Diseased Region Detection of Longitudinal Knee Magnetic Resonance Imaging Data

Abstract

Magnetic resonance imaging (MRI) has become an important imaging technique for quantifying the spatial location and magnitude/direction of longitudinal cartilage morphology changes in patients with osteoarthritis (OA). Although several analytical methods, such as subregion-based analysis, have been developed to refine and improve quantitative cartilage analyses, they can be suboptimal due to two major issues: the lack of spatial correspondence across subjects and time and the spatial heterogeneity of cartilage progression across subjects. The aim of this paper is to present a statistical method for longitudinal cartilage quantification in OA patients, while addressing these two issues. The 3D knee image data is preprocessed to establish spatial correspondence across subjects and/or time. Then, a Gaussian hidden Markov model (GHMM) is proposed to deal with the spatial heterogeneity of cartilage progression across both time and OA subjects. To estimate unknown parameters in GHMM, we employ a pseudo-likelihood function and optimize it by using an expectation-maximization (EM) algorithm. The proposed model can effectively detect diseased regions in each OA subject and present a localized analysis of longitudinal cartilage thickness within each latent subpopulation. Our GHMM integrates the strengths of two standard statistical methods including the local subregion-based analysis and the ordered value approach. We use simulation studies and the Pfizer longitudinal knee MRI dataset to evaluate the finite sample performance of GHMM in the quantification of longitudinal cartilage morphology changes. Our results indicate that GHMM significantly outperforms several standard analytical methods.

Statistical disease mapping for heterogeneous neuroimaging studies

“…. These assumptions are implicitly used in existing methods (Huang et al, 2015). The first assumption makes the normal controls a reference for detecting individual diseased regions, the second assumption plays a critical role in establishing spatial correspondence of diseased regions at different locations across patients, and the third assumption is an identifiability constraint that can deal with the label-switching issue in diseased region detection.…”

“…which approximately follows a 2 distribution with J degrees of freedom under the null hypothesis when the sample size is large enough. Here, the variance term Var[vec(B)] in TB can be obtained via the wild bootstrap method as described in Huang et al (2015), which proceeds as follows:…”

“…• This article proposes a special image-on-scalar regression model that integrates HMRFM with MVCM and preserves the key features from functional data analysis tools and Markov random field models. Specifically, our SDM can not only build up the relationship between functional phenotypes and a set of covariates of interest (Zhu et al, 2011;Zhu, Li & Kong, 2012;Huang et al, 2017) but also detect individual diseased regions (Geman & Geman, 1984;Huang et al, 2015). In addition, compared to voxel-wise analysis, our SDM can effectively capture the spatial smoothness and correlation within the imaging signals and model the heterogeneity among multiple imaging features.…”

“…Compared to the existing statistical methods for abnormal pattern detection in the literature (Van Leemput et al, 2001; Matteoli, Diani & Corsini, 2010; Penny et al, 2011; Ramteke & Monali, 2012; Goldstein & Uchida, 2016; Shaker et al, 2017), four major distinctive contributions of this article in terms of both methodology and application are as follows: This article proposes a special image‐on‐scalar regression model that integrates HMRFM with MVCM and preserves the key features from functional data analysis tools and Markov random field models. Specifically, our SDM can not only build up the relationship between functional phenotypes and a set of covariates of interest (Zhu et al, 2011; Zhu, Li & Kong, 2012; Huang et al, 2017) but also detect individual diseased regions (Geman & Geman, 1984; Huang et al, 2015). In addition, compared to voxel‐wise analysis, our SDM can effectively capture the spatial smoothness and correlation within the imaging signals and model the heterogeneity among multiple imaging features. Besides the subject‐specific diseased region detection, our SDM can also build the disease map at the group level.…”

Osteoarthritis year in review 2016: imaging

“…The method could be suitable especially in very large patient cohorts 43 . In the same area, another statistical method was tested to refine and improve computer aided quantitative cartilage analyses, that seemed to perform better than conventional voxelbased methods for detecting longitudinal cartilage changes in the Pfizer KOA cohort 44 . The applicability of automatic and semiautomatic cartilage segmentation methods must be further investigated with manual segmentation as the reference to ensure sufficient quality of measurement outcomes.…”