Stable Feature Selection from Brain sMRI

TL;DR

This paper introduces a nonnegative generalized fused lasso model that enhances stable feature selection in brain sMRI data for Alzheimer's diagnosis, emphasizing spatial cohesion and positive correlation, with an efficient optimization algorithm.

Contribution

It proposes a novel nonnegative fused lasso model incorporating pathological priors and a new optimization algorithm based on conic duality, improving feature stability.

Findings

The model outperforms existing methods in selecting stable features.

It effectively captures spatial and disease-related priors.

Experiments demonstrate improved reproducibility and accuracy.

Abstract

Neuroimage analysis usually involves learning thousands or even millions of variables using only a limited number of samples. In this regard, sparse models, e.g. the lasso, are applied to select the optimal features and achieve high diagnosis accuracy. The lasso, however, usually results in independent unstable features. Stability, a manifest of reproducibility of statistical results subject to reasonable perturbations to data and the model, is an important focus in statistics, especially in the analysis of high dimensional data. In this paper, we explore a nonnegative generalized fused lasso model for stable feature selection in the diagnosis of Alzheimer's disease. In addition to sparsity, our model incorporates two important pathological priors: the spatial cohesion of lesion voxels and the positive correlation between the features and the disease labels. To optimize the model, we propose an efficient algorithm by proving a novel link between total variation and fast network flow algorithms via conic duality. Experiments show that the proposed nonnegative model performs much better in exploring the intrinsic structure of data via selecting stable features compared with other state-of-the-arts.