Machine Learning classification of MRI features of Alzheimer's disease and mild cognitive impairment subjects to reduce the sample size in clinical trials.
ABSTRACT There is a need for objective tools to help clinicians to diagnose Alzheimer's Disease (AD) early and accurately and to conduct Clinical Trials (CTs) with fewer patients. Magnetic Resonance Imaging (MRI) is a promising AD biomarker but no single MRI feature is optimal for all disease stages. Machine Learning classification can address these challenges. In this study, we have investigated the classification of MRI features from AD, Mild Cognitive Impairment (MCI), and control subjects from ADNI with four techniques. The highest accuracy rates for the classification of controls against ADs and MCIs were 89.2% and 72.7%, respectively. Moreover, we used the classifiers to select AD and MCI subjects who are most likely to decline for inclusion in hypothetical CTs. Using the hippocampal volume as an outcome measure, we found that the required group sizes for the CTs were reduced from 197 to 117 AD patients and from 366 to 215 MCI subjects.
- SourceAvailable from: Javier EscuderoTrials 12/2011; 12 Suppl 1(Suppl 1):A18. DOI:10.1186/1745-6215-12-S1-A18
- [Show abstract] [Hide abstract]
ABSTRACT: OBJECTIVE: We describe the operationalization of the National Institute on Aging-Alzheimer's Association (NIA-AA) workgroup diagnostic guidelines pertaining to Alzheimer disease (AD) dementia in a large multicenter group of subjects with AD dementia. METHODS: Subjects with AD dementia from the Alzheimer's Disease Neuroimaging Initiative (ADNI) with at least 1 amyloid biomarker (n = 211) were included in this report. Biomarker data from CSF Aβ42, amyloid PET, fluorodeoxyglucose-PET, and MRI were examined. The biomarker results were assessed on a per-patient basis and the subject categorization as defined in the NIA-AA workgroup guidelines was determined. RESULTS: When using a requirement that subjects have a positive amyloid biomarker and single neuronal injury marker having an AD pattern, 87% (48% for both neuronal injury biomarkers) of the subjects could be categorized as "high probability" for AD. Amyloid status of the combined Pittsburgh compound B-PET and CSF results showed an amyloid-negative rate of 10% in the AD group. In the ADNI AD group, 5 of 92 subjects fit the category "dementia unlikely due to AD" when at least one neuronal injury marker was negative. CONCLUSIONS: A large proportion of subjects with AD dementia in ADNI may be categorized more definitively as high-probability AD using the proposed biomarker scheme in the NIA-AA criteria. A minority of subjects may be excluded from the diagnosis of AD by using biomarkers in clinically categorized AD subjects. In a well-defined AD dementia population, significant biomarker inconsistency can be seen on a per-patient basis.Neurology 05/2013; 80(23). DOI:10.1212/WNL.0b013e318295d6cf
- [Show abstract] [Hide abstract]
ABSTRACT: In recent years, numerous laboratories and consortia have used neuroimaging to evaluate the risk for and progression of Alzheimer's disease (AD). The Alzheimer's Disease Neuroimaging Initiative is a longitudinal, multicenter study that is evaluating a range of biomarkers for use in diagnosis of AD, prediction of patient outcomes, and clinical trials. These biomarkers include brain metrics derived from magnetic resonance imaging (MRI) and positron emission tomography scans as well as metrics derived from blood and cerebrospinal fluid. We focus on Alzheimer's Disease Neuroimaging Initiative studies published between 2011 and March 2013 for which structural MRI was a major outcome measure. Our main goal was to review key articles offering insights into progression of AD and the relationships of structural MRI measures to cognition and to other biomarkers in AD. In Supplement 1, we also discuss genetic and environmental risk factors for AD and exciting new analysis tools for the efficient evaluation of large-scale structural MRI data sets such as the Alzheimer's Disease Neuroimaging Initiative data.Biological psychiatry 11/2013; DOI:10.1016/j.biopsych.2013.11.020