期刊
COMPUTERS IN BIOLOGY AND MEDICINE
卷 147, 期 -, 页码 -出版社
PERGAMON-ELSEVIER SCIENCE LTD
DOI: 10.1016/j.compbiomed.2022.105705
关键词
Alzheimer?s disease; Disease progression; Feature selection; Incomplete labeled data
类别
资金
- National Key R&D Program of China [2019YFC1710300]
- Sichuan Science and Technology Program [2019YFS0019, 2020YFS0283, 2021YFS0152, 2021YJ0184]
- Alzheimer's Disease Neuroimaging Initiative (ADNI)
- USA (National Institutes of Health) [U01 AG024904]
- DOD ADNI, USA (Department of Defense) [W81XWH-12-2-0012]
- National Institute on Aging
- National Institute of Biomedical Imaging and Bioengineering
- AbbVie
- Alzheimer's Association
- Alzheimer's Drug Discovery Foundation
- Araclon Biotech
- BioClinica, Inc.
- Biogen
- BristolMyers Squibb Company
- CereSpir, Inc.
- Cogstate
- Eisai Inc.
- Elan Pharmaceuticals, Inc.
- Eli Lilly and Company
- EuroImmun
- F. Hoffmann-La Roche Ltd
- company Genentech, Inc.
- Fujirebio
- GE Healthcare
- IXICO Ltd.
- Janssen Alzheimer Immunotherapy Research & Development, LLC.
- Johnson & Johnson Pharmaceutical Research & Development LLC.
- Lumosity
- Lundbeck
- Merck Co., Inc.
- Meso Scale Diagnostics, LLC.
- NeuroRx Research
- Neurotrack Technologies
- Novartis Pharmaceuticals Corporation
- Pfizer Inc.
- Piramal Imaging
- Servier
- Takeda Pharmaceutical Company
- Transition Therapeutics
- Canadian Institutes of Health Research
The study introduces a new model LSFSIL for predicting cognitive performance and identifying neuroimaging markers with incomplete labeled data using MRI. Experimental results demonstrate that LSFSIL outperforms existing methods and selects consistent neuroimaging markers with previous studies.
Background: How to predict the cognitive performance of Alzheimer's disease (AD) and identify the informative neuroimaging markers is essential for timely treatment and possible delay of the disease. However, incomplete labeled samples and noises in neuroimaging data pose challenges to building reliable and robust prediction models. In this paper, we present a model named Low-rank Sparse Feature Selection with Incomplete Labels (LSFSIL) for predicting cognitive performance and identifying informative neuroimaging markers with MRI data and incomplete cognitive scores. Method: We propose a sparse matrix decomposition method to decompose the incomplete cognitive score matrix into two parts for recovering missing scores and utilizing incomplete labeled data. The former is the recovered cognitive score matrix without missing values. To make the recovered scores close to the real ones, a manifold regularizer is devised to fit the label distribution for capturing the label correlations locally. The latter is a l(1)-norm regularized matrix which represents the associated errors. Next, a low-rank regression model that regards the recovered matrix as the target is developed to increase the robustness to noises and outliers. Besides, l(2,1)-norm is introduced into the objective function as a sparse regularization to identify the important features. Results: Experimental results demonstrate that LSFSIL achieves higher performance and outperforms several state-of-the-art feature selection approaches. Moreover, the neuroimaging markers selected by LSFSIL are consistent with the previous AD studies. Conclusions: LSFSIL is effective in informative neuroimaging marker identification for cognitive performance prediction with incomplete labeled data.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据