Research has shown that the left and right hemispheres display different neural atrophy patterns in the progression of Alzheimer’s Disease (AD). Yet, in current research practices, asymmetrical...Show moreResearch has shown that the left and right hemispheres display different neural atrophy patterns in the progression of Alzheimer’s Disease (AD). Yet, in current research practices, asymmetrical patterns are not usually considered during classification analyses. Therefore, in this study we analysed the difference in classification accuracy between the left and right hemisphere for different stages of the disease. To further analyse the impact of asymmetric neural atrophy, we calculated an asymmetry score to include in the classification analyses. The sample consisted out of 173 healthy controls and 75 AD patients divided in two groups: 38 mild AD patients and 37 with severe AD (based on MMSE-scores). The classification analyses were performed using a logistic regression analysis with a LASSO penalty including a nested cross-validation procedure to improve the reliability of our analyses. We found that in the case of both mild and severe AD, the combined model (left + right hemisphere) showed that the left hemisphere was better in classifying AD. However, in severe AD the model included more predictors from the right hemisphere, indicating that the right hemisphere offers valuable information in classifying severe AD. We found that asymmetry scores did improve classification accuracy for mild AD. In contrast, this was not the case for severe AD, since asymmetric atrophy diminishes as AD progresses. Furthermore, asymmetry on its own was consistently shown to be a poor predictor of AD. This could possibly be attributed to the calculation method of asymmetry in our study. Overall, our findings contribute to the idea of asymmetric neural atrophy in AD and the emphasize the need for further investigation into the calculation of asymmetry scores.Show less
Research master thesis | Developmental Psychopathology in Education and Child Studies (research) (MSc)
closed access
Background: Empathy deficits are evident in Autism Spectrum Disorder (ASD) and Conduct Disorder (CD), and are linked to distinct brain structural abnormalities. Criticisms over the years highlight...Show moreBackground: Empathy deficits are evident in Autism Spectrum Disorder (ASD) and Conduct Disorder (CD), and are linked to distinct brain structural abnormalities. Criticisms over the years highlight that the DSM classifications of psychiatric disorders are primarily based on observable signs and symptoms, insufficiently based on causes and (neuro)biology, and rather ignorant of heterogeneity and overlap in symptoms. The current study therefore aimed for classification of ASD and CD in adolescents, based on brain morphology (BM) and social-emotional functioning (SEF). Methods: The sample included boys with ASD (n = 23) or CD (n = 51), and typically developing boys (TD; n = 36), aged 15-19 years. Participants’ empathy, aggression, psychopathy, problem behaviours, social function/cognition, and brain morphology (using an MRI scanning) were assessed. For the prediction of ASD and CD, a predictive regression with cross-validation comparing three models was performed, followed by several LASSO regressions. For clustering participants, K-means clustering was used with three clusters and K determined by the CH-index and ARI, followed by ANOVAs, T-tests and checking nestedness with crosstabs. Results: ASD and CD DSM-5 diagnoses can most accurately be predicted with a model based on SEF data, and least accurately with a model based on BM data. When using only SEF data, similar to the DSM, classification is erroneous in approximately one fifth of the participants. For an optimal prediction of ASD and CD, a combination of background, SEF, and BM variables is necessary. None of the cluster-solutions, theory-driven (three clusters) or data-driven (SEF data: four clusters; BM data: five clusters), were congruent with the original DSM clustering (ASD, CD, and TD). Conclusion: The current study shows that data-driven classification, based on BM and/or SEF, is not sufficiently accurate or congruent with the DSM classifications. This suggests that the DSM classifications, which are mainly based on SEF data, do not capture ASD and CD well enough. In order to better capture these disorders, combining factors across multiple domains (including background and BM data), is necessary. On top of that, this study shows less explored ways of analysing data in the field of social sciences. As such, this study may represent a stepping stone for the development of more accurate classifications with less negative implications.Show less
Introduction: Structural MRI has been widely used for Alzheimer’s disease (AD) diagnosis, but resting state functional MRI (RS-fMRI) and diffuse tensor imaging (DTI) can also be used for AD...Show moreIntroduction: Structural MRI has been widely used for Alzheimer’s disease (AD) diagnosis, but resting state functional MRI (RS-fMRI) and diffuse tensor imaging (DTI) can also be used for AD classification. Various studies compared several combinations of MRI types to assess whether combinations would increase classification accuracy, yet no study that researched these 3 MRI types compared all possible combinations for AD classification within one dataset. Methods: This thesis compares 7 models of MRI combinations from 1 dataset to classify AD patients (N = 76) and controls (N = 173). From the sMRI scans, 14 subcortical volumes, 68 regional cortical thickness values, and 48 regional grey matter density values were calculated, the RS-fMRI scans were used to calculate pairwise functional connectivity (FC) between 70 brain regions, and from the DTI scans fractional anisotropy (FA) values were calculated for 20 white matter regions. The Least Absolute Shrinkage and Selection Operator (LASSO) was used to set a penalty, with the goal to reduce overfitting by reducing the number of variables. With cross-validation, the models were trained and tested to estimate the classification accuracy of AD. Results: sMRI had an AUC of .931, sMRI + DTI resulted in an AUC of .929. The sMRI + RS-fMRI + DTI model had an AUC of .902. Conclusion: sMRI is the best single MRI type for classifying AD, and is better than any combination of MRI types. RS-fMRI and DTI were not helpful for increasing classification accuracy.Show less
The ability to correctly classify an individual’s clinical status could help pave the way for early treatment programs for disorders and illnesses of mental- and physical nature alike. Neuroimaging...Show moreThe ability to correctly classify an individual’s clinical status could help pave the way for early treatment programs for disorders and illnesses of mental- and physical nature alike. Neuroimaging data could serve as a basis in reaching this goal of accurate classification. The use of said data, however, does come with challenges. A prominent one of which is the fact that neuroimaging data is highly dimensional, meaning that the amount of features largely exceeds the number of subjects within the data set. Furthermore, research has indicated that the use of heterogeneous, but complimentary, data derived from multiple modalities can be an asset to a model used in a classification setting (Zhang et al., 2010; Schouten et al., 2016). The challenge arises in how to combine the different modalities within a single model. A solution could be the use of machine learning algorithms which search for patterns in the data to draw conclusions. Current literature is, however, lacking meaningful comparisons between different machine learning techniques. Within this project, three different algorithms (support vector machines, Gaussian process classification and multiple kernel learning) have been selected to get insight into (1) whether or not machine learning is able to cope with challenges in the use of neuroimaging data, (2) the difference in performance between these methods and (3) whether or not the use of multiple modalities leads to better results in classification. To this end, 16 alcohol-dependent respondents have been selected along with 32 age-matched healthy controls and have been subjected to both MRI and PET. Models have been trained on data from both separate modalities and on data combining the two modalities. The performances of the models have been assessed by leave-one-subject-out cross-validation and expressed in balanced accuracy and area under the curve. Results indicate that the chosen methods are effective in overcoming challenges arising in the use of neuroimaging data as a means of classification. High balanced accuracies have been found ranging from 76.56% (GPC using PET data) to 100% (GPC using MRI data). Different situations are cause for different solutions and the right choice of algorithm seems to be dependent on, for instance, the fact if either unimodal- or multimodal data is used. Also, settings/optimization of parameters within the model can make a large impact on accuracy. It is therefore advised that researchers try different algorithms and settings before selecting a technique. Different options need to be weighed in order to receive the best possible outcome.Show less
Research master thesis | Classics and Ancient Civilizations (research) (MA)
open access
By means of the sign known as Gardiner F27, theory on classification in ancient Egyptian written language is juxtaposed with orthography in both hieroglyphic and hieratic writing. Existing theory,...Show moreBy means of the sign known as Gardiner F27, theory on classification in ancient Egyptian written language is juxtaposed with orthography in both hieroglyphic and hieratic writing. Existing theory, currently a hotly debated topic in Egyptological scholarship, is tested and modified by means of source data primarily from the Ramesside period in Egypt. In this way the thesis opens up promising new avenues for future research.Show less
