Jie Biao, Liu Mingxia, Liu Jun, Zhang Daoqiang, Shen Dinggang
IEEE Trans Biomed Eng. 2017 Jan;64(1):238-249. doi: 10.1109/TBME.2016.2553663. Epub 2016 Apr 13.
Sparse learning has been widely investigated for analysis of brain images to assist the diagnosis of Alzheimer's disease and its prodromal stage, i.e., mild cognitive impairment. However, most existing sparse learning-based studies only adopt cross-sectional analysis methods, where the sparse model is learned using data from a single time-point. Actually, multiple time-points of data are often available in brain imaging applications, which can be used in some longitudinal analysis methods to better uncover the disease progression patterns. Accordingly, in this paper, we propose a novel temporallyconstrained group sparse learning method aiming for longitudinal analysis with multiple time-points of data. Specifically, we learn a sparse linear regression model by using the imaging data from multiple time-points, where a group regularization term is first employed to group the weights for the same brain region across different time-points together. Furthermore, to reflect the smooth changes between data derived from adjacent time-points, we incorporate two smoothness regularization terms into the objective function, i.e., one fused smoothness term thatrequires that the differences between two successive weight vectors from adjacent time-points should be small, and another output smoothness term thatrequires the differences between outputs of two successive models from adjacent time-points should also be small. We develop an efficient optimization algorithm to solve the proposed objective function. Experimental results on ADNI database demonstrate that, compared with conventional sparse learning-based methods, our proposed method can achieve improved regression performance and also help in discovering disease-related biomarkers.
稀疏学习已被广泛用于脑图像分析，以辅助阿尔茨海默病及其前驱阶段（即轻度认知障碍）的诊断。然而，大多数现有的基于稀疏学习的研究仅采用横断面分析方法，即使用来自单个时间点的数据学习稀疏模型。实际上，在脑成像应用中通常可以获得多个时间点的数据，这些数据可用于一些纵向分析方法，以更好地揭示疾病进展模式。因此，在本文中，我们提出了一种新颖的时间约束组稀疏学习方法，旨在对多个时间点的数据进行纵向分析。具体来说，我们通过使用来自多个时间点的成像数据来学习一个稀疏线性回归模型，其中首先采用组正则化项将不同时间点上同一脑区的权重分组在一起。此外，为了反映相邻时间点数据之间的平滑变化，我们在目标函数中纳入了两个平滑正则化项，即一个融合平滑项，要求相邻时间点的两个连续权重向量之间的差异应较小，另一个输出平滑项，要求相邻时间点的两个连续模型的输出之间的差异也应较小。我们开发了一种高效的优化算法来求解所提出的目标函数。在ADNI数据库上的实验结果表明，与传统的基于稀疏学习的方法相比，我们提出的方法可以实现更好的回归性能，并且有助于发现与疾病相关的生物标志物。
