Session

Saturday, June 24, 2017 - 09:00 to 15:45
Predicting Cognitive Recovery of Stroke Patients from the Structural MRI Connectome using a Naïve Bayesian Tree Classifier
Abstract: 
Successful post-stroke prognosis and recovery strategies are heavily dependent on our understanding about how the damage to one specific region may impact to other remote regions, as well as the various functional networks involved in efficient cognitive function. In this study, 27 consecutive ischemic stroke patients were recruited. Stroke patients underwent two complete neuropsychological assessments between the first 72 hours after stroke arrival and three months later. They were further evaluated with a MRI protocol at 3 months. Patients were splitted into two groups according to their level of cognitive recovery. A data mining technique was then applied to the probabilistic tractography data in order to determine whether the structural connectivity features can efficiently classify good from poor recovery. We found that the connectivity probability between the left Superior Parietal Gyrus and the left Angular Gyrus can describe the cognitive classification (good versus poor recovery) after stroke. Both regions are involved in higher cognitive functioning and their dysfunction has been related to mild cognitive impairment and dementia. Our findings suggest that cognitive prognosis, in stroke patients, mainly depends on the connection of these two regions. An accurate model for the early prediction of stroke recovery as the one presented herein is fundamental to develop early personalized rehabilitation strategies.
Rosalia Dacosta-Aguayo's picture
Rosalia Dacosta-Aguayo
Christian Stephan-Otto's picture
Christian Stephan-Otto
Tibor Auer's picture
Tibor Auer
Ic Clemente's picture
Ic Clemente
Antoni Davalos's picture
Antoni Davalos
Nuria Bargallo's picture
Nuria Bargallo
Maria Mataro's picture
Maria Mataro
Manousos A. Klados's picture
Manousos A. Klados
Aston University (UK)