Effect of Human Head Shape on the Risk of Traumatic Brain Injury: A Gaussian Process Regression-based Machine Learning Approach

INTRODUCTION

Computational head injury models are promising tools for understanding and predicting traumatic brain injuries. However, most available head injury models are "average" models that employ a single set of head geometry (e.g., 50th-percentile U.S. male) without considering variability in these parameters across the human population. A significant variability of head shapes exists in U.S. Army soldiers, evident from the Anthropometric Survey of U.S. Army Personnel (ANSUR II). The objective of this study is to elucidate the effects of head shape on the predicted risk of traumatic brain injury from computational head injury models.

MATERIALS AND METHODS

Magnetic resonance imaging scans of 25 human subjects are collected. These images are registered to the standard MNI152 brain atlas, and the resulting transformation matrix components (called head shape parameters) are used to quantify head shapes of the subjects. A generative machine learning model is used to generate 25 additional head shape parameter datasets to augment our database. Head injury models are developed for these head shapes, and a rapid injurious head rotation event is simulated to obtain several brain injury predictor variables (BIPVs): Peak cumulative maximum principal strain (CMPS), average CMPS, and the volume fraction of brain exceeding an injurious CMPS threshold. A Gaussian process regression model is trained between head shape parameters and BIPVs, which is then used to study the relative sensitivity of the various BIPVs on individual head shape parameters. We distinguish head shape parameters into 2 types: Scaling components ${T_{xx}}$, ${T_{yy}}$, and ${T_{zz}}$ that capture the breadth, length, and height of the head, respectively, and shearing components (${T_{xy}},{T_{xz}},{T_{yx}},{T_{yz}},{T_{zx}}$, and ${T_{zy}}$) that capture the relative skewness of the head shape.

RESULTS

An overall positive correlation is evident between scaling components and BIPVs. Notably, a very high, positive correlation is seen between the BIPVs and the head volume. As an example, a 57% increase in peak CMPS was noted between the smallest and the largest investigated head volume parameters. The variation in shearing components ${T_{xy}},{T_{xz}},{T_{yx}},{T_{yz}},{T_{zx}}$, and ${T_{zy}}$ on average does not cause notable changes in the BIPVs. From the Gaussian process regression model, all 3 BIPVs showed an increasing trend with each of the 3 scaling components, but the BIPVs are found to be most sensitive to the height dimension of the head. From the Sobol sensitivity analysis, the ${T_{zz}}$ scaling parameter contributes nearly 60% to the total variance in peak and average CMPS; ${T_{yy}}$ contributes approximately 20%, whereas ${T_{xx}}$ contributes less than 5%. The remaining contribution is from the 6 shearing components. Unlike peak and average CMPS, the VF-CMPS BIPV is associated with relatively evenly distributed Sobol indices across the 3 scaling parameters. Furthermore, the contribution of shearing components on the total variance in this case is negligible.

CONCLUSIONS

Head shape has a considerable influence on the injury predictions of computational head injury models. Available "average" head injury models based on a 50th-percentile U.S. male are likely associated with considerable uncertainty. In general, larger head sizes correspond to greater BIPV magnitudes, which point to potentially a greater injury risk under rapid neck rotation for people with larger heads.

Image harmonization improves consistency of intra-rater delineations of MS lesions in heterogeneous MRI

Clinical magnetic resonance images (MRIs) lack a standard intensity scale due to differences in scanner hardware and the pulse sequences used to acquire the images. When MRIs are used for quantification, as in the evaluation of white matter lesions (WMLs) in multiple sclerosis, this lack of intensity standardization becomes a critical problem affecting both the staging and tracking of the disease and its treatment. This paper presents a study of harmonization on WML segmentation consistency, which is evaluated using an object detection classification scheme that incorporates manual delineations from both the original and harmonized MRIs. A cohort of ten people scanned on two different imaging platforms was studied. An expert rater, blinded to the image source, manually delineated WMLs on images from both scanners before and after harmonization. It was found that there is closer agreement in both global and per-lesion WML volume and spatial distribution after harmonization, demonstrating the importance of image harmonization prior to the creation of manual delineations. These results could lead to better truth models in both the development and evaluation of automated lesion segmentation algorithms.

Effect of Direction and Frequency of Skull Motion on Mechanical Vulnerability of the Human Brain

Strain energy and kinetic energy in the human brain were estimated by magnetic resonance elastography (MRE) during harmonic excitation of the head, and compared to characterize the effect of loading direction and frequency on brain deformation. In brain MRE, shear waves are induced by external vibration of the skull and imaged by a modified MR imaging sequence; the resulting harmonic displacement fields are typically "inverted" to estimate mechanical properties, like stiffness or damping. However, measurements of tissue motion from MRE also illuminate key features of the response of the brain to skull loading. In this study, harmonic excitation was applied in two different directions and at five different frequencies from 20 to 90 Hz. Lateral loading induced primarily left-right head motion and rotation in the axial plane; occipital loading induced anterior-posterior head motion and rotation in the sagittal plane. The ratio of strain energy to kinetic energy (SE/KE) depended strongly on both direction and frequency. The ratio of SE/KE was approximately four times larger for lateral excitation than for occipital excitation and was largest at the lowest excitation frequencies studied. These results are consistent with clinical observations that suggest lateral impacts are more likely to cause injury than occipital or frontal impacts, and also with observations that the brain has low-frequency (∼10 Hz) natural modes of oscillation. The SE/KE ratio from brain MRE is potentially a simple and powerful dimensionless metric of brain vulnerability to deformation and injury.

Reproducibility evaluation of the effects of MRI defacing on brain segmentation

Purpose

Recent advances in magnetic resonance (MR) scanner quality and the rapidly improving nature of facial recognition software have necessitated the introduction of MR defacing algorithms to protect patient privacy. As a result, there are a number of MR defacing algorithms available to the neuroimaging community, with several appearing in just the last 5 years. While some qualities of these defacing algorithms, such as patient identifiability, have been explored in the previous works, the potential impact of defacing on neuroimage processing has yet to be explored.

Approach

We qualitatively evaluate eight MR defacing algorithms on 179 subjects from the OASIS-3 cohort and 21 subjects from the Kirby-21 dataset. We also evaluate the effects of defacing on two neuroimaging pipelines-SLANT and FreeSurfer-by comparing the segmentation consistency between the original and defaced images.

Results

Defacing can alter brain segmentation and even lead to catastrophic failures, which are more frequent with some algorithms, such as Quickshear, MRI_Deface, and FSL_deface. Compared to FreeSurfer, SLANT is less affected by defacing. On outputs that pass the quality check, the effects of defacing are less pronounced than those of rescanning, as measured by the Dice similarity coefficient.

Conclusions

The effects of defacing are noticeable and should not be disregarded. Extra attention, in particular, should be paid to the possibility of catastrophic failures. It is crucial to adopt a robust defacing algorithm and perform a thorough quality check before releasing defaced datasets. To improve the reliability of analysis in scenarios involving defaced MRIs, it is encouraged to include multiple brain segmentation pipelines.

Recurrent Self Fusion: Iterative Denoising for Consistent Retinal OCT Segmentation

Optical coherence tomography (OCT) is a valuable imaging technique in ophthalmology, providing high-resolution, cross-sectional images of the retina for early detection and monitoring of various retinal and neurological diseases. However, discrepancies in retinal layer thickness measurements among different OCT devices pose challenges for data comparison and interpretation, particularly in longitudinal analyses. This work introduces the idea of a recurrent self fusion (RSF) algorithm to address this issue. Our RSF algorithm, built upon the self fusion methodology, iteratively denoises retinal OCT images. A deep learning-based retinal OCT segmentation algorithm is employed for downstream analyses. A large dataset of paired OCT scans acquired on both a Spectralis and Cirrus OCT device are used for validation. The results demonstrate that the RSF algorithm effectively reduces speckle contrast and enhances the consistency of retinal OCT segmentation.

HACA3: A unified approach for multi-site MR image harmonization

The lack of standardization and consistency of acquisition is a prominent issue in magnetic resonance (MR) imaging. This often causes undesired contrast variations in the acquired images due to differences in hardware and acquisition parameters. In recent years, image synthesis-based MR harmonization with disentanglement has been proposed to compensate for the undesired contrast variations. The general idea is to disentangle anatomy and contrast information from MR images to achieve cross-site harmonization. Despite the success of existing methods, we argue that major improvements can be made from three aspects. First, most existing methods are built upon the assumption that multi-contrast MR images of the same subject share the same anatomy. This assumption is questionable, since different MR contrasts are specialized to highlight different anatomical features. Second, these methods often require a fixed set of MR contrasts for training (e.g., both T1-weighted and T2-weighted images), limiting their applicability. Lastly, existing methods are generally sensitive to imaging artifacts. In this paper, we present Harmonization with Attention-based Contrast, Anatomy, and Artifact Awareness (HACA3), a novel approach to address these three issues. HACA3 incorporates an anatomy fusion module that accounts for the inherent anatomical differences between MR contrasts. Furthermore, HACA3 can be trained and applied to any combination of MR contrasts and is robust to imaging artifacts. HACA3 is developed and evaluated on diverse MR datasets acquired from 21 sites with varying field strengths, scanner platforms, and acquisition protocols. Experiments show that HACA3 achieves state-of-the-art harmonization performance under multiple image quality metrics. We also demonstrate the versatility and potential clinical impact of HACA3 on downstream tasks including white matter lesion segmentation for people with multiple sclerosis and longitudinal volumetric analyses for normal aging subjects. Code is available at https://github.com/lianruizuo/haca3.

A reproducibility evaluation of the effects of MRI defacing on brain segmentation

Purpose

Recent advances in magnetic resonance (MR) scanner quality and the rapidly improving nature of facial recognition software have necessitated the introduction of MR defacing algorithms to protect patient privacy. As a result, there are a number of MR defacing algorithms available to the neuroimaging community, with several appearing in just the last five years. While some qualities of these defacing algorithms, such as patient identifiability, have been explored in previous works, the potential impact of defacing on neuroimage processing has yet to be explored.

Approach

We qualitatively evaluate eight MR defacing algorithms on 179 subjects from the OASIS-3 cohort and the 21 subjects from the Kirby-21 dataset. We also evaluate the effects of defacing on two neuroimaging pipelines-SLANT and FreeSurfer-by comparing the segmentation consistency between the original and defaced images.

Results

Defacing can alter brain segmentation and even lead to catastrophic failures, which are more frequent with some algorithms such as Quickshear, MRI_Deface, and FSL_deface. Compared to FreeSurfer, SLANT is less affected by defacing. On outputs that pass the quality check, the effects of defacing are less pronounced than those of rescanning, as measured by the Dice similarity coefficient.

Conclusions

The effects of defacing are noticeable and should not be disregarded. Extra attention, in particular, should be paid to the possibility of catastrophic failures. It is crucial to adopt a robust defacing algorithm and perform a thorough quality check before releasing defaced datasets. To improve the reliability of analysis in scenarios involving defaced MRIs, it's encouraged to include multiple brain segmentation pipelines.

Longitudinal deep network for consistent OCT layer segmentation

Retinal layer thickness is an important bio-marker for people with multiple sclerosis (PwMS). In clinical practice, retinal layer thickness changes in optical coherence tomography (OCT) are widely used for monitoring multiple sclerosis (MS) progression. Recent developments in automated retinal layer segmentation algorithms allow cohort-level retina thinning to be observed in a large study of PwMS. However, variability in these results make it difficult to identify patient-level trends; this prevents patient specific disease monitoring and treatment planning using OCT. Deep learning based retinal layer segmentation algorithms have achieved state-of-the-art accuracy, but the segmentation is performed on each individual scan without utilizing longitudinal information, which can be important in reducing segmentation error and reveal subtle changes in retinal layers. In this paper, we propose a longitudinal OCT segmentation network which achieves more accurate and consistent layer thickness measurements for PwMS.

ESPRESO: An algorithm to estimate the slice profile of a single magnetic resonance image

To reduce scan time, magnetic resonance (MR) images are often acquired using 2D multi-slice protocols with thick slices that may also have gaps between them. The resulting image volumes have lower resolution in the through-plane direction than in the in-plane direction, and the through-plane resolution is in part characterized by the protocol's slice profile which acts as a through-plane point spread function (PSF). Although super-resolution (SR) has been shown to improve the visualization and down-stream processing of 2D multi-slice MR acquisitions, previous algorithms are usually unaware of the true slice profile, which may lead to sub-optimal SR performance. In this work, we present an algorithm to estimate the slice profile of a 2D multi-slice acquisition given only its own image volume without any external training data. We assume that an anatomical image is isotropic in the sense that, after accounting for a correctly estimated slice profile, the image patches along different orientations have the same probability distribution. Our proposed algorithm uses a modified generative adversarial network (GAN) where the generator network estimates the slice profile to reduce the resolution of the in-plane direction, and the discriminator network determines whether a direction is generated or real low resolution. The proposed algorithm, ESPRESO, which stands for "estimating the slice profile for resolution enhancement of a single image only", was tested with a state-of-the-art internally supervised SR algorithm. Specifically, ESPRESO is used to create training data for this SR algorithm, and results show improvements when ESPRESO is used over commonly-used PSFs.

RAPID BRAIN MENINGES SURFACE RECONSTRUCTION WITH LAYER TOPOLOGY GUARANTEE

The meninges, located between the skull and brain, are composed of three membrane layers: the pia, the arachnoid, and the dura. Reconstruction of these layers can aid in studying volume differences between patients with neurodegenerative diseases and normal aging subjects. In this work, we use convolutional neural networks (CNNs) to reconstruct surfaces representing meningeal layer boundaries from magnetic resonance (MR) images. We first use the CNNs to predict the signed distance functions (SDFs) representing these surfaces while preserving their anatomical ordering. The marching cubes algorithm is then used to generate continuous surface representations; both the subarachnoid space (SAS) and the intracranial volume (ICV) are computed from these surfaces. The proposed method is compared to a state-of-the-art deformable model-based reconstruction method, and we show that our method can reconstruct smoother and more accurate surfaces using less computation time. Finally, we conduct experiments with volumetric analysis on both subjects with multiple sclerosis and healthy controls. For healthy and MS subjects, ICVs and SAS volumes are found to be significantly correlated to sex (p<0.01) and age (p ≤ 0.03) changes, respectively.

AUTOMATED VENTRICLE PARCELLATION AND EVAN'S RATIO COMPUTATION IN PRE- AND POST-SURGICAL VENTRICULOMEGALY

Normal pressure hydrocephalus (NPH) is a brain disorder associated with enlarged ventricles and multiple cognitive and motor symptoms. The degree of ventricular enlargement can be measured using magnetic resonance images (MRIs) and characterized quantitatively using the Evan's ratio (ER). Automatic computation of ER is desired to avoid the extra time and variations associated with manual measurements on MRI. Because shunt surgery is often used to treat NPH, it is necessary that this process be robust to image artifacts caused by the shunt and related implants. In this paper, we propose a 3D regions-of-interest aware (ROI-aware) network for segmenting the ventricles. The method achieves state-of-the-art performance on both pre-surgery MRIs and post-surgery MRIs with artifacts. Based on our segmentation results, we also describe an automated approach to compute ER from these results. Experimental results on multiple datasets demonstrate the potential of the proposed method to assist clinicians in the diagnosis and management of NPH.

Label Propagation via Random Walk for Training Robust Thalamus Nuclei Parcellation Model from Noisy Annotations

Data-driven thalamic nuclei parcellation depends on high-quality manual annotations. However, the small size and low contrast changes among thalamic nuclei, yield annotations that are often incomplete, noisy, or ambiguously labelled. To train a robust thalamic nuclei parcellation model with noisy annotations, we propose a label propagation algorithm based on random walker to refine the annotations before model training. A two-step model was trained to generate first the whole thalamus and then the nuclei masks. We conducted experiments on a mild traumatic brain injury~(mTBI) dataset with noisy thalamic nuclei annotations. Our model outperforms current state-of-the-art thalamic nuclei parcellations by a clear margin. We believe our method can also facilitate the training of other parcellation models with noisy labels.

Analysis of Tongue Muscle Strain During Speech From Multimodal Magnetic Resonance Imaging

PURPOSE

Muscle groups within the tongue in healthy and diseased populations show different behaviors during speech. Visualizing and quantifying strain patterns of these muscle groups during tongue motion can provide insights into tongue motor control and adaptive behaviors of a patient.

METHOD

We present a pipeline to estimate the strain along the muscle fiber directions in the deforming tongue during speech production. A deep convolutional network estimates the crossing muscle fiber directions in the tongue using diffusion-weighted magnetic resonance imaging (MRI) data acquired at rest. A phase-based registration algorithm is used to estimate motion of the tongue muscles from tagged MRI acquired during speech. After transforming both muscle fiber directions and motion fields into a common atlas space, strain tensors are computed and projected onto the muscle fiber directions, forming so-called strains in the line of actions (SLAs) throughout the tongue. SLAs are then averaged over individual muscles that have been manually labeled in the atlas space using high-resolution T2-weighted MRI. Data were acquired, and this pipeline was run on a cohort of eight healthy controls and two glossectomy patients.

RESULTS

The crossing muscle fibers reconstructed by the deep network show orthogonal patterns. The strain analysis results demonstrate consistency of muscle behaviors among some healthy controls during speech production. The patients show irregular muscle patterns, and their tongue muscles tend to show more extension than the healthy controls.

CONCLUSIONS

The study showed visual evidence of correlation between two muscle groups during speech production. Patients tend to have different strain patterns compared to the controls. Analysis of variations in muscle strains can potentially help develop treatment strategies in oral diseases.

SUPPLEMENTAL MATERIAL

https://doi.org/10.23641/asha.21957011.

Investigation of probability maps in deep-learning-based brain ventricle parcellation

Normal Pressure Hydrocephalus (NPH) is a brain disorder associated with ventriculomegaly. Accurate segmentation of the ventricle system into its sub-compartments from magnetic resonance images (MRIs) could help evaluate NPH patients for surgical intervention. In this paper, we modify a 3D U-net utilizing probability maps to perform accurate ventricle parcellation, even with grossly enlarged ventricles and post-surgery shunt artifacts, from MRIs. Our method achieves a mean dice similarity coefficient (DSC) on whole ventricles for healthy controls of 0.864 ± 0.047 and 0.961 ± 0.024 for NPH patients. Furthermore, with the benefit of probability maps, the proposed method provides superior performance on MRI with grossly enlarged ventricles (mean DSC value of 0.965 ± 0.027) or post-surgery shunt artifacts (mean DSC value of 0.964 ± 0.031). Results indicate that our method provides a high robust parcellation tool on the ventricular systems which is comparable to other state-of-the-art methods.

Disentangled Representation Learning for OCTA Vessel Segmentation With Limited Training Data

Optical coherence tomography angiography (OCTA) is an imaging modality that can be used for analyzing retinal vasculature. Quantitative assessment of en face OCTA images requires accurate segmentation of the capillaries. Using deep learning approaches for this task faces two major challenges. First, acquiring sufficient manual delineations for training can take hundreds of hours. Second, OCTA images suffer from numerous contrast-related artifacts that are currently inherent to the modality and vary dramatically across scanners. We propose to solve both problems by learning a disentanglement of an anatomy component and a local contrast component from paired OCTA scans. With the contrast removed from the anatomy component, a deep learning model that takes the anatomy component as input can learn to segment vessels with a limited portion of the training images being manually labeled. Our method demonstrates state-of-the-art performance for OCTA vessel segmentation.

Development and validation of subject-specific 3D human head models based on a nonlinear visco-hyperelastic constitutive framework

Computational head models are promising tools for understanding and predicting traumatic brain injuries. Most available head models are developed using inputs (i.e. head geometry, material properties and boundary conditions) from experiments on cadavers or animals and employ hereditary integral-based constitutive models that assume linear viscoelasticity in part of the rate-sensitive material response. This leads to high uncertainty and poor accuracy in capturing the nonlinear brain tissue response. To resolve these issues, a framework for the development of subject-specific three-dimensional head models is proposed, in which all inputs are derived in vivo from the same living human subject: head geometry via magnetic resonance imaging (MRI), brain tissue properties via magnetic resonance elastography (MRE), and full-field strain-response of the brain under rapid head rotation via tagged MRI. A nonlinear, viscous dissipation-based visco-hyperelastic constitutive model is employed to capture brain tissue response. Head models are validated using quantitative metrics that compare spatial strain distribution, temporal strain evolution, and the magnitude of strain maxima, with the corresponding experimental observations from tagged MRI. Results show that our head models accurately capture the strain-response of the brain. Further, employment of the nonlinear visco-hyperelastic constitutive framework provides improvements in the prediction of peak strains and temporal strain evolution over hereditary integral-based models.

Bi-directional Synthesis of Pre- and Post-contrast MRI via Guided Feature Disentanglement

Magnetic resonance imaging (MRI) with gadolinium contrast is widely used for tissue enhancement and better identification of active lesions and tumors. Recent studies have shown that gadolinium deposition can accumulate in tissues including the brain, which raises safety concerns. Prior works have tried to synthesize post-contrast T1-weighted MRIs from pre-contrast MRIs to avoid the use of gadolinium. However, contrast and image representations are often entangled during the synthesis process, resulting in synthetic post-contrast MRIs with undesirable contrast enhancements. Moreover, the synthesis of pre-contrast MRIs from post-contrast MRIs which can be useful for volumetric analysis is rarely investigated in the literature. To tackle pre- and post- contrast MRI synthesis, we propose a BI-directional Contrast Enhancement Prediction and Synthesis (BICEPS) network that enables disentanglement of contrast and image representations via a bi-directional image-to-image translation(I2I)model. Our proposed model can perform both pre-to-post and post-to-pre contrast synthesis, and provides an interpretable synthesis process by predicting contrast enhancement maps from the learned contrast embedding. Extensive experiments on a multiple sclerosis dataset demonstrate the feasibility of applying our bidirectional synthesis and show that BICEPS outperforms current methods.

Data-driven Uncertainty Quantification in Computational Human Head Models

Computational models of the human head are promising tools for estimating the impact-induced response of the brain, and thus play an important role in the prediction of traumatic brain injury. The basic constituents of these models (i.e., model geometry, material properties, and boundary conditions) are often associated with significant uncertainty and variability. As a result, uncertainty quantification (UQ), which involves quantification of the effect of this uncertainty and variability on the simulated response, becomes critical to ensure reliability of model predictions. Modern biofidelic head model simulations are associated with very high computational cost and high-dimensional inputs and outputs, which limits the applicability of traditional UQ methods on these systems. In this study, a two-stage, data-driven manifold learning-based framework is proposed for UQ of computational head models. This framework is demonstrated on a 2D subject-specific head model, where the goal is to quantify uncertainty in the simulated strain fields (i.e., output), given variability in the material properties of different brain substructures (i.e., input). In the first stage, a data-driven method based on multi-dimensional Gaussian kernel-density estimation and diffusion maps is used to generate realizations of the input random vector directly from the available data. Computational simulations of a small number of realizations provide input-output pairs for training data-driven surrogate models in the second stage. The surrogate models employ nonlinear dimensionality reduction using Grassmannian diffusion maps, Gaussian process regression to create a low-cost mapping between the input random vector and the reduced solution space, and geometric harmonics models for mapping between the reduced space and the Grassmann manifold. It is demonstrated that the surrogate models provide highly accurate approximations of the computational model while significantly reducing the computational cost. Monte Carlo simulations of the surrogate models are used for uncertainty propagation. UQ of the strain fields highlights significant spatial variation in model uncertainty, and reveals key differences in uncertainty among commonly used strain-based brain injury predictor variables.

Evaluating the impact of MR image harmonization on thalamus deep network segmentation

Medical image segmentation is one of the core tasks of medical image analysis. Automatic segmentation of brain magnetic resonance images (MRIs) can be used to visualize and track changes of the brain's anatomical structures that may occur due to normal aging or disease. Machine learning techniques are widely used in automatic structure segmentation. However, the contrast variation between the training and testing data makes it difficult for segmentation algorithms to generate consistent results. To address this problem, an image-to-image translation technique called MR image harmonization can be used to match the contrast between different data sets. It is important for the harmonization to transform image intensity while maintaining the underlying anatomy. In this paper, we present a 3D U-Net algorithm to segment the thalamus from multiple MR image modalities and investigate the impact of harmonization on the segmentation algorithm. Manual delineations of thalamic nuclei on two data sets are available. However, we aim to analyze the thalamus in another large data set where ground truth labels are lacking. We trained two segmentation networks, one with unharmonized images and the other with harmonized images, on one data set with manual labels, and compared their performances on the other data set with manual labels. These two data groups were diagnosed with two brain disorders and were acquired with similar imaging protocols. The harmonization target is the large data set without manual labels, which also has a different imaging protocol. The networks trained on unharmonized and harmonized data showed no significant difference when evaluating on the other data set; demonstrating that image harmonization can maintain the anatomy and does not affect the segmentation task. The two networks were evaluated on the harmonization target data set and the network trained on harmonized data showed significant improvement over the network trained on unharmonized data. Therefore, the network trained on harmonized data provides the potential to process large amounts of data from other sites, even in the absence of site-specific training data.

Cranial Meninges Reconstruction Based on Convolutional Networks and Deformable Models: Applications to Longitudinal Study of Normal Aging

The cranial meninges are membranes enveloping the brain. The space between these membranes contains mainly cerebrospinal fluid. It is of interest to study how the volumes of this space change with respect to normal aging. In this work, we propose to combine convolutional neural networks (CNNs) with nested topology-preserving geometric deformable models (NTGDMs) to reconstruct meningeal surfaces from magnetic resonance (MR) images. We first use CNNs to predict implicit representations of these surfaces then refine them with NTGDMs to achieve sub-voxel accuracy while maintaining spherical topology and the correct anatomical ordering. MR contrast harmonization is used to match the contrasts between training and testing images. We applied our algorithm to a subset of healthy subjects from the Baltimore Longitudinal Study of Aging for demonstration purposes and conducted longitudinal statistical analysis of the intracranial volume (ICV) and subarachnoid space (SAS) volume. We found a statistically significant decrease in the ICV and an increase in the SAS volume with respect to normal aging.

Unsupervised MR harmonization by learning disentangled representations using information bottleneck theory

In magnetic resonance (MR) imaging, a lack of standardization in acquisition often causes pulse sequence-based contrast variations in MR images from site to site, which impedes consistent measurements in automatic analyses. In this paper, we propose an unsupervised MR image harmonization approach, CALAMITI (Contrast Anatomy Learning and Analysis for MR Intensity Translation and Integration), which aims to alleviate contrast variations in multi-site MR imaging. Designed using information bottleneck theory, CALAMITI learns a globally disentangled latent space containing both anatomical and contrast information, which permits harmonization. In contrast to supervised harmonization methods, our approach does not need a sample population to be imaged across sites. Unlike traditional unsupervised harmonization approaches which often suffer from geometry shifts, CALAMITI better preserves anatomy by design. The proposed method is also able to adapt to a new testing site with a straightforward fine-tuning process. Experiments on MR images acquired from ten sites show that CALAMITI achieves superior performance compared with other harmonization approaches.

DUAL-CYCLE CONSTRAINED BIJECTIVE VAE-GAN FOR TAGGED-TO-CINE MAGNETIC RESONANCE IMAGE SYNTHESIS

Tagged magnetic resonance imaging (MRI) is a widely used imaging technique for measuring tissue deformation in moving organs. Due to tagged MRI's intrinsic low anatomical resolution, another matching set of cine MRI with higher resolution is sometimes acquired in the same scanning session to facilitate tissue segmentation, thus adding extra time and cost. To mitigate this, in this work, we propose a novel dual-cycle constrained bijective VAE-GAN approach to carry out tagged-to-cine MR image synthesis. Our method is based on a variational autoencoder backbone with cycle reconstruction constrained adversarial training to yield accurate and realistic cine MR images given tagged MR images. Our framework has been trained, validated, and tested using 1,768, 416, and 1,560 subject-independent paired slices of tagged and cine MRI from twenty healthy subjects, respectively, demonstrating superior performance over the comparison methods. Our method can potentially be used to reduce the extra acquisition time and cost, while maintaining the same workflow for further motion analyses.

MR Imaging of Human Brain Mechanics In Vivo: New Measurements to Facilitate the Development of Computational Models of Brain Injury

Computational models of the brain and its biomechanical response to skull accelerations are important tools for understanding and predicting traumatic brain injuries (TBIs). However, most models have been developed using experimental data collected on animal models and cadaveric specimens, both of which differ from the living human brain. Here we describe efforts to noninvasively measure the biomechanical response of the human brain with MRI-at non-injurious strain levels-and generate data that can be used to develop, calibrate, and evaluate computational brain biomechanics models. Specifically, this paper reports on a project supported by the National Institute of Neurological Disorders and Stroke to comprehensively image brain anatomy and geometry, mechanical properties, and brain deformations that arise from impulsive and harmonic skull loadings. The outcome of this work will be a publicly available dataset ( http://www.nitrc.org/projects/bbir ) that includes measurements on both males and females across an age range from adolescence to older adulthood. This article describes the rationale and approach for this study, the data available, and how these data may be used to develop new computational models and augment existing approaches; it will serve as a reference to researchers interested in using these data.

Accurate Estimation of Total Intracranial Volume in MRI using a Multi-tasked Image-to-Image Translation Network

Total intracranial volume (TIV) is the volume enclosed inside the cranium, inclusive of the meninges and the brain. TIV is extensively used to correct variations in inter-subject head size for the evaluation of neurodegenerative diseases. In this work, we present an automatic method to generate a TIV mask from MR images while synthesizing a CT image to be used in subsequent analysis. In addition, we propose an alternative way to obtain ground truth TIV masks using a semi-manual approach, which results in significant time savings. We train a conditional generative adversarial network (cGAN) using 2D MR slices to realize our tasks. The quantitative evaluation showed that the model was able to synthesize CT and generate TIV masks that closely approximate the reference images. This study also provides a comparison of the described method against skull stripping tools that output a mask enclosing the cranial volume, using MRI scan. In particular, highlighting the deficiencies in using such tools to approximate the volume using MRI scan.

Joint Image and Label Self-Super-Resolution

We propose a method to jointly super-resolve an anisotropic image volume along with its corresponding voxel labels without external training data. Our method is inspired by internally trained superresolution, or self-super-resolution (SSR) techniques that target anisotropic, low-resolution (LR) magnetic resonance (MR) images. While resulting images from such methods are quite useful, their corresponding LR labels-derived from either automatic algorithms or human raters-are no longer in correspondence with the super-resolved volume. To address this, we develop an SSR deep network that takes both an anisotropic LR MR image and its corresponding LR labels as input and produces both a super-resolved MR image and its super-resolved labels as output. We evaluated our method with 50 T ₁-weighted brain MR images 4× down-sampled with 10 automatically generated labels. In comparison to other methods, our method had superior Dice across all labels and competitive metrics on the MR image. Our approach is the first reported method for SSR of paired anisotropic image and label volumes.

Autoencoder based self-supervised test-time adaptation for medical image analysis

Deep neural networks have been successfully applied to medical image analysis tasks like segmentation and synthesis. However, even if a network is trained on a large dataset from the source domain, its performance on unseen test domains is not guaranteed. The performance drop on data obtained differently from the network's training data is a major problem (known as domain shift) in deploying deep learning in clinical practice. Existing work focuses on retraining the model with data from the test domain, or harmonizing the test domain's data to the network training data. A common practice is to distribute a carefully-trained model to multiple users (e.g., clinical centers), and then each user uses the model to process their own data, which may have a domain shift (e.g., varying imaging parameters and machines). However, the lack of availability of the source training data and the cost of training a new model often prevents the use of known methods to solve user-specific domain shifts. Here, we ask whether we can design a model that, once distributed to users, can quickly adapt itself to each new site without expensive retraining or access to the source training data? In this paper, we propose a model that can adapt based on a single test subject during inference. The model consists of three parts, which are all neural networks: a task model (T) which performs the image analysis task like segmentation; a set of autoencoders (AEs); and a set of adaptors (As). The task model and autoencoders are trained on the source dataset and can be computationally expensive. In the deployment stage, the adaptors are trained to transform the test image and its features to minimize the domain shift as measured by the autoencoders' reconstruction loss. Only the adaptors are optimized during the testing stage with a single test subject thus is computationally efficient. The method was validated on both retinal optical coherence tomography (OCT) image segmentation and magnetic resonance imaging (MRI) T1-weighted to T2-weighted image synthesis. Our method, with its short optimization time for the adaptors (10 iterations on a single test subject) and its additional required disk space for the autoencoders (around 15 MB), can achieve significant performance improvement. Our code is publicly available at: https://github.com/YufanHe/self-domain-adapted-network.

Structured layer surface segmentation for retina OCT using fully convolutional regression networks

Optical coherence tomography (OCT) is a noninvasive imaging modality with micrometer resolution which has been widely used for scanning the retina. Retinal layers are important biomarkers for many diseases. Accurate automated algorithms for segmenting smooth continuous layer surfaces with correct hierarchy (topology) are important for automated retinal thickness and surface shape analysis. State-of-the-art methods typically use a two step process. Firstly, a trained classifier is used to label each pixel into either background and layers or boundaries and non-boundaries. Secondly, the desired smooth surfaces with the correct topology are extracted by graph methods (e.g., graph cut). Data driven methods like deep networks have shown great ability for the pixel classification step, but to date have not been able to extract structured smooth continuous surfaces with topological constraints in the second step. In this paper, we combine these two steps into a unified deep learning framework by directly modeling the distribution of the surface positions. Smooth, continuous, and topologically correct surfaces are obtained in a single feed forward operation. The proposed method was evaluated on two publicly available data sets of healthy controls and subjects with either multiple sclerosis or diabetic macular edema, and is shown to achieve state-of-the art performance with sub-pixel accuracy.

Integrating material properties from magnetic resonance elastography into subject-specific computational models for the human brain

Advances in brain imaging and computational methods have facilitated the creation of subject-specific computational brain models that aid researchers in investigating brain trauma using simulated impacts. The emergence of magnetic resonance elastography (MRE) as a non-invasive mechanical neuroimaging tool has enabled in vivo estimation of material properties at low-strain, harmonic loading. An open question in the field has been how this data can be integrated into computational models. The goals of this study were to use a novel MRI dataset acquired in human volunteers to generate models with subject-specific anatomy and material properties, and then to compare simulated brain deformations to subject-specific brain deformation data under non-injurious loading. Models of five subjects were simulated with linear viscoelastic (LVE) material properties estimated directly from MRE data. Model predictions were compared to experimental brain deformation acquired in the same subjects using tagged MRI. Outcomes from the models matched the spatial distribution and magnitude of the measured peak strain components as well as the 95^th percentile in-plane peak strains within 0.005 mm/mm and maximum principal strain within 0.012 mm/mm. Sensitivity to material heterogeneity was also investigated. Simulated brain deformations from a model with homogenous brain properties and a model with brain properties discretized with up to ten regions were very similar (a mean absolute difference less than 0.0015 mm/mm in peak strains). Incorporating material properties directly from MRE into a biofidelic subject-specific model is an important step toward future investigations of higher-order model features and simulations under more severe loading conditions.

Unsupervised MR-to-CT Synthesis Using Structure-Constrained CycleGAN

Synthesizing a CT image from an available MR image has recently emerged as a key goal in radiotherapy treatment planning for cancer patients. CycleGANs have achieved promising results on unsupervised MR-to-CT image synthesis; however, because they have no direct constraints between input and synthetic images, cycleGANs do not guarantee structural consistency between these two images. This means that anatomical geometry can be shifted in the synthetic CT images, clearly a highly undesirable outcome in the given application. In this paper, we propose a structure-constrained cycleGAN for unsupervised MR-to-CT synthesis by defining an extra structure-consistency loss based on the modality independent neighborhood descriptor. We also utilize a spectral normalization technique to stabilize the training process and a self-attention module to model the long-range spatial dependencies in the synthetic images. Results on unpaired brain and abdomen MR-to-CT image synthesis show that our method produces better synthetic CT images in both accuracy and visual quality as compared to other unsupervised synthesis methods. We also show that an approximate affine pre-registration for unpaired training data can improve synthesis results.

Longitudinal analysis of regional cerebellum volumes during normal aging

Some cross-sectional studies suggest reduced cerebellar volumes with aging, but there have been few longitudinal studies of age changes in cerebellar subregions in cognitively healthy older adults. In this work, 2,023 magnetic resonance (MR) images of 822 cognitively normal participants from the Baltimore Longitudinal Study of Aging (BLSA) were analyzed. Participants ranged in age from 50 to 95 years (mean 70.7 years) at the baseline assessment. Follow-up intervals were 1-9 years (mean 3.7 years) for participants with two or more visits. We used a recently developed cerebellum parcellation algorithm based on convolutional neural networks to divide the cerebellum into 28 subregions. Linear mixed effects models were applied to the volume of each cerebellar subregion to investigate cross-sectional and longitudinal age effects, as well as effects of sex and their interactions, after adjusting for intracranial volume. Our findings suggest spatially varying atrophy patterns across the cerebellum with respect to age and sex both cross-sectionally and longitudinally.

Automatic cerebellum anatomical parcellation using U-Net with locally constrained optimization

The cerebellum plays a central role in sensory input, voluntary motor action, and many neuropsychological functions and is involved in many brain diseases and neurological disorders. Cerebellar parcellation from magnetic resonance images provides a way to study regional cerebellar atrophy and also provides an anatomical map for functional imaging. In a recent comparison, a multi-atlas approach proved to be superior to other parcellation methods including some based on convolutional neural networks (CNNs) which have a considerable speed advantage. In this work, we developed an alternative CNN design for cerebellar parcellation, yielding a method that achieves the leading performance to date. The proposed method was evaluated on multiple data sets to show its broad applicability, and a Singularity container has been made publicly available.

Evaluating White Matter Lesion Segmentations with Refined Sørensen-Dice Analysis

The Sørensen-Dice index (SDI) is a widely used measure for evaluating medical image segmentation algorithms. It offers a standardized measure of segmentation accuracy which has proven useful. However, it offers diminishing insight when the number of objects is unknown, such as in white matter lesion segmentation of multiple sclerosis (MS) patients. We present a refinement for finer grained parsing of SDI results in situations where the number of objects is unknown. We explore these ideas with two case studies showing what can be learned from our two presented studies. Our first study explores an inter-rater comparison, showing that smaller lesions cannot be reliably identified. In our second case study, we demonstrate fusing multiple MS lesion segmentation algorithms based on the insights into the algorithms provided by our analysis to generate a segmentation that exhibits improved performance. This work demonstrates the wealth of information that can be learned from refined analysis of medical image segmentations.

Applications of a deep learning method for anti-aliasing and super-resolution in MRI

Magnetic resonance (MR) images with both high resolutions and high signal-to-noise ratios (SNRs) are desired in many clinical and research applications. However, acquiring such images takes a long time, which is both costly and susceptible to motion artifacts. Acquiring MR images with good in-plane resolution and poor through-plane resolution is a common strategy that saves imaging time, preserves SNR, and provides one viewpoint with good resolution in two directions. Unfortunately, this strategy also creates orthogonal viewpoints that have poor resolution in one direction and, for 2D MR acquisition protocols, also creates aliasing artifacts. A deep learning approach called SMORE that carries out both anti-aliasing and super-resolution on these types of acquisitions using no external atlas or exemplars has been previously reported but not extensively validated. This paper reviews the SMORE algorithm and then demonstrates its performance in four applications with the goal to demonstrate its potential for use in both research and clinical scenarios. It is first shown to improve the visualization of brain white matter lesions in FLAIR images acquired from multiple sclerosis patients. Then it is shown to improve the visualization of scarring in cardiac left ventricular remodeling after myocardial infarction. Third, its performance on multi-view images of the tongue is demonstrated and finally it is shown to improve performance in parcellation of the brain ventricular system. Both visual and selected quantitative metrics of resolution enhancement are demonstrated.

DeepHarmony: A deep learning approach to contrast harmonization across scanner changes

Magnetic resonance imaging (MRI) is a flexible medical imaging modality that often lacks reproducibility between protocols and scanners. It has been shown that even when care is taken to standardize acquisitions, any changes in hardware, software, or protocol design can lead to differences in quantitative results. This greatly impacts the quantitative utility of MRI in multi-site or long-term studies, where consistency is often valued over image quality. We propose a method of contrast harmonization, called DeepHarmony, which uses a U-Net-based deep learning architecture to produce images with consistent contrast. To provide training data, a small overlap cohort (n = 8) was scanned using two different protocols. Images harmonized with DeepHarmony showed significant improvement in consistency of volume quantification between scanning protocols. A longitudinal MRI dataset of patients with multiple sclerosis was also used to evaluate the effect of a protocol change on atrophy calculations in a clinical research setting. The results show that atrophy calculations were substantially and significantly affected by protocol change, whereas such changes have a less significant effect and substantially reduced overall difference when using DeepHarmony. This establishes that DeepHarmony can be used with an overlap cohort to reduce inconsistencies in segmentation caused by changes in scanner protocol, allowing for modernization of hardware and protocol design in long-term studies without invalidating previously acquired data.

Fully Convolutional Boundary Regression for Retina OCT Segmentation

A major goal of analyzing retinal optical coherence tomography (OCT) images is retinal layer segmentation. Accurate automated algorithms for segmenting smooth continuous layer surfaces, with correct hierarchy (topology) are desired for monitoring disease progression. State-of-the-art methods use a trained classifier to label each pixel into background, layer, or surface pixels. The final step of extracting the desired smooth surfaces with correct topology are mostly performed by graph methods (e.g. shortest path, graph cut). However, manually building a graph with varying constraints by retinal region and pathology and solving the minimization with specialized algorithms will degrade the flexibility and time efficiency of the whole framework. In this paper, we directly model the distribution of surface positions using a deep network with a fully differentiable soft argmax to obtain smooth, continuous surfaces in a single feed forward operation. A special topology module is used in the deep network both in the training and testing stages to guarantee the surface topology. An extra deep network output branch is also used for predicting lesion and layers in a pixel-wise labeling scheme. The proposed method was evaluated on two publicly available data sets of healthy controls, subjects with multiple sclerosis, and diabetic macular edema; it achieves state-of-the art sub-pixel results.

Hierarchical Parcellation of the Cerebellum

Parcellation of the cerebellum in an MR image has been used to study regional associations with both motion and cognitive functions. Despite the fact that the division of the cerebellum is defined hierarchically-i.e., the cerebellum can be divided into lobes and the lobes can be further divided into lobules-previous automatic methods to parcellate the cerebellum do not utilize this information. In this work, we propose a method based on convolutional neural networks (CNNs) to explicitly incorporate the hierarchical organization of the cerebellum. The network is constructed in a tree structure with each node representing a cerebellar region and having child nodes that further subdivide the region into finer substructures. Thus, our CNN is aware of the hierarchical organization of the cerebellum. Furthermore, by selecting tree nodes to represent the hierarchical properties of a given training sample, our network can be trained with heterogeneous training data that are labeled to different hierarchical depths. The proposed method was compared with a state-of-the-art cerebellum parcellation network. Our approach shows promising results as a first parcellation method to take the cerebellar hierarchical organization into consideration.

Deep learning based topology guaranteed surface and MME segmentation of multiple sclerosis subjects from retinal OCT

Optical coherence tomography (OCT) is a noninvasive imaging modality that can be used to obtain depth images of the retina. Patients with multiple sclerosis (MS) have thinning retinal nerve fiber and ganglion cell layers, and approximately 5% of MS patients will develop microcystic macular edema (MME) within the retina. Segmentation of both the retinal layers and MME can provide important information to help monitor MS progression. Graph-based segmentation with machine learning preprocessing is the leading method for retinal layer segmentation, providing accurate surface delineations with the correct topological ordering. However, graph methods are time-consuming and they do not optimally incorporate joint MME segmentation. This paper presents a deep network that extracts continuous, smooth, and topology-guaranteed surfaces and MMEs. The network learns shape priors automatically during training rather than being hard-coded as in graph methods. In this new approach, retinal surfaces and MMEs are segmented together with two cascaded deep networks in a single feed forward propagation. The proposed framework obtains retinal surfaces (separating the layers) with sub-pixel surface accuracy comparable to the best existing graph methods and MMEs with better accuracy than the state-of-the-art method. The full segmentation operation takes only ten seconds for a 3D volume.

Brain ventricle parcellation using a deep neural network: Application to patients with ventriculomegaly

Numerous brain disorders are associated with ventriculomegaly, including both neuro-degenerative diseases and cerebrospinal fluid disorders. Detailed evaluation of the ventricular system is important for these conditions to help understand the pathogenesis of ventricular enlargement and elucidate novel patterns of ventriculomegaly that can be associated with different diseases. One such disease is normal pressure hydrocephalus (NPH), a chronic form of hydrocephalus in older adults that causes dementia. Automatic parcellation of the ventricular system into its sub-compartments in patients with ventriculomegaly is quite challenging due to the large variation of the ventricle shape and size. Conventional brain labeling methods are time-consuming and often fail to identify the boundaries of the enlarged ventricles. We propose a modified 3D U-Net method to perform accurate ventricular parcellation, even with grossly enlarged ventricles, from magnetic resonance images (MRIs). We validated our method on a data set of healthy controls as well as a cohort of 95 patients with NPH with mild to severe ventriculomegaly and compared with several state-of-the-art segmentation methods. On the healthy data set, the proposed network achieved mean Dice similarity coefficient (DSC) of 0.895 ± 0.03 for the ventricular system. On the NPH data set, we achieved mean DSC of 0.973 ± 0.02, which is significantly (p < 0.005) higher than four state-of-the-art segmentation methods we compared with. Furthermore, the typical processing time on CPU-base implementation of the proposed method is 2 min, which is much lower than the several hours required by the other methods. Results indicate that our method provides: 1) highly robust parcellation of the ventricular system that is comparable in accuracy to state-of-the-art methods on healthy controls; 2) greater robustness and significantly more accurate results on cases of ventricular enlargement; and 3) a tool that enables computation of novel imaging biomarkers for dilated ventricular spaces that characterize the ventricular system.

Evaluating the Impact of Intensity Normalization on MR Image Synthesis

Image synthesis learns a transformation from the intensity features of an input image to yield a different tissue contrast of the output image. This process has been shown to have application in many medical image analysis tasks including imputation, registration, and segmentation. To carry out synthesis, the intensities of the input images are typically scaled-i.e., normalized-both in training to learn the transformation and in testing when applying the transformation, but it is not presently known what type of input scaling is optimal. In this paper, we consider seven different intensity normalization algorithms and three different synthesis methods to evaluate the impact of normalization. Our experiments demonstrate that intensity normalization as a preprocessing step improves the synthesis results across all investigated synthesis algorithms. Furthermore, we show evidence that suggests intensity normalization is vital for successful deep learning-based MR image synthesis.

Layer boundary evolution method for macular OCT layer segmentation

Optical coherence tomography (OCT) is used to produce high resolution depth images of the retina and is now the standard of care for in-vivo ophthalmological assessment. It is also increasingly being used for evaluation of neurological disorders such as multiple sclerosis (MS). Automatic segmentation methods identify the retinal layers of the macular cube providing consistent results without intra- and inter-rater variation and is faster than manual segmentation. In this paper, we propose a fast multi-layer macular OCT segmentation method based on a fast level set method. Our framework uses contours in an optimized approach specifically for OCT layer segmentation over the whole macular cube. Our algorithm takes boundary probability maps from a trained random forest and iteratively refines the prediction to subvoxel precision. Evaluation on both healthy and multiple sclerosis subjects shows that our method is statistically better than a state-of-the-art graph-based method.

Retinal layer parcellation of optical coherence tomography images: Data resource for multiple sclerosis and healthy controls

This paper presents optical coherence tomography (OCT) images of the human retina and manual delineations of eight retinal layers. The data includes 35 human retina scans acquired on a Spectralis OCT system (Heidelberg Engineering, Heidelberg, Germany), 14 of which are healthy controls (HC) and 21 have a diagnosis of multiple sclerosis (MS). The provided data includes manually delineation of eight retina layers, which were independently reviewed and edited. The data presented in this article was used to validate automatic segmentation algorithms (Lang et al., 2013).

Cerebellum Parcellation with Convolutional Neural Networks

To better understand cerebellum-related diseases and functional mapping of the cerebellum, quantitative measurements of cerebellar regions in magnetic resonance (MR) images have been studied in both clinical and neurological studies. Such studies have revealed that different spinocerebellar ataxia (SCA) subtypes have different patterns of cerebellar atrophy and that atrophy of different cerebellar regions is correlated with specific functional losses. Previous methods to automatically parcellate the cerebellum-that is, to identify its sub-regions-have been largely based on multi-atlas segmentation. Recently, deep convolutional neural network (CNN) algorithms have been shown to have high speed and accuracy in cerebral sub-cortical structure segmentation from MR images. In this work, two three-dimensional CNNs were used to parcellate the cerebellum into 28 regions. First, a locating network was used to predict a bounding box around the cerebellum. Second, a parcellating network was used to parcellate the cerebellum using the entire region within the bounding box. A leave-one-out cross validation of fifteen manually delineated images was performed. Compared with a previously reported state-of-the-art algorithm, the proposed algorithm shows superior Dice coefficients. The proposed algorithm was further applied to three MR images of a healthy subject and subjects with SCA6 and SCA8, respectively. A Singularity container of this algorithm is publicly available.

Why rankings of biomedical image analysis competitions should be interpreted with care

International challenges have become the standard for validation of biomedical image analysis methods. Given their scientific impact, it is surprising that a critical analysis of common practices related to the organization of challenges has not yet been performed. In this paper, we present a comprehensive analysis of biomedical image analysis challenges conducted up to now. We demonstrate the importance of challenges and show that the lack of quality control has critical consequences. First, reproducibility and interpretation of the results is often hampered as only a fraction of relevant information is typically provided. Second, the rank of an algorithm is generally not robust to a number of variables such as the test data used for validation, the ranking scheme applied and the observers that make the reference annotations. To overcome these problems, we recommend best practice guidelines and define open research questions to be addressed in the future.

Comparing fully automated state-of-the-art cerebellum parcellation from magnetic resonance images

The human cerebellum plays an essential role in motor control, is involved in cognitive function (i.e., attention, working memory, and language), and helps to regulate emotional responses. Quantitative in-vivo assessment of the cerebellum is important in the study of several neurological diseases including cerebellar ataxia, autism, and schizophrenia. Different structural subdivisions of the cerebellum have been shown to correlate with differing pathologies. To further understand these pathologies, it is helpful to automatically parcellate the cerebellum at the highest fidelity possible. In this paper, we coordinated with colleagues around the world to evaluate automated cerebellum parcellation algorithms on two clinical cohorts showing that the cerebellum can be parcellated to a high accuracy by newer methods. We characterize these various methods at four hierarchical levels: coarse (i.e., whole cerebellum and gross structures), lobe, subdivisions of the vermis, and the lobules. Due to the number of labels, the hierarchy of labels, the number of algorithms, and the two cohorts, we have restricted our analyses to the Dice measure of overlap. Under these conditions, machine learning based methods provide a collection of strategies that are efficient and deliver parcellations of a high standard across both cohorts, surpassing previous work in the area. In conjunction with the rank-sum computation, we identified an overall winning method.

Shortcomings of Ventricle Segmentation Using Deep Convolutional Networks

Normal Pressure Hydrocephalus (NPH) is a brain disorder that can present with ventriculomegaly and dementia-like symptoms, which often can be reversed through surgery. Having accurate segmentation of the ventricular system into its sub-compartments from magnetic resonance images (MRI) would be beneficial to better characterize the condition of NPH patients. Previous segmentation algorithms need long processing time and often fail to accurately segment severely enlarged ventricles in NPH patients. Recently, deep convolutional neural network (CNN) methods have been reported to have fast and accurate performance on medical image segmentation tasks. In this paper, we present a 3D U-net CNN-based network to segment the ventricular system in MRI. We trained three networks on different data sets and compared their performances. The networks trained on healthy controls (HC) failed in patients with NPH pathology, even in patients with normal appearing ventricles. The network trained on images from HC and NPH patients provided superior performance against state-of-the-art methods when evaluated on images from both data sets.

A Deep Learning Based Anti-aliasing Self Super-resolution Algorithm for MRI

High resolution magnetic resonance (MR) images are desired in many clinical applications, yet acquiring such data with an adequate signal-to-noise ratio requires a long time, making them costly and susceptible to motion artifacts. A common way to partly achieve this goal is to acquire MR images with good in-plane resolution and poor through-plane resolution (i.e., large slice thickness). For such 2D imaging protocols, aliasing is also introduced in the through-plane direction, and these high-frequency artifacts cannot be removed by conventional interpolation. Super-resolution (SR) algorithms which can reduce aliasing artifacts and improve spatial resolution have previously been reported. State-of-the-art SR methods are mostly learning-based and require external training data consisting of paired low resolution (LR) and high resolution (HR) MR images. However, due to scanner limitations, such training data are often unavailable. This paper presents an anti-aliasing (AA) and self super-resolution (SSR) algorithm that needs no external training data. It takes advantage of the fact that the in-plane slices of those MR images contain high frequency information. Our algorithm consists of three steps: 1) We build a self AA (SAA) deep network followed by 2) an SSR deep network, both of which can be applied along different orientations within the original images, and 3) recombine the multiple orientations output from Steps 1 and 2 using Fourier burst accumulation. We perform our SAA+SSR algorithm on a diverse collection of MR data without modification or preprocessing other than N4 inhomogeneity correction, and demonstrate significant improvement compared to competing SSR methods.

Preoperative Mapping of the Supplementary Motor Area in Patients with Brain Tumor Using Resting-State fMRI with Seed-Based Analysis

BACKGROUND AND PURPOSE

The supplementary motor area can be a critical region in the preoperative planning of patients undergoing brain tumor resection because it plays a role in both language and motor function. While primary motor regions have been successfully identified using resting-state fMRI, there is variability in the literature regarding the identification of the supplementary motor area for preoperative planning. The purpose of our study was to compare resting-state fMRI to task-based fMRI for localization of the supplementary motor area in a large cohort of patients with brain tumors presenting for preoperative brain mapping.

MATERIALS AND METHODS

Sixty-six patients with brain tumors were evaluated with resting-state fMRI using seed-based analysis of hand and orofacial motor regions. Rates of supplementary motor area localization were compared with those in healthy controls and with localization results by task-based fMRI.

RESULTS

Localization of the supplementary motor area using hand motor seed regions was more effective than seeding using orofacial motor regions for both patients with brain tumor (95.5% versus 34.8%, P < .001) and controls (95.2% versus 45.2%, P < .001). Bilateral hand motor seeding was superior to unilateral hand motor seeding in patients with brain tumor for either side (95.5% versus 75.8%/75.8% for right/left, P < .001). No difference was found in the ability to identify the supplementary motor area between patients with brain tumors and controls.

CONCLUSIONS

In addition to task-based fMRI, seed-based analysis of resting-state fMRI represents an equally effective method for supplementary motor area localization in patients with brain tumors, with the best results obtained with bilateral hand motor region seeding.

Multi-layer Fast Level Set Segmentation for Macular OCT

Segmenting optical coherence tomography (OCT) images of the retina is important in the diagnosis, staging, and tracking of ophthalmological diseases. Whereas automatic segmentation methods are typically much faster than manual segmentation, they may still take several minutes to segment a three-dimensional macular scan, and this can be prohibitive for routine clinical application. In this paper, we propose a fast, multi-layer macular OCT segmentation method based on a fast level set method. In our framework, the boundary evolution operations are computationally fast, are specific to each boundary between retinal layers, guarantee proper layer ordering, and avoid level set computation during evolution. Subvoxel resolution is achieved by reconstructing the level set functions after convergence. Experiments demonstrate that our method reduces the computation expense by 90% compared to graph-based methods and produces comparable accuracy to both graph-based and level set retinal OCT segmentation methods.

Inverse Biomechanical Modeling of the Tongue via Machine Learning and Synthetic Training Data

The tongue's deformation during speech can be measured using tagged magnetic resonance imaging, but there is no current method to directly measure the pattern of muscles that activate to produce a given motion. In this paper, the activation pattern of the tongue's muscles is estimated by solving an inverse problem using a random forest. Examples describing different activation patterns and the resulting deformations are generated using a finite-element model of the tongue. These examples form training data for a random forest comprising 30 decision trees to estimate contractions in 262 contractile elements. The method was evaluated on data from tagged magnetic resonance data from actual speech and on simulated data mimicking flaps that might have resulted from glossectomy surgery. The estimation accuracy was modest (5.6% error), but it surpassed a semi-manual approach (8.1% error). The results suggest that a machine learning approach to contraction pattern estimation in the tongue is feasible, even in the presence of flaps.