The ATIA engineers have developed and are still expanding a set of tools that can be used for inductive and deductive reasoning:
These tools are especially useful to solve complex problems in the biomedical field, as the past three years of applying them have demonstrated.
They can discover interactions, meaning that one variable, for example anxiety, is only relevant if another variable, for example a certain type of brain activity, does not indicate a negative treatment outcome to start with.
The tools can also discover non-linear effects, for example that a certain rehabilitative treatment can not only start too late, but also too early after surgery.
The output of one tool/agent, may be used as input for another tool/agent, thus creating a multi agent system.
Since no two datasets are the same and no two problems are identical, ATIA has built a platform that allows us to do Advanced Analytics that we have named ADVANA. The software of this platform finds characteristics in complex datasets and provides advisory information about new ( individual) cases to clinicians and patiensts. Using ADVANA, we can quickly build a reasoning configuration that prepares your data, mines your data for patterns and rules and provides the means to implement these as Decision support in your systems.
New treatment approaches were found when ATIA analysed treatment outcome of combined repetitive Transcranial Magnetic Stimulation with psychotherapy in depressive patients for Brainclinics.
As expected for such a combined treatment, the predictors of success were both biological and psychological. However, the major gain here was the prediction of non-response, which appeared to depend on psychological (co-morbid anxiety) and biological (brainwaves indicative of impaired vigilance regulation) predictors. The latter could be interpreted to suggest psychostimulant treatment, this matches a recent finding that EEG-guided treatment of depression with psychostimulants is more effective than treatment by protocol.
Decision trees could give valuable information about responsers and non-responders.
ATIA has made predictive models that show that valuable tests, are ordered for diagnostic purposes, do not contribute information to take the required clinical decision: a chance for cost reduction in health care.
The ATIA tools also support our own applied biomedical research: Dr. Koen Böcker is using homemade software that analyzes thousands of voxels in a brainscan image for non-linear associations with personal characteristics related to intelligence, emotional regulation, eating and sexual behavior. Here is an example image of a brain scan showing a positive relation to social rise relative to parents.