assigned on a spectrum of ADHD and comorbidities (e.g., spectrum is a continuous function S(Ter, v, a, z), where ADHD characteristics can be described by a unique combination of RDM model parameters [Ter, v, a, z]) rather than classified into discrete categories.
Thus far, the analysis of neurocognitive tasks is based on accuracy and/or response times (RTs) using summary statistics. Some studies have begun to use descriptive distributions (e.g., ex-Gaussian). However, they mostly examine RT distributions of either correct or error responses, but rarely of both.
The Ratcliff Diffusion Model (RDM; Ratcliff, 1978) seems to be promising as it is a well-validated model across a range of research fields and its parameters have validated cognitive interpretations. It also has already informed clinical research in other domains. The RDM incorporates more information than most other models as it simultaneously accounts for RTs distributions of correct and error responses as well as for accuracy. Eventually, the model may link cognitive concepts to neuronal measures. In sum, the parameters index the following cognitive properties:
Non-decision time (Ter): time for sensory input, cue retrieval and motor response
Drift rate (v): quality of evidence accumulation
Starting point (z): initial expectation favoring one decision over the other
Boundary separation (a): preference for doing task quickly rather than accurately
(e.g., degree of conservatism in decision criteria)
Contact me for more information.