There are many active research projects accessing and applying shared ADNI data. Use the search above to find specific research focuses on the active ADNI investigations. This information is requested annually as a requirement for data access.
| Principal Investigator | |
| Principal Investigator's Name: | Yang Feng |
| Institution: | New York University |
| Department: | Biostatistics |
| Country: | |
| Proposed Analysis: | Alzheimer’s disease (AD) is a common chronic neurodegenerative disease with an estimated 5.8 million Americans (one in 10 people aged 65 and older) having AD in 2019. Early and accurate diagnosis of AD is important because many conditions (e.g., depression, adverse drug reactions, metabolic changes) can cause dementia- like symptoms, some of which are treatable or reversible. There has thus been extensive research on developing accurate classifiers (e.g., linear discriminant analysis, support vector machines, multi-class logistic regression, random forests, boosting, convolutional neural network) to automatically assign people into different classes, including mild cognitive impairment (MCI), AD, and normal control (NC). In most existing work, the classifiers are designed to reach high overall accuracy. However, different types of classification error may have various degrees of consequences. For example, misclassifying AD as NC may lead to the missing of timely treatment while mis- classifying NC as AD may lead to underpowered clinical trials. As a result, it is desirable in such applications to create multiclass classifiers with prioritized error control. For instance, a practitioner may want to maximize the probability of correctly classifying AD as AD while controlling the probability of misclassifying AD as NC under 1% and the probability of misclassifying MCI as NC under 5%. The main goal of this project is introduce a general framework for multiclass classification that can fit the practitioners’ specific needs for controlling selected types of misclassification errors, as well as imposing various (relative) costs for another set of misclassification error types. Within this unified framework, we will develop an efficient umbrella algorithm that converts established classification methods to a new classifier that takes into account both the prioritized error controls and various costs involved. We will apply the proposed algorithm to classify AD patients using the Alzheimer’s Disease Neuroimaging Initiative (ADNI) data. It can be used to optimize treatment for AD patients and expedite drug discovery for AD. |
| Additional Investigators |
