Journal of Alzheimer's Disease - Volume 26, issue s3

Authors: Haller, Sven | Lovblad, Karl O. | Giannakopoulos, Panteleimon

Article Type: Research Article

Abstract: The majority of advanced neuroimaging studies implement group level analyses contrasting a group of patients versus a group of controls, or two groups of patients. Such analyses may identify for example changes in grey matter in specific regions associated with a given disease. Although such group investigations provided key contributions to the understanding of the pathological process surrounding a wide range of diseases, they are of limited utility at an individual level. Recently, there is a trend towards individual classification analyses, representing a fundamental shift of the research paradigm. In contrast to group comparisons, these latter studies do not provide …insights on vulnerable brain areas but may allow for an early (and ideally preclinical) identification of at risk individuals in routine clinical setting. One currently very popular method in this domain are support vector machines (SVM), yet this method is only one of many available methods in the field of individual classification analyses. The current manuscript reviews the fundamental properties and features of such individual level classification analyses in neurodegenerative diseases. Show more

Keywords: SVM (support vector machine), MVPA (multi voxel pattern analysis), artificial intelligence, machine learning, individual classification

DOI: 10.3233/JAD-2011-0014

Citation: Journal of Alzheimer's Disease, vol. 26, no. s3, pp. 389-394, 2011

Authors: Furney, Simon J. | Kronenberg, Deborah | Simmons, Andrew | Güntert, Andreas | Dobson, Richard J. | Proitsi, Petroula | Wahlund, Lars Olof | Kloszewska, Iwona | Mecocci, Patrizia | Soininen, Hilkka | Tsolaki, Magda | Vellas, Bruno | Spenger, Christian | Lovestone, Simon

Article Type: Research Article

Abstract: Progression of people presenting with Mild Cognitive Impairment (MCI) to dementia is not certain and it is not possible for clinicians to predict which people are most likely to convert. The inability of clinicians to predict progression limits the use of MCI as a syndrome for treatment in prevention trials and, as more people present with this syndrome in memory clinics, and as earlier diagnosis is a major goal of health services, this presents an important clinical problem. Some data suggest that CSF biomarkers and functional imaging using PET might act as markers to facilitate prediction of conversion. However, both …techniques are costly and not universally available. The objective of our study was to investigate the potential added benefit of combining biomarkers that are more easily obtained in routine clinical practice to predict conversion from MCI to Alzheimer's disease. To explore this we combined automated regional analysis of structural MRI with analysis of plasma cytokines and chemokines and compared these to measures of APOE genotype and clinical assessment to assess which best predict progression. In a total of 205 people with MCI, 77 of whom subsequently converted to Alzheimer's disease, we find biochemical markers of inflammation to be better predictors of conversion than APOE genotype or clinical measures (Area under the curve (AUC) 0.65, 0.62, 0.59 respectively). In a subset of subjects who also had MRI scans the combination of serum markers of inflammation and MRI automated imaging analysis provided the best predictor of conversion (AUC 0.78). These results show that the combination of imaging and cytokine biomarkers provides an improvement in prediction of MCI to AD conversion compared to either datatype alone, APOE genotype or clinical data and an accuracy of prediction that would have clinical utility. Show more

Keywords: Proteomics, MRI, mild cognitive impairment, Alzheimer's disease, biomarkers

DOI: 10.3233/JAD-2011-0044

Citation: Journal of Alzheimer's Disease, vol. 26, no. s3, pp. 395-405, 2011