Each and every one of us are inevitably and constantly aging. Although the individual differences are substantial, with increasing age, most cognitive processes, such as processing speed, attention and working memory, decline. In addition to normal aging, age-related brain diseases can accelerate or lead to sudden cognitive decline. Stroke constitutes a good example of sudden loss of functions following a vascular cerebral insult for which the recovery of cognitive functions can have a major impact on the individual’s quality of life and independence. There is a need for better treatment programs targeting cognitive function following disease, as well as programs preventing or slowing down the “normal” age-related decline in order to increase the health span of the population together with the increasing life span. In order to develop and evaluate better treatment or preventive programs, we need reliable and sensitive measures of brain health. In this thesis, we assessed two candidate markers of brain and cognitive health, namely the concept of brain age prediction based on neuroimaging data (Franke, Ziegler, Kloppel, Gaser, & Alzheimer's Disease Neuroimaging, 2010), and a computational model of visual attention based on the theory of visual attention (TVA; Bundesen, 1990).
In order to evaluate the sensitivity and reliability of brain age prediction and computational parameters obtained from TVA as measures of brain health and cognitive functions in response to cognitive training, we used a combination of cross-sectional and longitudinal data from a group of healthy individuals and a group of chronic stroke patients who underwent a computerized cognitive training (CCT) program in combination with either sham or active transcranial current stimulation (tDCS), respectively.
More specifically, in Paper I, using a cross-sectional design, we investigated the association between 1) various brain age prediction models based on sub-sets of diffusion tensor imaging (DTI) measures of white matter properties and Freesurfer-based T1-weighted MRI measures, and 2) a range of cognitive measures in 265 healthy adults aged 20 to 88 years. Briefly, our results demonstrated good age prediction across our 11 trained models, and differential sensitivity to cognitive measures. These results suggest that modality-specific brain age estimation might better capture the heterogeneity of the aging brain by providing partly independent information.
Building on the results from Paper I, in Paper II, using a longitudinal design, we explored the clinical utility and feasibility of automated brain age estimation in stroke patients. More specifically, we investigated the reliability of global and regional brain age prediction in a group of 54 chronic stroke patients who underwent three weeks of CCT together with either sham or active tDCS. We assessed the sensitivity of brain age prediction to cognitive performance at baseline, as well as response to the CCT. Briefly, our results confirmed the high reliability and feasibility of brain age estimation in clinical group; however, we did not find significant association between brain age estimation and cognitive test performance or response to CCT. The implications and limitations of our study are discussed, including our sample bias towards the highly functioning end of the stroke population.
Together, Paper I and II supports the use of brain age models based on different subset of features as partly complementary measures of brain health and integrity; however, replications and further refinement of the methods are needed to increase sensitivity and specificity. Importantly, our results also demonstrate the clinical feasibility of automated brain age estimation in patients with brain lesions, highlighting the robustness of the technique. Yet, the clinical utility toward individualized care requires further investigations.
In Paper III, based on the critical role of attentional abilities in complex everyday functions, and in functional recovery following stroke, we assessed the sensitivity to detect attentional impairment in chronic stroke patients and the reliability and predictive value of TVA parameters reflecting short-term memory capacity (K), processing speed (C) and perceptual threshold (t0) in response to cognitive training. To do so, we used cross-sectional case-control data, which included data from 70 stroke patients and 140 controls matched by age and sex data. In addition, we used longitudinal data collected from 54 stroke patients who underwent six repeated assessments over the course of a three-weeks intensive CCT intervention. Briefly, our results revealed poorer storage capacity, lower processing speed and higher visual threshold in chronic stroke patients compared to age-matched healthy controls. In addition, we demonstrated high reliability of the TVA parameters in stroke patients across six test sessions, and showed that higher processing speed at baseline was associated with larger improvement during the course of CCT.
Overall, our findings do not support the concept of a single marker of cognitive health. Rather, models of cognitive health ought to incorporate and use a range of complementary
markers in order to achieve predictability at an individual level. Exploring a variety of markers may eventually enable personalized preventive or treatments programs that address the increasing societal challenges caused by cognitive decline associated with normal aging and aging-related brain disorders such as stroke.