Dynamical network stability analysis of multiple biological ages provides a framework for understanding the aging process.

Widespread interest in non-destructive biomarkers of aging has led to a multitude of biological ages that each proffers a 'true' health-adjusted individual age. While each measure provides salient information on the aging process, they are each univariate, in contrast to the "hallmark" and "pillar" theories of aging which are explicitly multidimensional, multicausal and multiscale. Fortunately, multiple biological ages can be systematically combined into a multidimensional network representation. The interaction network between these biological ages permits analysis of the multidimensional effects of aging, as well as quantification of causal influences during both natural aging and, potentially, after anti-aging intervention. The behaviour of the system as a whole can then be explored using dynamical network stability analysis which identifies new, efficient biomarkers that quantify long term resilience scores on the timescale between measurements (years). We demonstrate this approach using a set of 8 biological ages from the longitudinal Swedish Adoption/Twin Study of Aging (SATSA). After extracting an interaction network between these biological ages, we observed that physiological age, a proxy for cardiometabolic health, serves as a central node in the network, implicating it as a key vulnerability for slow, age-related decline. We furthermore show that while the system as a whole is stable, there is a weakly stable direction along which recovery is slow -- on the timescale of a human lifespan. This slow direction provides an aging biomarker which correlates strongly with chronological age and predicts longitudinal decline in health - suggesting that it estimates an important driver of age-related changes.

Network dynamical stability analysis reveals key "mallostatic" natural variables that erode homeostasis and drive age-related decline of health.

Using longitudinal study data, we dynamically model how aging affects homeostasis in both mice and humans. We operationalize homeostasis as a multivariate mean-reverting stochastic process. We hypothesize that biomarkers have stable equilibrium values, but that deviations from equilibrium of each biomarker affects other biomarkers through an interaction network-this precludes univariate analysis. We therefore looked for age-related changes to homeostasis using dynamic network stability analysis, which transforms observed biomarker data into independent "natural" variables and determines their associated recovery rates. Most natural variables remained near equilibrium and were essentially constant in time. A small number of natural variables were unable to equilibrate due to a gradual drift with age in their homeostatic equilibrium, i.e. allostasis. This drift caused them to accumulate over the lifespan course and makes them natural aging variables. Their rate of accumulation was correlated with risk of adverse outcomes: death or dementia onset. We call this tendency for aging organisms to drift towards an equilibrium position of ever-worsening health "mallostasis". We demonstrate that the effects of mallostasis on observed biomarkers are spread out through the interaction network. This could provide a redundancy mechanism to preserve functioning until multi-system dysfunction emerges at advanced ages.

Network dynamical stability analysis reveals key "mallostatic" natural variables that erode homeostasis and drive age-related decline of health.

Using longitudinal study data, we dynamically model how aging affects homeostasis in both mice and humans. We operationalize homeostasis as a multivariate mean-reverting stochastic process. We hypothesize that biomarkers have stable equilibrium values, but that deviations from equilibrium of each biomarker affects other biomarkers through an interaction network - this precludes univariate analysis. We therefore looked for age-related changes to homeostasis using dynamic network stability analysis, which transforms observed biomarker data into independent "natural" variables and determines their associated recovery rates. Most natural variables remained near equilibrium and were essentially constant in time. A small number of natural variables were unable to equilibrate due to a gradual drift with age in their homeostatic equilibrium, i.e. allostasis. This drift caused them to accumulate over the lifespan course and makes them natural aging variables. Their rate of accumulation was correlated with risk of adverse outcomes: death or dementia onset. We call this tendency for aging organisms to drift towards an equilibrium position of ever-worsening health "mallostasis". We demonstrate that the effects of mallostasis on observed biomarkers are spread out through the interaction network. This could provide a redundancy mechanism to preserve functioning until multi-system dysfunction emerges at advanced ages.

Modelling lifespan reduction in an exogenous damage model of generic disease.

We model the effects of disease and other exogenous damage during human aging. Even when the exogenous damage is repaired at the end of acute disease, propagated secondary damage remains. We consider both short-term mortality effects due to (acute) exogenous damage and long-term mortality effects due to propagated damage within the context of a generic network model (GNM) of individual aging that simulates a U.S. population. Across a wide range of disease durations and severities we find that while excess short-term mortality is highest for the oldest individuals, the long-term years of life lost are highest for the youngest individuals. These appear to be universal effects of human disease. We support this conclusion with a phenomenological model coupling damage and mortality. Our results are consistent with previous lifetime mortality studies of atom bomb survivors and post-recovery health studies of COVID-19. We suggest that short-term health impact studies could complement lifetime mortality studies to better characterize the lifetime impacts of disease on both individuals and populations.

Efficient representations of binarized health deficit data: the frailty index and beyond.

We investigated efficient representations of binarized health deficit data using the 2001-2002 National Health and Nutrition Examination Survey (NHANES). We compared the abilities of features to compress health deficit data and to predict adverse outcomes. We used principal component analysis (PCA) and several other dimensionality reduction techniques, together with several varieties of the frailty index (FI). We observed that the FI approximates the first - primary - component obtained by PCA and other compression techniques. Most adverse outcomes were well predicted using only the FI. While the FI is therefore a useful technique for compressing binary deficits into a single variable, additional dimensions were needed for high-fidelity compression of health deficit data. Moreover, some outcomes - including inflammation and metabolic dysfunction - showed high-dimensional behaviour. We generally found that clinical data were easier to compress than lab data. Our results help to explain the success of the FI as a simple dimensionality reduction technique for binary health data. We demonstrate how PCA extends the FI, providing additional health information, and allows us to explore system dimensionality and complexity. PCA is a promising tool for determining and exploring collective health features from collections of binarized biomarkers.

Texture analysis in brain T2 and diffusion MRI differentiates histology-verified grey and white matter pathology types in multiple sclerosis.

BACKGROUND: Multiple sclerosis (MS) is a co mplex disease of the central nervous system involving several types of brain pathology that are difficult to characterize using conventional imaging methods. NEW METHOD: We originated novel texture analysis and machine learning approaches for classifying MS pathology subtypes as compared with 2 common advanced MRI measures: magnetization transfer ratio (MTR) and fractional anisotropy (FA). Texture analysis used an optimized grey level co-occurrence matrix method with histology-informed 7T T2-weighted magnetic resonance imaging (MRI) and diffusion tensor imaging (DTI) from 15 MS and 12 control brain specimens. DTI analysis took an innovative approach that assessed the texture across diffusion directions upsampled from 30 to 90. Tissue types included de- and re-myelinated lesions and normal-appearing areas in both grey and white matter, and diffusely abnormal white matter. Data analyses were stepwise, including: (1) group-wise classification using random forest algorithms based on all or individual imaging parameters; (2) parameter importance ranking; and (3) pairwise analysis using top-ranked features. RESULTS: Texture analysis performed better than MTR and FA, with T2 texture performed the best. T2 texture measures ranked the highest in classifying most grey and white matter tissue types, including de- versus re-myelinated lesions and among grey matter lesion subtypes (accuracy=0.86-0.59; kappa=0.60-0.41). Diffusion texture best differentiated normal appearing and control white matter. COMPARISON WITH EXISTING METHODS: There is no established method in imaging for differentiating MS pathology subtypes. In combined texture analysis and machine learning studies, there is also no direct evidence comparing conventional with advanced MRI measures for assessing MS pathology. Further, this study is unique in conducting innovative texture analysis with DTI following data-augmentation using robust methods. CONCLUSIONS: T2 and diffusion MRI texture analysis integrated with machine learning may be valuable approaches for characterizing MS pathology.

Strategies for handling missing data that improve Frailty Index estimation and predictive power: lessons from the NHANES dataset.

Missing data are ubiquitous in aging studies. Combining the National Health and Nutrition Examination Survey (NHANES) 2003/2004 and 2005/2006 cross-sectional aging studies (N = 9307), we investigated the effects of both real and simulated missing data on the Frailty Index (FI) and survival analysis, along with several mitigation strategies. We observed distinct block patterns of missing variables in the dataset. These blocks showed significant hazard rate (HR) differences when they were missing versus present, indicating that missingness cannot be simply ignored. Simulations of this patterned missingness produced a bias of 0.0112 ± 0.0008 to the mean FI when missing values were ignored, representing a change in hazard of 1.09 ± 0.01. A similar bias of 0.0106 ± 0.0001 was estimated in the real missingness. Imputation was able to correct the bias using the multivariate imputation by chained equations (MICE) method via the classification and regression tree (CART) prediction model together with rule-based imputation. Using auxiliary variables (CART+Aux) improved the performance of CART. Well-performing imputation models, especially CART+Aux, were able to increase the FI predictive power and the reliability of the HR estimates. In contrast, the default MICE models, predictive mean matching/logistic regression (PMM/logreg), caused even stronger biases to the FI. Our results demonstrate that calibration of the FI as a mortality predictor depends on how missing data are handled. Ignoring missing values when calculating the FI may be an acceptable strategy for clinical settings where the FI is used as a rough predictor of adverse outcomes. Where the FI is to be compared across studies or populations, judicious imputation - cognizant of the risks carried by poor imputation - should be used to ensure reliability and precision of statistical estimates and conclusions.

A metric learning method for estimating myelin content based on T2-weighted MRI from a de- and re-myelination model of multiple sclerosis.

Myelin plays a critical role in the pathogenesis of neurological disorders but is difficult to characterize in vivo using standard analysis methods. Our goal was to develop a novel analytical framework for estimating myelin content using T2-weighted magnetic resonance imaging (MRI) based on a de- and re-myelination model of multiple sclerosis. We examined 18 mice with lysolecithin induced demyelination and spontaneous remyelination in the ventral white matter of thoracic spinal cord. Cohorts of 6 mice underwent 9.4T MRI at days 7 (peak demyelination), 14 (ongoing recovery), and 28 (near complete recovery), as well as histological analysis of myelin and the associated cellularity at corresponding timepoints. Our MRI framework took an unsupervised learning approach, including tissue segmentation using a Gaussian Markov random field (GMRF), and myelin and cellularity feature estimation based on the Mahalanobis distance. For comparison, we also investigated 2 regression-based supervised learning approaches, one using our GMRF results, and another using a freely available generalized additive model (GAM). Results showed that GMRF segmentation was 73.2% accurate, and our unsupervised learning method achieved a correlation coefficient of 0.67 (top quartile: 0.78) with histological myelin, similar to 0.70 (top quartile: 0.78) obtained using supervised analyses. Further, the area under the receiver operator characteristic curve of our unsupervised myelin feature (0.883, 95% CI: 0.874-0.891) was significantly better than any of the supervised models in detecting white matter myelin as compared to histology. Collectively, metric learning using standard MRI may prove to be a new alternative method for estimating myelin content, which ultimately can improve our disease monitoring ability in a clinical setting.

Grad-CAM helps interpret the deep learning models trained to classify multiple sclerosis types using clinical brain magnetic resonance imaging.

BACKGROUND: Deep learning using convolutional neural networks (CNNs) has shown great promise in advancing neuroscience research. However, the ability to interpret the CNNs lags far behind, confounding their clinical translation. NEW METHOD: We interrogated 3 heatmap-generating techniques that have increasing generalizability for CNN interpretation: class activation mapping (CAM), gradient (Grad)-CAM, and Grad-CAM++. To investigate the impact of CNNs on heatmap generation, we also examined 6 different models trained to classify brain magnetic resonance imaging into 3 types: relapsing-remitting multiple sclerosis (RRMS), secondary progressive MS (SPMS), and control. Further, we designed novel methods to visualize and quantify the heatmaps to improve interpretability. RESULTS: Grad-CAM showed the best heatmap localizing ability, and CNNs with a global average pooling layer and pretrained weights had the best classification performance. Based on the best-performing CNN model, called VGG19, the 95th percentile values of Grad-CAM in SPMS were significantly higher than RRMS, indicating greater heterogeneity. Further, voxel-wise analysis of the thresholded Grad-CAM confirmed the difference identified visually between RRMS and SPMS in discriminative brain regions: occipital versus frontal and occipital, or temporal/parietal. COMPARISON WITH EXISTING METHODS: No study has examined the CAM methods together using clinical images. There is also lack of study on the impact of CNN architecture on heatmap outcomes, and of technologies to quantify heatmap patterns in clinical settings. CONCLUSIONS: Grad-CAM outperforms CAM and Grad-CAM++. Integrating Grad-CAM, novel heatmap quantification approaches, and robust CNN models may be an effective strategy in identifying the most crucial brain areas underlying disease development in MS.

Evaluation of discrete orthogonal versus polar Stockwell Transform for local multi-resolution texture analysis using brain MRI of multiple sclerosis patients.

The Stockwell Transform has the potential to perform multi-resolution texture analysis in magnetic resonance imaging (MRI). However, it is computationally intensive and memory demanding. The polar Stockwell Transform (PST) is rotation-invariant and relatively memory efficient, but still computationally demanding. The new Discrete Orthogonal Stockwell Transform (DOST) appears to have addressed both the computation and storage challenges; however, its utility in localized texture analysis remains unclear. Our goal was to investigate the theory and texture analysis ability of the DOST versus PST using both synthetic and MR images, and explore the relative importance of the associated texture features using a simple classification example based on clinical brain MRI of six multiple sclerosis patients. MRI texture analysis focused on FLAIR images, and the classification used a machine learning algorithm, random forest, that differentiated regions of interest (ROIs) into 2 classes: white matter lesions, and the contralateral normal-appearing white matter (control). Our results showed that the PST features had a greater ability in detecting subtle changes in image structure than the DOST and polar-index DOST (PDOST). Quantitatively, based on 187 lesion and 187 control ROIs, both the PST and the rotation-invariant radial PST performed better in the classification than the DOST and PDOST, where the latter were no better than guessing (p = 0.65 and 0.98). Further analysis using a hierarchical random forest showed that combining MRI signal intensity with the PST or DOST predictions increased the classification performance, with the accuracy, sensitivity, and specificity all improved to >85% in the tests. Collectively, the DOST is less competitive than the PST in localized image texture analysis. The PST features may help with texture-based lesion classification in MS based on clinical brain MRI scans following further verification.

Acoustic Radiation Force Impulse and Conventional Ultrasound in the Prediction of Cirrhosis Complicating Fatty Liver: Does Body Mass Index Independently Alter the Results?

We investigated whether ultrasound (US) could quantify steatosis and fibrosis in non-alcoholic fatty liver disease (NAFLD). Estimates of fat by gray-scale, hepatorenal index (HRI) and fibrosis by acoustic radiation force impulse (ARFI) were made using the interquartile range (IQR)/median for ARFI quality. Biopsy was the gold standard. US fat assessment correlated with histologic grade and predicted steatosis. HRI predicted steatosis but did not improve accuracy. ARFI of good quality was highly sensitive toward severe fibrosis. The median ARFI value depended linearly on body mass index (BMI). Poor quality ARFI data had higher histologic steatosis, leading to higher mean steatosis grades in rejected data (p = 0.018). The ARFI quality cut with IQR/median >0.15 or >0.3 excluded many more patients with severe steatosis versus normal, influenced by increasing BMI. By combining the baseline US with ARFI, patients can be concurrently diagnosed for steatosis and fibrosis, two of the key pathologies of NAFLD and non-alcoholic steatohepatitis (NASH). However, severe steatosis and high BMI may falsely alter ARFI results.

Characterizing Structural Changes With Devolving Remyelination Following Experimental Demyelination Using High Angular Resolution Diffusion MRI and Texture Analysis.

BACKGROUND: Changes in myelin integrity are associated with the pathophysiology of many neurological diseases, including multiple sclerosis. However, noninvasive measurement of myelin injury and repair remains challenging. Advanced MRI techniques including diffusion tensor imaging (DTI), neurite orientation dispersion and density index (NODDI), and texture analysis have shown promise in quantifying subtle abnormalities in white matter structure. PURPOSE: To determine whether and how these advanced imaging methods help understand remyelination changes after demyelination using a mouse model. STUDY TYPE: Prospective, longitudinal. ANIMAL MODEL: Demyelination was induced in the thoracic spinal cord of 21 mice using the chemical toxin lysolecithin. FIELD STRENGTH/SEQUENCES: 9.4T ASSESSMENT: Imaging was done at day 7 (demyelination) and days 14 to 35 (ongoing remyelination) postsurgery, followed by histology. Image analysis focused on both lesions and peri-lesional areas where remyelination began. In histology, we quantified the complexity of tissue alignment using angular entropy, in addition to staining area. STATISTICAL ANALYSIS: Two-way analysis of variance was performed for assessing differences between tissue types and across timepoints, followed by post-hoc analysis to correct for multiple comparisons (P < 0.05). RESULTS: All diffusion and texture parameters were worse in lesions than the control tissue (P < 0.05) except orientation dispersion index (ODI) and neurite density index (NDI) over late remyelination. Longitudinally, ODI decreased and NDI increased persistently in both lesions and peri-lesion regions (P < 0.05). Fractional anisotropy showed a mild decrease at day 35 after increase, when lesion texture heterogeneity showed a trend to decrease (P > 0.05). Both lesion size and angular entropy decreased over time, and no change in any measure in the control tissue. DATA CONCLUSION: Diffusion and MRI texture metrics may provide compensatory information on myelin repair and ODI and NDI could be sensitive measures of evolving remyelination, deserving further validation. LEVEL OF EVIDENCE: 1 Technical Efficacy: Stage 1 J. Magn. Reson. Imaging 2019;49:1750-1759.

A discrete polar Stockwell transform for enhanced characterization of tissue structure using MRI.

PURPOSE: The purpose of this study was to present an effective algorithm for computing the discrete polar Stockwell transform (PST), investigate its unique multiscale and multi-orientation features, and explore potentially new applications including denoising and tissue segmentation. THEORY AND METHODS: We investigated PST responses using both synthetic and MR images. Moreover, we compared the features of PST with both Gabor and Morlet wavelet transforms, and compared the PST with two wavelet approaches for denoising using MRI. Using a synthetic image, we also tested the edge effect of PST through signal-padding. Then, we constructed a partially supervised classifier using radial, marginal PST spectra of T2-weighted MRI, acquired from postmortem brains with multiple sclerosis. The classification involved three histology-verified tissue types: normal appearing white matter (NAWM), lesion, or other, along with 5-fold cross-validation. RESULTS: The PST generated a series of images with varying orientations or rotation-invariant scales. Radial frequencies highlighted image structures of different size, and angular frequencies enhanced structures by orientation. Signal-padding helped suppress boundary artifacts but required attention to incidental artifacts. In comparison, the Gabor transform produced more redundant images and the wavelet spectra appeared less spatially smooth than the PST. In addition, the PST demonstrated lower root-mean-square errors than other transforms in denoising and achieved a 93% accuracy for NAWM pixels (296/317), and 88% accuracy for lesion pixels (165/188) in MRI segmentation. CONCLUSIONS: The PST is a unique local spectral density-assessing tool which is sensitive to both structure orientations and scales. This may facilitate multiple new applications including advanced characterization of tissue structure in standard MRI.