The publication data currently available has been vetted by Vanderbilt faculty, staff, administrators and trainees. The data itself is retrieved directly from NCBI's PubMed and is automatically updated on a weekly basis to ensure accuracy and completeness.
If you have any questions or comments, please contact us.
PURPOSE - As deep neural networks achieve more success in the wide field of computer vision, greater emphasis is being placed on the generalizations of these models for production deployment. With sufficiently large training datasets, models can typically avoid overfitting their data; however, for medical imaging it is often difficult to obtain enough data from a single site. Sharing data between institutions is also frequently nonviable or prohibited due to security measures and research compliance constraints, enforced to guard protected health information (PHI) and patient anonymity.
METHODS - In this paper, we implement cyclic weight transfer with independent datasets from multiple geographically disparate sites without compromising PHI. We compare results between single-site learning (SSL) and multisite learning (MSL) models on testing data drawn from each of the training sites as well as two other institutions.
RESULTS - The MSL model attains an average dice similarity coefficient (DSC) of 0.690 on the holdout institution datasets with a volume correlation of 0.914, respectively corresponding to a 7% and 5% statistically significant improvement over the average of both SSL models, which attained an average DSC of 0.646 and average correlation of 0.871.
CONCLUSIONS - We show that a neural network can be efficiently trained on data from two physically remote sites without consolidating patient data to a single location. The resulting network improves model generalization and achieves higher average DSCs on external datasets than neural networks trained on data from a single source.
© 2019 American Association of Physicists in Medicine.
Brown adipose tissue undergoes a dynamic, heterogeneous response to cold exposure that can include the simultaneous synthesis, uptake, and oxidation of fatty acids. The purpose of this work was to quantify these changes in brown adipose tissue lipid content (fat-signal fraction (FSF)) using fat-water magnetic resonance imaging during individualized cooling to 3 °C above a participant's shiver threshold. Eight healthy men completed familiarization, perception-based cooling, and MRI-cooling visits. FSF maps of the supraclavicular region were acquired in thermoneutrality and during cooling (59.5 ± 6.5 min). Brown adipose tissue regions of interest were defined, and voxels were grouped into FSF decades (0-10%, 10-20%…90-100%) according to their initial value. Brown adipose tissue contained a heterogeneous morphology of lipid content. Voxels with initial FSF values of 60-100% (P < 0.05) exhibited a significant decrease in FSF while a simultaneous increase in FSF occurred in voxels with initial FSF values of 0-30% (P < 0.05). These data suggest that in healthy young men, cold exposure elicits a dynamic and heterogeneous response in brown adipose tissue, with areas initially rich with lipid undergoing net lipid loss and areas of low initial lipid undergoing a net lipid accumulation.
Cognitive impairment (CI) is a major manifestation of multiple sclerosis (MS) and is responsible for extensively hindering patient quality of life. Cortical gray matter (cGM) damage is a significant contributor to CI, but is poorly characterized by conventional MRI let alone with quantitative MRI, such as quantitative magnetization transfer (qMT). Here we employed high-resolution qMT at 7T via the selective inversion recovery (SIR) method, which provides tissue-specific indices of tissue macromolecular content, such as the pool size ratio (PSR) and the rate of MT exchange (kmf). These indices could represent expected demyelination that occurs in the presence of gray matter damage. We utilized selective inversion recovery (SIR) qMT which provides a low SAR estimate of macromolecular-bulk water interactions using a tailored, B1 and B0 robust inversion recovery (IR) sequence acquired at multiple inversion times (TI) at 7T and fit to a two-pool model of magnetization exchange. Using this sequence, we evaluated qMT indices across relapsing-remitting multiple sclerosis patients (N = 19) and healthy volunteers (N = 37) and derived related associations with neuropsychological measures of cognitive impairment. We found a significant reduction in k in cGM of MS patients (15.5%, p = 0.002), unique association with EDSS (ρ = -0.922, p = 0.0001), and strong correlation with cognitive performance (ρ = -0.602, p = 0.0082). Together these findings indicate that the rate of MT exchange (k) may be a significant biomarker of cGM damage relating to CI in MS.
Copyright © 2019 Elsevier Inc. All rights reserved.
The combination of sodium salt doping of a tissue section along with the sublimation of the matrix 2,5-dihydrobenzoic acid (DHB) was found to be an effective coating for the simultaneous detection of neutral lipids and phospholipids using matrix-assisted laser desorption/ionization (MALDI) imaging mass spectrometry in positive ionization mode. Lithium, sodium, and potassium acetate were initially screened for their ability to cationize difficult to analyze neutral lipids such as cholesterol esters, cerebrosides, and triglycerides directly from a tissue section. The combination of sodium salt and DHB sublimation was found to be an effective cation/matrix combination for detection of neutral lipids. Further experimental optimizations revealed that sodium carbonate or sodium phosphate followed by DHB sublimation increases the signal intensity of the neutral lipids studied depending on the specific lipid family and tissue type by 10-fold to 140-fold compared with that of previously published methods. Application of sodium carbonate tissue doping and DHB sublimation resulted in crystal sizes ≤2 μm. We were thus able to image a mouse brain cerebellum at a high spatial resolution and detected 37 cerebrosides in a single run using a MALDI-TOF instrument. The combination of sodium doping and DHB sublimation offer a targeted and sensitive approach for the detection of neutral lipids that do not typically ionize well under normal MALDI conditions.
To characterize cell types, cellular functions and intracellular processes, an understanding of the differences between individual cells is required. Although microscopy approaches have made tremendous progress in imaging cells in different contexts, the analysis of these imaging data sets is a long-standing, unsolved problem. The few robust cell segmentation approaches that exist often rely on multiple cellular markers and complex time-consuming image analysis. Recently developed deep learning approaches can address some of these challenges, but they require tremendous amounts of data and well-curated reference data sets for algorithm training. We propose an alternative experimental and computational approach, called CellDissect, in which we first optimize specimen preparation and data acquisition prior to image processing to generate high quality images that are easier to analyze computationally. By focusing on fixed suspension and dissociated adherent cells, CellDissect relies only on widefield images to identify cell boundaries and nuclear staining to automatically segment cells in two dimensions and nuclei in three dimensions. This segmentation can be performed on a desktop computer or a computing cluster for higher throughput. We compare and evaluate the accuracy of different nuclear segmentation approaches against manual expert cell segmentation for different cell lines acquired with different imaging modalities.
We present hierarchical spherical deformation for a group-wise shape correspondence to address template selection bias and to minimize registration distortion. In this work, we aim at a continuous and smooth deformation field to guide accurate cortical surface registration. In conventional spherical registration methods, a global rigid alignment and local deformation are independently performed. Motivated by the composition of precession and intrinsic rotation, we simultaneously optimize global rigid rotation and non-rigid local deformation by utilizing spherical harmonics interpolation of local composite rotations in a single framework. To this end, we indirectly encode local displacements by such local composite rotations as functions of spherical locations. Furthermore, we introduce an additional regularization term to the spherical deformation, which maximizes its rigidity while reducing registration distortion. To improve surface registration performance, we employ the second order approximation of the energy function that enables fast convergence of the optimization. In the experiments, we validate our method on healthy normal subjects with manual cortical surface parcellation in registration accuracy and distortion. We show an improved shape correspondence with high accuracy in cortical surface parcellation and significantly low registration distortion in surface area and edge length. In addition to validation, we discuss parameter tuning, optimization, and implementation design with potential acceleration.
Copyright © 2019 Elsevier B.V. All rights reserved.
Imaging apoptosis could provide an early and specific means to monitor tumor responses to treatment. To date, despite numerous attempts to develop molecular imaging approaches, there is still no widely-accepted and reliable method for in vivo imaging of apoptosis. We hypothesized that the distinct cellular morphologic changes associated with treatment-induced apoptosis, such as cell shrinkage, cytoplasm condensation, and DNA fragmentation, can be detected by temporal diffusion spectroscopy imaging (TDSI). Cetuximab-induced apoptosis was assessed in vitro and in vivo with cetuximab-sensitive (DiFi) and insensitive (HCT-116) human colorectal cancer cell lines by TDSI. TDSI findings were complemented by flow cytometry and immunohistochemistry. Cell cycle analysis and flow cytometry detected apoptotic cell shrinkage in cetuximab-treated DiFi cells, and significant apoptosis was confirmed by histology. TDSI-derived parameters quantified key morphological changes including cell size decreases during apoptosis in responsive tumors that occurred earlier than gross tumor volume regression. TDSI provides a unique measurement of apoptosis by identifying cellular characteristics, particularly cell shrinkage. The method will assist in understanding the underlying biology of solid tumors and predict tumor response to therapies. TDSI is free of any exogenous agent or radiation, and hence is very suitable to be incorporated into clinical applications.
PURPOSE - Manually tracing regions of interest (ROIs) within the liver is the de facto standard method for measuring liver attenuation on computed tomography (CT) in diagnosing nonalcoholic fatty liver disease (NAFLD). However, manual tracing is resource intensive. To address these limitations and to expand the availability of a quantitative CT measure of hepatic steatosis, we propose the automatic liver attenuation ROI-based measurement (ALARM) method for automated liver attenuation estimation.
METHODS - The ALARM method consists of two major stages: (a) deep convolutional neural network (DCNN)-based liver segmentation and (b) automated ROI extraction. First, liver segmentation was achieved using our previously developed SS-Net. Then, a single central ROI (center-ROI) and three circles ROI (periphery-ROI) were computed based on liver segmentation and morphological operations. The ALARM method is available as an open source Docker container (https://github.com/MASILab/ALARM).
RESULTS - Two hundred and forty-six subjects with 738 abdomen CT scans from the African American-Diabetes Heart Study (AA-DHS) were used for external validation (testing), independent from the training and validation cohort (100 clinically acquired CT abdominal scans). From the correlation analyses, the proposed ALARM method achieved Pearson correlations = 0.94 with manual estimation on liver attenuation estimations. When evaluating the ALARM method for detection of nonalcoholic fatty liver disease (NAFLD) using the traditional cut point of < 40 HU, the center-ROI achieved substantial agreements (Kappa = 0.79) with manual estimation, while the periphery-ROI method achieved "excellent" agreement (Kappa = 0.88) with manual estimation. The automated ALARM method had reduced variability compared to manual measurements as indicated by a smaller standard deviation.
CONCLUSIONS - We propose a fully automated liver attenuation estimation method termed ALARM by combining DCNN and morphological operations, which achieved "excellent" agreement with manual estimation for fatty liver detection. The entire pipeline is implemented as a Docker container which enables users to achieve liver attenuation estimation in five minutes per CT exam.
© 2019 American Association of Physicists in Medicine.
Correlations between fluctuations in resting state BOLD fMRI signals are interpreted as measures of functional connectivity (FC), but the neural basis of their origins and their relationships to specific features of underlying electrophysiologic activity, have not been fully established. In particular, the dependence of FC metrics on different frequency bands of local field potentials (LFPs), and the relationship of dynamic changes in BOLD FC to underlying temporal variations of LFP correlations, are not known. We compared the spatial profiles of resting state coherences of different frequency bands of LFP signals, with high resolution resting state BOLD FC measurements. We also compared the probability distributions of temporal variations of connectivity in both modalities using a Markov chain model-based approach. We analyzed data obtained from the primary somatosensory (S1) cortex of monkeys. We found that in areas 3b and 1 of S1 cortex, low frequency LFP signal fluctuations were the main contributions to resting state LFP coherence. Additionally, the dynamic changes of BOLD FC behaved most similarly to the LFP low frequency signal coherence. These results indicate that, within the S1 cortex meso-scale circuit studied, resting state FC measures from BOLD fMRI mainly reflect contributions from low frequency LFP signals and their dynamic changes.
BACKGROUND - Human cortical primary sulci are relatively stable landmarks and commonly observed across the population. Despite their stability, the primary sulci exhibit phenotypic variability.
NEW METHOD - We propose a fully automated pipeline that integrates both sulcal curve extraction and labeling. In this study, we use a large normal control population (n = 1424) to train neural networks for accurately labeling the primary sulci. Briefly, we use sulcal curve distance map, surface parcellation, mean curvature and spectral features to delineate their sulcal labels. We evaluate the proposed method with 8 primary sulcal curves in the left and right hemispheres compared to an established multi-atlas curve labeling method.
RESULTS - Sulcal labels by the proposed method reasonably well agree with manual labeling. The proposed method outperforms the existing multi-atlas curve labeling method.
COMPARISON WITH EXISTING METHOD - Significantly improved sulcal labeling results are achieved with over 12.5 and 20.6 percent improvement on labeling accuracy in the left and right hemispheres, respectively compared to that of a multi-atlas curve labeling method in eight curves (p≪0.001, two-sample t-test).
CONCLUSION - The proposed method offers a computationally efficient and robust labeling of major sulci.
Copyright © 2019 Elsevier B.V. All rights reserved.