Multi-modal Pulmonary Imaging: Using Complementary Information from CT and Hyperpolarized 129Xe MRI to Evaluate Lung Structure-Function.

Hyperpolarized 129Xe gas MRI is an emerging technique to evaluate and measure regional lung function including pulmonary gas distribution and gas exchange. Chest computed tomography (CT) still remains the clinical gold standard for imaging of the lungs, though, in part due to the rapid CT protocols that acquire high-resolution images in seconds and the widespread availability of CT scanners. Quantitative approaches have enabled the extraction of structural lung parenchymal, airway and vascular measurements from chest CT that have been evaluated in many clinical research studies. Together, CT and 129Xe MRI provide complementary information that can be used to evaluate regional lung structure and function, resulting in new insights into lung health and disease. 129Xe MR-CT image registration can be performed to measure regional lung structure-function to better understand lung disease pathophysiology, and to perform image-guided pulmonary interventions. Here, a method for 129Xe MRI-CT registration is outlined to support implementation in research or clinical settings. Registration methods and applications that have been employed to date in the literature are also summarized, and suggestions are provided for future directions that may further overcome technical challenges related to 129Xe MR-CT image registration and facilitate broader implementation of regional lung structure-function evaluation.

Accounting for uncertainty in training data to improve machine learning performance in predicting new disease activity in early multiple sclerosis.

Introduction: Machine learning (ML) has great potential for using health data to predict clinical outcomes in individual patients. Missing data are a common challenge in training ML algorithms, such as when subjects withdraw from a clinical study, leaving some samples with missing outcome labels. In this study, we have compared three ML models to determine whether accounting for label uncertainty can improve a model's predictions. Methods: We used a dataset from a completed phase-III clinical trial that evaluated the efficacy of minocycline for delaying the conversion from clinically isolated syndrome to multiple sclerosis (MS), using the McDonald 2005 diagnostic criteria. There were a total of 142 participants, and at the 2-year follow-up 81 had converted to MS, 29 remained stable, and 32 had uncertain outcomes. In a stratified 7-fold cross-validation, we trained three random forest (RF) ML models using MRI volumetric features and clinical variables to predict the conversion outcome, which represented new disease activity within 2 years of a first clinical demyelinating event. One RF was trained using subjects with the uncertain labels excluded (RFexclude), another RF was trained using the entire dataset but with assumed labels for the uncertain group (RFnaive), and a third, a probabilistic RF (PRF, a type of RF that can model label uncertainty) was trained on the entire dataset, with probabilistic labels assigned to the uncertain group. Results: Probabilistic random forest outperformed both the RF models with the highest AUC (0.76, compared to 0.69 for RFexclude and 0.71 for RFnaive) and F1-score (86.6% compared to 82.6% for RFexclude and 76.8% for RFnaive). Conclusion: Machine learning algorithms capable of modeling label uncertainty can improve predictive performance in datasets in which a substantial number of subjects have unknown outcomes.

Myelin and Physical Activity in Older Adults With Cerebral Small Vessel Disease and Mild Cognitive Impairment.

BACKGROUND: Myelin loss is a feature of cerebral small vessel disease (cSVD). Although physical activity levels may exert protective effects over cSVD pathology, its specific relationship with myelin content in people living with the cSVD is unknown. Thus, we investigated whether physical activity levels are associated with myelin in community-dwelling older adults with cSVD and mild cognitive impairment. METHODS: Cross-sectional data from 102 individuals with cSVD and mild cognitive impairment were analyzed (mean age [SD] = 74.7 years [5.5], 63.7% female). Myelin was measured using a magnetic resonance gradient and spin echo sequence. Physical activity was estimated using the Physical Activity Scale for the Elderly. Hierarchical regression models adjusting for total intracranial volume, age, sex, body mass index, and education were conducted to determine the associations between myelin content and physical activity. Significant models were further adjusted for white matter hyperintensity volume. RESULTS: In adjusted models, greater physical activity was linked to higher myelin content in the whole-brain white matter (R2change = .04, p = .048). Greater physical activity was also associated with myelin content in the sagittal stratum (R2change = .08, p = .004), anterior corona radiata (R2change = .04, p = .049), and genu of the corpus callosum (R2change = .05, p = .018). Adjusting for white matter hyperintensity volume did not change any of these associations. CONCLUSIONS: Physical activity may be a strategy to maintain myelin in older adults with cSVD and mild cognitive impairment. Future randomized controlled trials of exercise are needed to determine whether exercise increases myelin content.

Myelin Content and Gait Impairment in Older Adults with Cerebral Small Vessel Disease and Mild Cognitive Impairment.

We investigated whether myelin is associated with gait parameters in older adults with cerebral small vessel disease (cSVD). Cross-sectional data from sixty-four participants with cSVD and mild cognitive impairment were analyzed. Myelin was assessed via MRI multi-echo gradient and spin echo T2 relaxation sequence, indexed as myelin water fraction (MWF). Gait was assessed using an electronic walkway. Hierarchical regression models adjusting for total intracranial volume, age, sex, Mini-Mental State Examination, and body mass index were conducted to determine associations between MWF and gait parameters. Significant models were further adjusted for white matter hyperintensities. Sixty-four participants were included (mean [SD], age = 75.2y [5.4], 62.5% female). In adjusted models, lower MWF in the cingulum (p = 0.015), superior longitudinal fasciculus (p = 0.034), posterior corona radiata (p = 0.039), and body of the corpus callosum (p = 0.040) was associated with higher cycle time variability. White matter hyperintensities weakened these associations. Lower myelin in specific white matter tracts may contribute to higher gait variability, increasing the overall risk of mobility impairment.

Reshaping the path of vascular cognitive impairment with resistance training: a study protocol for a randomized controlled trial.

BACKGROUND: Subcortical ischemic vascular cognitive impairment (SIVCI) is the most common form of vascular cognitive impairment. Importantly, SIVCI is considered the most treatable form of cognitive impairment in older adults, due to its modifiable risk factors such as hypertension, diabetes mellitus, and hypercholesterolemia. Exercise training is a promising intervention to delay the progression of SIVCI, as it actively targets these cardiometabolic risk factors. Despite the demonstrated benefits of resistance training on cognitive function and emerging evidence suggesting resistance training may reduce the progression of white matter hyperintensities (WMHs), research on SIVCI has predominantly focused on the use of aerobic exercise. Thus, the primary aim of this proof-of-concept randomized controlled trial is to investigate the efficacy of a 12-month, twice-weekly progressive resistance training program on cognitive function and WMH progression in adults with SIVCI. We will also assess the efficiency of the intervention. METHODS: Eighty-eight community-dwelling adults, aged > 55 years, with SIVCI from metropolitan Vancouver will be recruited to participate in this study. SIVCI will be determined by the presence of cognitive impairment (Montreal Cognitive Assessment < 26) and cerebral small vessel disease using computed tomography or magnetic resonance imaging. Participants will be randomly allocated to a twice-weekly exercise program of (1) progressive resistance training or (2) balance and tone training (i.e., active control). The primary outcomes are cognitive function measured by the Alzheimer's Disease Assessment Scale-Cognitive-Plus (ADAS-Cog-13 with additional cognitive tests) and WMH progression. DISCUSSION: The burden of SIVCI is immense, and to our knowledge, this will be the first study to quantify the effect of progressive resistance training on cognitive function and WMH progression among adults with SIVCI. Slowing the rate of cognitive decline and WMH progression could preserve functional independence and quality of life. This could lead to reduced health care costs and avoidance of early institutional care. TRIAL REGISTRATION: ClinicalTrials.gov NCT02669394 . Registered on February 1, 2016.

Towards large-scale case-finding: training and validation of residual networks for detection of chronic obstructive pulmonary disease using low-dose CT.

BACKGROUND: Chronic obstructive pulmonary disease (COPD) is underdiagnosed in the community. Thoracic CT scans are widely used for diagnostic and screening purposes for lung cancer. In this proof-of-concept study, we aimed to evaluate a software pipeline for the automated detection of COPD, based on deep learning and a dataset of low-dose CTs that were performed for early detection of lung cancer. METHODS: We examined the use of deep residual networks, a type of artificial residual network, for the automated detection of COPD. Three versions of the residual networks were independently trained to perform COPD diagnosis using random subsets of CT scans collected from the PanCan study, which enrolled ex-smokers and current smokers at high risk of lung cancer, and evaluated the networks using three-fold cross-validation experiments. External validation was performed using 2153 CT scans acquired from a separate cohort of individuals with COPD in the ECLIPSE study. Spirometric data were used to define COPD, with stages defined according to the GOLD criteria. FINDINGS: The best performing networks achieved an area under the receiver operating characteristic curve (AUC) of 0·889 (SD 0·017) in three-fold cross-validation experiments. When the same set of networks was applied to the ECLIPSE cohort without any modifications to the trained models, they achieved an AUC of 0·886 (0·017), a positive predictive value of 0·847 (0·056), and a negative predictive value of 0·755 (0·097), which is a greater performance than the best quantitative CT measure, the percentage of lung volumes of less than or equal to -950 Hounsfield units (AUC 0·742). INTERPRETATION: Our proposed approach could identify patients with COPD among ex-smokers and current smokers without a previous diagnosis of COPD, with clinically acceptable performance. The use of deep residual networks on chest CT scans could be an effective case-finding tool for COPD detection and diagnosis, particularly in ex-smokers and current smokers who are being screened for lung cancer. FUNDING: Data Science Institute, University of British Columbia; Canadian Institutes of Health Research.

Deep learning of joint myelin and T1w MRI features in normal-appearing brain tissue to distinguish between multiple sclerosis patients and healthy controls.

Myelin imaging is a form of quantitative magnetic resonance imaging (MRI) that measures myelin content and can potentially allow demyelinating diseases such as multiple sclerosis (MS) to be detected earlier. Although focal lesions are the most visible signs of MS pathology on conventional MRI, it has been shown that even tissues that appear normal may exhibit decreased myelin content as revealed by myelin-specific images (i.e., myelin maps). Current methods for analyzing myelin maps typically use global or regional mean myelin measurements to detect abnormalities, but ignore finer spatial patterns that may be characteristic of MS. In this paper, we present a machine learning method to automatically learn, from multimodal MR images, latent spatial features that can potentially improve the detection of MS pathology at early stage. More specifically, 3D image patches are extracted from myelin maps and the corresponding T1-weighted (T1w) MRIs, and are used to learn a latent joint myelin-T1w feature representation via unsupervised deep learning. Using a data set of images from MS patients and healthy controls, a common set of patches are selected via a voxel-wise t-test performed between the two groups. In each MS image, any patches overlapping with focal lesions are excluded, and a feature imputation method is used to fill in the missing values. A feature selection process (LASSO) is then utilized to construct a sparse representation. The resulting normal-appearing features are used to train a random forest classifier. Using the myelin and T1w images of 55 relapse-remitting MS patients and 44 healthy controls in an 11-fold cross-validation experiment, the proposed method achieved an average classification accuracy of 87.9% (SD = 8.4%), which is higher and more consistent across folds than those attained by regional mean myelin (73.7%, SD = 13.7%) and T1w measurements (66.7%, SD = 10.6%), or deep-learned features in either the myelin (83.8%, SD = 11.0%) or T1w (70.1%, SD = 13.6%) images alone, suggesting that the proposed method has strong potential for identifying image features that are more sensitive and specific to MS pathology in normal-appearing brain tissues.

Fast computation of myelin maps from MRI T2 relaxation data using multicore CPU and graphics card parallelization.

PURPOSE: To develop a fast algorithm for computing myelin maps from multiecho T2 relaxation data using parallel computation with multicore CPUs and graphics processing units (GPUs). MATERIALS AND METHODS: Using an existing MATLAB (MathWorks, Natick, MA) implementation with basic (nonalgorithm-specific) parallelism as a guide, we developed a new version to perform the same computations but using C++ to optimize the hybrid utilization of multicore CPUs and GPUs, based on experimentation to determine which algorithmic components would benefit from CPU versus GPU parallelization. Using 32-echo T2 data of dimensions 256 × 256 × 7 from 17 multiple sclerosis patients and 18 healthy subjects, we compared the two methods in terms of speed, myelin values, and the ability to distinguish between the two patient groups using Student's t-tests. RESULTS: The new method was faster than the MATLAB implementation by 4.13 times for computing a single map and 14.36 times for batch-processing 10 scans. The two methods produced very similar myelin values, with small and explainable differences that did not impact the ability to distinguish the two patient groups. CONCLUSION: The proposed hybrid multicore approach represents a more efficient alternative to MATLAB, especially for large-scale batch processing.

Improving the clinical correlation of multiple sclerosis black hole volume change by paired-scan analysis.

The change in T 1-hypointense lesion ("black hole") volume is an important marker of pathological progression in multiple sclerosis (MS). Black hole boundaries often have low contrast and are difficult to determine accurately and most (semi-)automated segmentation methods first compute the T 2-hyperintense lesions, which are a superset of the black holes and are typically more distinct, to form a search space for the T 1w lesions. Two main potential sources of measurement noise in longitudinal black hole volume computation are partial volume and variability in the T 2w lesion segmentation. A paired analysis approach is proposed herein that uses registration to equalize partial volume and lesion mask processing to combine T 2w lesion segmentations across time. The scans of 247 MS patients are used to compare a selected black hole computation method with an enhanced version incorporating paired analysis, using rank correlation to a clinical variable (MS functional composite) as the primary outcome measure. The comparison is done at nine different levels of intensity as a previous study suggests that darker black holes may yield stronger correlations. The results demonstrate that paired analysis can strongly improve longitudinal correlation (from -0.148 to -0.303 in this sample) and may produce segmentations that are more sensitive to clinically relevant changes.

Optimizing the use of radiologist seed points for improved multiple sclerosis lesion segmentation.

Many current methods for multiple sclerosis (MS) lesion segmentation require radiologist seed points as input, but do not necessarily allow the expert to work in an intuitive or efficient way. Ironically, most methods also assume that the points are placed optimally. This paper examines how seed points can be processed with intuitive heuristics, which provide improved segmentation accuracy while facilitating quick and natural point placement. Using a large set of MRIs from an MS clinical trial, two radiologists are asked to seed the lesions while unaware that the points would be fed into a classifier, based on Parzen windows, that automatically delineates each marked lesion. To evaluate the impact of the new heuristics, an interactive region-growing method is used to provide ground truth and the Dice coefficient (DC) and Spearman's rank correlation are used as the primary measures of agreement. A stratified analysis is performed to determine the effect on scans with low-, medium-, and high lesion loads. Compared to the unenhanced classifier, the heuristics dramatically improve the DC (+32.91 pt.) and correlation (+0.50) for the scans with low lesion loads, and also improve the DC (+14.55 pt.) and correlation (+0.15) for the scans with medium lesion loads, while having aminimal effect for the scans with high lesion loads, which are already segmented accurately by Parzen windows.With the heuristics, the DC is close to 80% and the correlation is above 0.9 for all three load categories.

Detection and measurement of coverage loss in interleaved multi-acquisition brain MRIs due to motion-induced inter-slice misalignment.

In MRI scans that are acquired in a slice-by-slice manner, patient motion during scanning can cause adjacent slices to overlap, resulting in duplicate coverage in some areas and missing coverage in others. Scans in which multiple slices are acquired simultaneously and interleaved with other sets of slices are particularly vulnerable because a single movement can result in the misalignment and overlap of many slices. Despite the fact that considerable data losses can occur even with few visible artifacts, this problem has received very little attention from MRI researchers. The primary goals of this paper are: (1) to raise awareness of the problem in the MRI community and (2) to present an efficient multiscale algorithm that accurately quantifies the amount of data loss. Validation of the algorithm's accuracy is performed on 200 scans with simulated patient motion so that the true amount of data loss is known for each scan. The motion parameters are chosen to simulate scans that have significant data loss (mean missing coverage=14.39% of head volume, SD=6.61%, range=2.76-32.98%) but with few visual indications of the problem. The algorithm is shown to be very accurate, yielding estimates that differ from the true values by a mean of only 1.1% point (SD=0.98pt, range=0.00-6.54pt). The algorithm is also shown to be consistent and robust when tested on a large set of scans from a recent multiple sclerosis clinical trial.