Contributed Talk Sessions | Poster Sessions | All Posters | Search Papers

Poster Session B: Wednesday, August 13, 1:00 – 4:00 pm, de Brug & E‑Hall

Artificial Neural Networks trained on human cognitive data as network models of cognition in health and mental disorders

Oliver Frank¹; ¹Zentralinstitut für Seelische Gesundheit

Presenter: Oliver Frank

Cognitive functions are mental processes essential for goal-directed behavior. Impairments in these functions are common in psychiatric disorders and significantly impact quality of life. Artificial Neural Networks (ANNs), trained on cognitive test data from human individuals, offer a new model-based approach to study potential causal links between brain network structure, cognitive function and brain architecture. In this study, we collected longitudinal cognitive data from healthy individuals and patients (schizophrenia, depression, autism spectrum disorder) to train individualized ANNs and analyse their emerging network properties. Our results show that ANNs can learn participants’ behavior and, when initialized with suitable architectures, exhibit a balance of integration and segregation in their hidden layers, mirroring the brain’s topological organization. Network topologies remain mostly robust across randomized training iterations, and topological marker distributions differ significantly (5 out of 6 comparisons (t-test), p < .05). Our findings suggest that ANNs trained on cognitive-behavioral data may serve as tools to understand (brain) network properties underlying human cognitive function in health and mental disorder.

Topic Area: Brain Networks & Neural Dynamics

Extended Abstract: Full Text PDF