Patch-based segmentation using expert priors: application to hippocampus and ventricle segmentation

VentRa: distinguishing frontotemporal dementia from psychiatric disorders

The volume of the lateral ventricles is a reliable and sensitive indicator of brain atrophy and disease progression in behavioural variant frontotemporal dementia. In this study, we validate our previously developed automated tool using ventricular features (known as VentRa) for the classification of behavioural variant frontotemporal dementia versus a mixed cohort of neurodegenerative, vascular and psychiatric disorders from a clinically representative independent dataset. Lateral ventricles were segmented for 1110 subjects-14 behavioural variant frontotemporal dementia, 30 other frontotemporal dementia, 70 Lewy body disease, 898 Alzheimer's disease, 62 vascular brain injury and 36 primary psychiatric disorder from the publicly accessible National Alzheimer's Coordinating Center dataset to assess the performance of VentRa. Using ventricular features to discriminate behavioural variant frontotemporal dementia subjects from primary psychiatric disorders, VentRa achieved an accuracy rate of 84%, a sensitivity rate of 71% and a specificity rate of 89%. VentRa was able to identify behavioural variant frontotemporal dementia from a mixed age-matched cohort (i.e. other frontotemporal dementia, Lewy body disease, Alzheimer's disease, vascular brain injury and primary psychiatric disorders) and to correctly classify other disorders as 'not compatible with behavioral variant frontotemporal dementia' with a specificity rate of 83%. The specificity rates against each of the other individual cohorts were 80% for other frontotemporal dementia, 83% for Lewy body disease, 83% for Alzheimer's disease, 84% for vascular brain injury and 89% for primary psychiatric disorders. VentRa is a robust and generalizable tool with potential usefulness for improving the diagnostic certainty of behavioural variant frontotemporal dementia, particularly for the differential diagnosis with primary psychiatric disorders.

Sex differences in risk factors, burden, and outcomes of cerebrovascular disease in Alzheimer's disease populations

BACKGROUND

White matter hyperintensities (WMHs) are associated with cognitive decline and progression to mild cognitive impairment (MCI) and dementia. It remains unclear if sex differences influence WMH progression or the relationship between WMH and cognition.

METHODS

Linear mixed models examined the relationship between risk factors, WMHs, and cognition in males and females.

RESULTS

Males exhibited increased WMH progression in occipital, but lower progression in frontal, total, and deep than females. For males, history of hypertension was the strongest contributor, while in females, the vascular composite was the strongest contributor to WMH burden. WMH burden was more strongly associated with decreases in global cognition, executive functioning, memory, and functional activities in females than males.

DISCUSSION

Controlling vascular risk factors may reduce WMH in both males and females. For males, targeting hypertension may be most important to reduce WMHs. The results have implications for therapies/interventions targeting cerebrovascular pathology and subsequent cognitive decline.

HIGHLIGHTS

Hypertension is the main vascular risk factor associated with WMH in males A combination of vascular risk factors contributes to WMH burden in females Only small WMH burden differences were observed between sexes Females' cognition was more negatively impacted by WMH burden than males Females with WMHs may have less resilience to future pathology.

Analyzing the cortical fine structure as revealed by ex-vivo anatomical MRI

The human cortex has a rich fiber structure as revealed by myelin-staining of histological slices. Myelin also contributes to the image contrast in Magnetic Resonance Imaging (MRI). Recent advances in Magnetic Resonance (MR) scanner and imaging technology allowed the acquisition of an ex-vivo data set at an isotropic resolution of 100 µm. This study focused on a computational analysis of this data set with the aim of bridging between histological knowledge and MRI-based results. This work highlights: (1) the design and implementation of a processing chain that extracts intracortical features from a high-resolution MR image; (2) a demonstration of the correspondence between MRI-based cortical intensity profiles and the myelo-architectonic layering of the cortex; (3) the characterization and classification of four basic myelo-architectonic profile types; (4) the distinction of cortical regions based on myelo-architectonic features; and (5) the segmentation of cortical modules in the entorhinal cortex.

Spiking neural networks fine-tuning for brain image segmentation

Introduction

The field of machine learning has undergone a significant transformation with the progress of deep artificial neural networks (ANNs) and the growing accessibility of annotated data. ANNs usually require substantial power and memory usage to achieve optimal performance. Spiking neural networks (SNNs) have recently emerged as a low-power alternative to ANNs due to their sparsity nature. Despite their energy efficiency, SNNs are generally more difficult to be trained than ANNs.

Methods

In this study, we propose a novel three-stage SNN training scheme designed specifically for segmenting human hippocampi from magnetic resonance images. Our training pipeline starts with optimizing an ANN to its maximum capacity, then employs a quick ANN-SNN conversion to initialize the corresponding spiking network. This is followed by spike-based backpropagation to fine-tune the converted SNN. In order to understand the reason behind performance decline in the converted SNNs, we conduct a set of experiments to investigate the output scaling issue. Furthermore, we explore the impact of binary and ternary representations in SNN networks and conduct an empirical evaluation of their performance through image classification and segmentation tasks.

Results and discussion

By employing our hybrid training scheme, we observe significant advantages over both ANN-SNN conversion and direct SNN training solutions in terms of segmentation accuracy and training efficiency. Experimental results demonstrate the effectiveness of our model in achieving our design goals.

Temporal Uncertainty Localization to Enable Human-in-the-Loop Analysis of Dynamic Contrast-Enhanced Cardiac MRI Datasets

Dynamic contrast-enhanced (DCE) cardiac magnetic resonance imaging (CMRI) is a widely used modality for diagnosing myocardial blood flow (perfusion) abnormalities. During a typical free-breathing DCE-CMRI scan, close to 300 time-resolved images of myocardial perfusion are acquired at various contrast "wash in/out" phases. Manual segmentation of myocardial contours in each time-frame of a DCE image series can be tedious and time-consuming, particularly when non-rigid motion correction has failed or is unavailable. While deep neural networks (DNNs) have shown promise for analyzing DCE-CMRI datasets, a "dynamic quality control" (dQC) technique for reliably detecting failed segmentations is lacking. Here we propose a new space-time uncertainty metric as a dQC tool for DNN-based segmentation of free-breathing DCE-CMRI datasets by validating the proposed metric on an external dataset and establishing a human-in-the-loop framework to improve the segmentation results. In the proposed approach, we referred the top 10% most uncertain segmentations as detected by our dQC tool to the human expert for refinement. This approach resulted in a significant increase in the Dice score (p < 0.001) and a notable decrease in the number of images with failed segmentation (16.2% to 11.3%) whereas the alternative approach of randomly selecting the same number of segmentations for human referral did not achieve any significant improvement. Our results suggest that the proposed dQC framework has the potential to accurately identify poor-quality segmentations and may enable efficient DNN-based analysis of DCE-CMRI in a human-in-the-loop pipeline for clinical interpretation and reporting of dynamic CMRI datasets.

Automated precision localization of peripherally inserted central catheter tip through model-agnostic multi-stage networks

BACKGROUND

Peripherally inserted central catheters (PICCs) have been widely used as one of the representative central venous lines (CVCs) due to their long-term intravascular access with low infectivity. However, PICCs have a fatal drawback of a high frequency of tip mispositions, increasing the risk of puncture, embolism, and complications such as cardiac arrhythmias. To automatically and precisely detect it, various attempts have been made by using the latest deep learning (DL) technologies. However, even with these approaches, it is still practically difficult to determine the tip location because the multiple fragments phenomenon (MFP) occurs in the process of predicting and extracting the PICC line required before predicting the tip.

OBJECTIVE

This study aimed to develop a system generally applied to existing models and to restore the PICC line more exactly by removing the MFs of the model output, thereby precisely localizing the actual tip position for detecting its misposition.

METHODS

To achieve this, we proposed a multi-stage DL-based framework post-processing the PICC line extraction result of the existing technology. Our method consists of the following three stages: 1. Existing PICC line segmentation network for a baseline, 2. Patch-based PICC line refinement network, 3. PICC line reconnection network. The proposed second and third-stage models address MFs caused by the sparseness of the PICC line and the line disconnection due to confusion with anatomical structures respectively, thereby enhancing tip detection.

RESULTS

To verify the objective performance of the proposed MFCN, internal validation and external validation were conducted. For internal validation, learning (130 samples) and verification (150 samples) were performed with 280 data, including PICC among Chest X-ray (CXR) images taken at our institution. External validation was conducted using a public dataset called the Royal Australian and New Zealand College of Radiologists (RANZCR), and training (130 samples) and validation (150 samples) were performed with 280 data of CXR images, including PICC, which has the same number as that for internal validation. The performance was compared by root mean squared error (RMSE) and the ratio of single fragment images (RatioSFI) (i.e., the rate at which model predicts PICC as multiple sub-lines) according to whether or not MFCN is applied to seven conventional models (i.e., FCDN, UNET, AUNET, TUNET, FCDN-HT, UNET-ELL, and UNET-RPN). In internal validation, when MFCN was applied to the existing single model, MFP was improved by an average of 45 %. The RMSE improved over 63% from an average of 27.54 mm (17.16 to 35.80 mm) to 9.77 mm (9.11 to 10.98 mm). In external validation, when MFCN was applied, the MFP incidence rate decreased by an average of 32% and the RMSE decreased by an average of 65%. Therefore, by applying the proposed MFCN, we observed the consistent detection performance improvement of PICC tip location compared to the existing model.

CONCLUSION

In this study, we applied the proposed technique to the existing technique and demonstrated that it provides high tip detection performance, proving its high versatility and superiority. Therefore, we believe, in countries and regions where radiologists are scarce, that the proposed DL approach will be able to effectively detect PICC misposition on behalf of radiologists.

Hippocampus segmentation after brain tumor resection via postoperative region synthesis

PURPOSE

Accurately segmenting the hippocampus is an essential step in brain tumor radiotherapy planning. Some patients undergo brain tumor resection beforehand, which can significantly alter the postoperative regions' appearances and intensity of the 3D MR images. However, there are limited tumor resection patient images for deep neural networks to be effective.

METHODS

We propose a novel automatic hippocampus segmentation framework via postoperative image synthesis. The variational generative adversarial network consists of intensity alignment and a weight-map-guided feature fusion module, which transfers the postoperative regions to the preoperative images. In addition, to further boost the performance of hippocampus segmentation, We design a joint training strategy to optimize the image synthesis network and the segmentation task simultaneously.

RESULTS

Comprehensive experiments demonstrate that our proposed method on the dataset with 48 nasopharyngeal carcinoma patients and 67 brain tumor patients observes consistent improvements over state-of-the-art methods.

CONCLUSION

The proposed postoperative image synthesis method act as a novel and powerful scheme to generate additional training data. Compared with existing deep learning methods, it achieves better accuracy for hippocampus segmentation of brain tumor patients who have undergone brain tumor resection. It can be used as an automatic contouring tool for hippocampus delineation in hippocampus-sparing radiotherapy.

Longitudinal changes in hippocampal texture from healthy aging to Alzheimer's disease

Early detection of Alzheimer's disease is essential to develop preventive treatment strategies. Detectible change in brain volume emerges relatively late in the pathogenic progression of disease, but microstructural changes caused by early neuropathology may cause subtle changes in the MR signal, quantifiable using texture analysis. Texture analysis quantifies spatial patterns in an image, such as smoothness, randomness and heterogeneity. We investigated whether the MRI texture of the hippocampus, an early site of Alzheimer's disease pathology, is sensitive to changes in brain microstructure before the onset of cognitive impairment. We also explored the longitudinal trajectories of hippocampal texture across the Alzheimer's continuum in relation to hippocampal volume and other biomarkers. Finally, we assessed the ability of texture to predict future cognitive decline, over and above hippocampal volume. Data were acquired from the Alzheimer's Disease Neuroimaging Initiative. Texture was calculated for bilateral hippocampi on 3T T1-weighted MRI scans. Two hundred and ninety-three texture features were reduced to five principal components that described 88% of total variance within cognitively unimpaired participants. We assessed cross-sectional differences in these texture components and hippocampal volume between four diagnostic groups: cognitively unimpaired amyloid-β- (n = 406); cognitively unimpaired amyloid-β+ (n = 213); mild cognitive impairment amyloid-β+ (n = 347); and Alzheimer's disease dementia amyloid-β+ (n = 202). To assess longitudinal texture change across the Alzheimer's continuum, we used a multivariate mixed-effects spline model to calculate a 'disease time' for all timepoints based on amyloid PET and cognitive scores. This was used as a scale on which to compare the trajectories of biomarkers, including volume and texture of the hippocampus. The trajectories were modelled in a subset of the data: cognitively unimpaired amyloid-β- (n = 345); cognitively unimpaired amyloid-β+ (n = 173); mild cognitive impairment amyloid-β+ (n = 301); and Alzheimer's disease dementia amyloid-β+ (n = 161). We identified a difference in texture component 4 at the earliest stage of Alzheimer's disease, between cognitively unimpaired amyloid-β- and cognitively unimpaired amyloid-β+ older adults (Cohen's d = 0.23, Padj = 0.014). Differences in additional texture components and hippocampal volume emerged later in the disease continuum alongside the onset of cognitive impairment (d = 0.30-1.22, Padj < 0.002). Longitudinal modelling of the texture trajectories revealed that, while most elements of texture developed over the course of the disease, noise reduced sensitivity for tracking individual textural change over time. Critically, however, texture provided additional information than was provided by volume alone to more accurately predict future cognitive change (d = 0.32-0.63, Padj < 0.0001). Our results support the use of texture as a measure of brain health, sensitive to Alzheimer's disease pathology, at a time when therapeutic intervention may be most effective.

RBS-Net: Hippocampus segmentation using multi-layer feature learning with the region, boundary and structure loss

Hippocampus has great influence over the Alzheimer's disease (AD) research because of its essential role as a biomarker in the human brain. Thus the performance of hippocampus segmentation influences the development of clinical research for brain disorders. Deep learning using U-net-like networks becomes prevalent in hippocampus segmentation on Magnetic Resonance Imaging (MRI) due to its efficiency and accuracy. However, current methods lose sufficient detailed information during pooling, which hinders the segmentation results. And weak supervision on the details like edges or positions results in fuzzy and coarse boundary segmentation, causing great differences between the segmentation and ground-truth. In view of these drawbacks, we propose a Region-Boundary and Structure Net (RBS-Net), which consists of a primary net and an auxiliary net. (1) Our primary net focuses on the region distribution of hippocampus and introduces a distance map for boundary supervision. Furthermore the primary net adds a multi-layer feature learning module to compensate the information loss during pooling and strengthen the differences between the foreground and background, improving the region and boundary segmentation. (2) The auxiliary net concentrates on the structure similarity and also utilizes the multi-layer feature learning module, and this parallel task can refine encoders by similarizing the structure of the segmentation and ground-truth. We train and test our network using 5-fold cross-validation on HarP, a public available hippocampus dataset. Experimental results demonstrate that our proposed RBS-Net achieves a Dice of 89.76% in average, outperforming several state-of-the-art hippocampus segmentation methods. Furthermore, in few shot circumstances, our proposed RBS-Net achieves better results in terms of a comprehensive evaluation compared to several state-of-the-art deep learning-based methods. Finally we can observe that visual segmentation results for the boundary and detailed regions are improved by our proposed RBS-Net.

HGM-cNet: Integrating hippocampal gray matter probability map into a cascaded deep learning framework improves hippocampus segmentation

A robust cascaded deep learning framework with integrated hippocampal gray matter (HGM) probability map was developed to improve the hippocampus segmentation (called HGM-cNet) due to its significance in various neuropsychiatric disorders such as Alzheimer's disease (AD). Particularly, the HGM-cNet cascaded two identical convolutional neural networks (CNN), where each CNN was devised by incorporating Attention Block, Residual Block, and DropBlock into the typical encoder-decoder architecture. The two CNNs were skip-connected between encoder components at each scale. The adoption of the cascaded deep learning framework was to conveniently incorporate the HGM probability map with the feature map generated by the first CNN. Experiments on 135T1-weighted MRI scans and manual hippocampal labels from publicly available ADNI-HarP dataset demonstrated that the proposed HGM-cNet outperformed seven multi-atlas-based hippocampus segmentation methods and six deep learning methods under comparison in most evaluation metrics. The Dice (average > 0.89 for both left and right hippocampus) was increased by around or more than 1% over other methods. The HGM-cNet also achieved a superior hippocampus segmentation performance in each group of cognitive normal, mild cognitive impairment, and AD. The stability, conveniences and generalizability of the cascaded deep learning framework with integrated HGM probability map in improving hippocampus segmentation was validated by replacing the proposed CNN with 3D-UNet, Atten-UNet, HippoDeep, QuickNet, DeepHarp, and TransBTS models. The integration of the HGM probability map in the cascaded deep learning framework was also demonstrated to facilitate capturing hippocampal atrophy more accurately than alternative methods in AD analysis. The codes are publicly available at https://github.com/Liu1436510768/HGM-cNet.git.

A review on Alzheimer's disease classification from normal controls and mild cognitive impairment using structural MR images

Alzheimer's disease (AD) is an irreversible neurodegenerative brain disorder that degrades the memory and cognitive ability in elderly people. The main reason for memory loss and reduction in cognitive ability is the structural changes in the brain that occur due to neuronal loss. These structural changes are most conspicuous in the hippocampus, cortex, and grey matter and can be assessed by using neuroimaging techniques viz. Positron Emission Tomography (PET), structural Magnetic Resonance Imaging (MRI) and functional MRI (fMRI), etc. Out of these neuroimaging techniques, structural MRI has evolved as the best technique as it indicates the best soft tissue contrast and high spatial resolution which is important for AD detection. Currently, the focus of researchers is on predicting the conversion of Mild Cognitive Impairment (MCI) into AD. MCI represents the transition state between expected cognitive changes with normal aging and Alzheimer's disease. Not every MCI patient progresses into Alzheimer's disease. MCI can develop into stable MCI (sMCI, patients are called non-converters) or into progressive MCI (pMCI, patients are diagnosed as MCI converters). This paper discusses the prognosis of MCI to AD conversion and presents a review of structural MRI-based studies for AD detection. AD detection framework includes feature extraction, feature selection, and classification process. This paper reviews the studies for AD detection based on different feature extraction methods and machine learning algorithms for classification. The performance of various feature extraction methods has been compared and it has been observed that the wavelet transform-based feature extraction method would give promising results for AD classification. The present study indicates that researchers are successful in classifying AD from Normal Controls (N_rmC) but, it still requires a lot of work to be done for MCI/ N_rmC and MCI/AD, which would help in detecting AD at its early stage.

Deep label fusion: A generalizable hybrid multi-atlas and deep convolutional neural network for medical image segmentation

Deep convolutional neural networks (DCNN) achieve very high accuracy in segmenting various anatomical structures in medical images but often suffer from relatively poor generalizability. Multi-atlas segmentation (MAS), while less accurate than DCNN in many applications, tends to generalize well to unseen datasets with different characteristics from the training dataset. Several groups have attempted to integrate the power of DCNN to learn complex data representations and the robustness of MAS to changes in image characteristics. However, these studies primarily focused on replacing individual components of MAS with DCNN models and reported marginal improvements in accuracy. In this study we describe and evaluate a 3D end-to-end hybrid MAS and DCNN segmentation pipeline, called Deep Label Fusion (DLF). The DLF pipeline consists of two main components with learnable weights, including a weighted voting subnet that mimics the MAS algorithm and a fine-tuning subnet that corrects residual segmentation errors to improve final segmentation accuracy. We evaluate DLF on five datasets that represent a diversity of anatomical structures (medial temporal lobe subregions and lumbar vertebrae) and imaging modalities (multi-modality, multi-field-strength MRI and Computational Tomography). These experiments show that DLF achieves comparable segmentation accuracy to nnU-Net (Isensee et al., 2020), the state-of-the-art DCNN pipeline, when evaluated on a dataset with similar characteristics to the training datasets, while outperforming nnU-Net on tasks that involve generalization to datasets with different characteristics (different MRI field strength or different patient population). DLF is also shown to consistently improve upon conventional MAS methods. In addition, a modality augmentation strategy tailored for multimodal imaging is proposed and demonstrated to be beneficial in improving the segmentation accuracy of learning-based methods, including DLF and DCNN, in missing data scenarios in test time as well as increasing the interpretability of the contribution of each individual modality.

Deep Grey Matter Volume is Reduced in Amateur Boxers as Compared to Healthy Age-matched Controls

Purpose

Mild traumatic brain injuries (mTBI) sustained during contact sports like amateur boxing are found to have long-term sequelae, being linked to an increased risk of developing neurological conditions like Parkinson’s disease. The aim of this study was to assess differences in volume of anatomical brain structures between amateur boxers and control subjects with a special interest in the affection of deep grey matter structures.

Methods

A total of 19 amateur boxers and 19 healthy controls (HC), matched for age and intelligence quotient (IQ), underwent 3T magnetic resonance imaging (MRI) as well as neuropsychological testing. Body mass index (BMI) was evaluated for every subject and data about years of boxing training and number of fights were collected for each boxer. The acquired 3D high resolution T1 weighted MR images were analyzed to measure the volumes of cortical grey matter (GM), white matter (WM), cerebrospinal fluid (CSF) and deep grey matter structures. Multivariate analysis was applied to reveal differences between groups referencing deep grey matter structures to normalized brain volume (NBV) to adjust for differences in head size and brain volume as well as adding BMI as cofactor.

Results

Total intracranial volume (TIV), comprising GM, WM and CSF, was lower in boxers compared to controls (by 7.1%, P = 0.009). Accordingly, GM (by 5.5%, P = 0.038) and WM (by 8.4%, P = 0.009) were reduced in boxers. Deep grey matter showed statistically lower volumes of the thalamus (by 8.1%, P = 0.006), caudate nucleus (by 11.1%, P = 0.004), putamen (by 8.1%, P = 0.011), globus pallidus (by 9.6%, P = 0.017) and nucleus accumbens (by 13.9%, P = 0.007) but not the amygdala (by 5.5%, P = 0.221), in boxers compared to HC.

Conclusion

Several deep grey matter structures were reduced in volume in the amateur boxer group. Furthermore, longitudinal studies are needed to determine the damage pattern affecting deep grey matter structures and its neuropsychological relevance.

Assessment of Bioenergetic Deficits in Patients With Parkinson Disease and Progressive Supranuclear Palsy Using 31P-MRSI

BACKGROUND AND OBJECTIVE

Bioenergetic disturbance, mainly caused by mitochondrial dysfunction, is an established pathophysiologic phenomenon in neurodegenerative movement disorders. The in vivo assessment of brain energy metabolism by 31phosphorus magnetic resonance spectroscopy imaging could provide pathophysiologic insights and serve in the differential diagnosis of parkinsonian disorders. In this study, we investigated such aspects of the underlying pathophysiology in patients with idiopathic Parkinson disease (PwPD) and progressive supranuclear palsy (PwPSP).

METHODS

In total, 30 PwPD, 16 PwPSP, and 25 healthy control subjects (HCs) underwent a clinical examination, structural magnetic resonance imaging, and 31phosphorus magnetic resonance spectroscopy imaging of the forebrain and basal ganglia in a cross-sectional study.

RESULTS

High-energy phosphate metabolites were remarkably decreased in PwPD, particularly in the basal ganglia (-42% compared with HCs and -43% compared with PwPSP, p < 0.0001). This result was not confounded by morphometric brain differences. By contrast, PwPSP had normal levels of high-energy energy metabolites. Thus, the combination of morphometric and metabolic neuroimaging was able to discriminate PwPD from PwPSP with an accuracy of up to 0.93 [95%-CI: 0.91-0.94].

DISCUSSION

Our study shows that mitochondrial dysfunction and bioenergetic depletion contribute to idiopathic Parkinson disease pathophysiology but not to progressive supranuclear palsy. Combined morphometric and metabolic imaging could serve as an accompanying diagnostic biomarker in the neuroimaging-guided differential diagnosis of these parkinsonian disorders.

CLASSIFICATION OF EVIDENCE

This study provides Class III evidence that 31phosphorus magnetic resonance spectroscopy imaging combined with morphometric MRI can differentiate PwPD from PwPSP.

CerebNet: A fast and reliable deep-learning pipeline for detailed cerebellum sub-segmentation

Quantifying the volume of the cerebellum and its lobes is of profound interest in various neurodegenerative and acquired diseases. Especially for the most common spinocerebellar ataxias (SCA), for which the first antisense oligonculeotide-base gene silencing trial has recently started, there is an urgent need for quantitative, sensitive imaging markers at pre-symptomatic stages for stratification and treatment assessment. This work introduces CerebNet, a fully automated, extensively validated, deep learning method for the lobular segmentation of the cerebellum, including the separation of gray and white matter. For training, validation, and testing, T1-weighted images from 30 participants were manually annotated into cerebellar lobules and vermal sub-segments, as well as cerebellar white matter. CerebNet combines FastSurferCNN, a UNet-based 2.5D segmentation network, with extensive data augmentation, e.g. realistic non-linear deformations to increase the anatomical variety, eliminating additional preprocessing steps, such as spatial normalization or bias field correction. CerebNet demonstrates a high accuracy (on average 0.87 Dice and 1.742mm Robust Hausdorff Distance across all structures) outperforming state-of-the-art approaches. Furthermore, it shows high test-retest reliability (average ICC >0.97 on OASIS and Kirby) as well as high sensitivity to disease effects, including the pre-ataxic stage of spinocerebellar ataxia type 3 (SCA3). CerebNet is compatible with FreeSurfer and FastSurfer and can analyze a 3D volume within seconds on a consumer GPU in an end-to-end fashion, thus providing an efficient and validated solution for assessing cerebellum sub-structure volumes. We make CerebNet available as source-code (https://github.com/Deep-MI/FastSurfer).

Comparison of atlas-based and deep learning methods for organs at risk delineation on head-and-neck CT images using an automated treatment planning system

BACKGROUND AND PURPOSE

To investigate the performance of head-and-neck (HN) organs-at-risk (OAR) automatic segmentation (AS) using four atlas-based (ABAS) and two deep learning (DL) solutions.

MATERIAL AND METHODS

All patients underwent iodine contrast-enhanced planning CT. Fourteen OAR were manually delineated. DL.1 and DL.2 solutions were trained with 63 mono-centric patients and > 1000 multi-centric patients, respectively. Ten and 15 patients with varied anatomies were selected for the atlas library and for testing, respectively. The evaluation was based on geometric indices (DICE coefficient and 95th percentile-Hausdorff Distance (HD_95%)), time needed for manual corrections and clinical dosimetric endpoints obtained using automated treatment planning.

RESULTS

Both DICE and HD_95% results indicated that DL algorithms generally performed better compared with ABAS algorithms for automatic segmentation of HN OAR. However, the hybrid-ABAS (ABAS.3) algorithm sometimes provided the highest agreement to the reference contours compared with the 2 DL. Compared with DL.2 and ABAS.3, DL.1 contours were the fastest to correct. For the 3 solutions, the differences in dose distributions obtained using AS contours and AS + manually corrected contours were not statistically significant. High dose differences could be observed when OAR contours were at short distances to the targets. However, this was not always interrelated.

CONCLUSION

DL methods generally showed higher delineation accuracy compared with ABAS methods for AS segmentation of HN OAR. Most ABAS contours had high conformity to the reference but were more time consuming than DL algorithms, especially when considering the computing time and the time spent on manual corrections.

An Efficient Optimization Approach for Glioma Tumor Segmentation in Brain MRI

Glioma is an aggressive type of cancer that develops in the brain or spinal cord. Due to many differences in its shape and appearance, accurate segmentation of glioma for identifying all parts of the tumor and its surrounding cancerous tissues is a challenging task. In recent researches, the combination of multi-atlas segmentation and machine learning methods provides robust and accurate results by learning from annotated atlas datasets. To overcome the side effects of limited existing information on atlas-based segmentation, and the long training phase of learning methods, we proposed a semi-supervised unified framework for multi-label segmentation that formulates this problem in terms of a Markov Random Field energy optimization on a parametric graph. To evaluate the proposed framework, we apply it to publicly available BRATS datasets, including low- and high-grade glioma tumors. Experimental results indicate competitive performance compared to the state-of-the-art methods. Compared with the top ranked methods, the proposed framework obtains the best dice score for segmenting of "whole tumor" (WT), "tumor core" (TC ) and "enhancing active tumor" (ET) regions. The achieved accuracy is 94[Formula: see text] characterized by the mean dice score. The motivation of using MRF graph is to map the segmentation problem to an optimization model in a graphical environment. Therefore, by defining perfect graph structure and optimum constraints and flows in the continuous max-flow model, the segmentation is performed precisely.

SynthStrip: skull-stripping for any brain image

The removal of non-brain signal from magnetic resonance imaging (MRI) data, known as skull-stripping, is an integral component of many neuroimage analysis streams. Despite their abundance, popular classical skull-stripping methods are usually tailored to images with specific acquisition properties, namely near-isotropic resolution and T1-weighted (T1w) MRI contrast, which are prevalent in research settings. As a result, existing tools tend to adapt poorly to other image types, such as stacks of thick slices acquired with fast spin-echo (FSE) MRI that are common in the clinic. While learning-based approaches for brain extraction have gained traction in recent years, these methods face a similar burden, as they are only effective for image types seen during the training procedure. To achieve robust skull-stripping across a landscape of imaging protocols, we introduce SynthStrip, a rapid, learning-based brain-extraction tool. By leveraging anatomical segmentations to generate an entirely synthetic training dataset with anatomies, intensity distributions, and artifacts that far exceed the realistic range of medical images, SynthStrip learns to successfully generalize to a variety of real acquired brain images, removing the need for training data with target contrasts. We demonstrate the efficacy of SynthStrip for a diverse set of image acquisitions and resolutions across subject populations, ranging from newborn to adult. We show substantial improvements in accuracy over popular skull-stripping baselines - all with a single trained model. Our method and labeled evaluation data are available at https://w3id.org/synthstrip.

How segmentation methods affect hippocampal radiomic feature accuracy in Alzheimer's disease analysis?

OBJECTIVES

Hippocampal radiomic features (HRFs) can serve as biomarkers in Alzheimer's disease (AD). However, how different hippocampal segmentation methods affect HRFs in AD is still unknown. The aim of the study was to investigate how different segmentation methods affect HRF accuracy in AD analysis.

METHODS

A total of 1650 subjects were identified from the Alzheimer's Disease Neuroimaging Initiative database (ADNI). The mini-mental state examination (MMSE) and Alzheimer's disease assessment scale (ADAS-cog13) were also adopted. After calculating the HRFs of intensity, shape, and textural features from each side of the hippocampus in structural magnetic resonance imaging (sMRI), the consistency of HRFs calculated from 7 different hippocampal segmentation methods was validated, and the performance of machine learning-based classification of AD vs. normal control (NC) adopting the different HRFs was also examined. Additional 571 subjects from the European DTI Study on Dementia database (EDSD) were to validate the consistency of results.

RESULTS

Between different segmentations, HRFs showed a high measurement consistency (R > 0.7), a high significant consistency between NC, mild cognitive impairment (MCI), and AD (T-value plot, R > 0.8), and consistent significant correlations between HRFs and MMSE/ADAS-cog13 (p < 0.05). The best NC vs. AD classification was obtained when the hippocampus was sufficiently segmented by primitive majority voting (threshold = 0.2). High consistent results were reproduced from independent EDSD cohort.

CONCLUSIONS

HRFs exhibited high consistency across different hippocampal segmentation methods, and the best performance in AD classification was obtained when HRFs were extracted by the naïve majority voting method with a more sufficient segmentation and relatively low hippocampus segmentation accuracy.

KEY POINTS

• The hippocampal radiomic features exhibited high measurement/statistical/clinical consistency across different hippocampal segmentation methods. • The best performance in AD classification was obtained when hippocampal radiomics were extracted by the naïve majority voting method with a more sufficient segmentation and relatively low hippocampus segmentation accuracy.

Altered Cerebral Microstructure in Adults With Atrial Septal Defect and Ventricular Septal Defect Repaired in Childhood

Background

Delayed brain development, brain injury, and neurodevelopmental disabilities are commonly observed in infants operated for complex congenital heart defect. Our previous findings of poorer neurodevelopmental outcomes in individuals operated for simple congenital heart defects calls for further etiological clarification. Hence, we examined the microstructural tissue composition in cerebral cortex and subcortical structures in comparison to healthy controls and whether differences were associated with neurodevelopmental outcomes.

Methods and Results

Adults (n=62) who underwent surgical closure of an atrial septal defect (n=33) or a ventricular septal defect (n=29) in childhood and a group of healthy, matched controls (n=38) were enrolled. Brain diffusional kurtosis imaging and neuropsychological assessment were performed. Cortical and subcortical tissue microstructure were assessed using mean kurtosis tensor and mean diffusivity and compared between groups and tested for associations with neuropsychological outcomes. Alterations in microstructural tissue composition were found in the parietal, temporal, and occipital lobes in the congenital heart defects, with distinct mean kurtosis tensor cluster‐specific changes in the right visual cortex (pericalcarine gyrus, P=0.002; occipital part of fusiform and lingual gyri, P=0.019). Altered microstructural tissue composition in the subcortical structures was uncovered in atrial septal defects but not in ventricular septal defects. Associations were found between altered cerebral microstructure and social recognition and executive function.

Conclusions

Children operated for simple congenital heart defects demonstrated altered microstructural tissue composition in the cerebral cortex and subcortical structures during adulthood when compared with healthy peers. Alterations in cerebral microstructural tissue composition were associated with poorer neuropsychological performance.

Registration

URL: https://www.clinicaltrials.gov; Unique identifier: NCT03871881.

vol2Brain: A New Online Pipeline for Whole Brain MRI Analysis

Automatic and reliable quantitative tools for MR brain image analysis are a very valuable resource for both clinical and research environments. In the past few years, this field has experienced many advances with successful techniques based on label fusion and more recently deep learning. However, few of them have been specifically designed to provide a dense anatomical labeling at the multiscale level and to deal with brain anatomical alterations such as white matter lesions (WML). In this work, we present a fully automatic pipeline (vol2Brain) for whole brain segmentation and analysis, which densely labels (N &gt; 100) the brain while being robust to the presence of WML. This new pipeline is an evolution of our previous volBrain pipeline that extends significantly the number of regions that can be analyzed. Our proposed method is based on a fast and multiscale multi-atlas label fusion technology with systematic error correction able to provide accurate volumetric information in a few minutes. We have deployed our new pipeline within our platform volBrain (www.volbrain.upv.es), which has been already demonstrated to be an efficient and effective way to share our technology with the users worldwide.

BrainSec: Automated Brain Tissue Segmentation Pipeline for Scalable Neuropathological Analysis

As neurodegenerative disease pathological hallmarks have been reported in both grey matter (GM) and white matter (WM) with different density distributions, automating the segmentation process of GM/WM would be extremely advantageous for aiding in neuropathologic deep phenotyping. Standard segmentation methods typically involve manual annotations, where a trained researcher traces the delineation of GM/WM in ultra-high-resolution Whole Slide Images (WSIs). This method can be time-consuming and subjective, preventing a scalable analysis on pathology images. This paper proposes an automated segmentation pipeline (BrainSec) combining a Convolutional Neural Network (CNN) module for segmenting GM/WM regions and a post-processing module to remove artifacts/residues of tissues. The final output generates XML annotations that can be visualized via Aperio ImageScope. First, we investigate two baseline models for medical image segmentation: FCN, and U-Net. Then we propose a patch-based approach, BrainSec, to classify the GM/WM/background regions. We demonstrate BrainSec is robust and has reliable performance by testing it on over 180 WSIs that incorporate numerous unique cases as well as distinct neuroanatomic brain regions. We also apply gradient-weighted class activation mapping (Grad-CAM) to interpret the segmentation masks and provide relevant explanations and insights. In addition, we have integrated BrainSec with an existing Amyloid-β pathology classification model into a unified framework (without incurring significant computation complexity) to identify pathologies, visualize their distributions, and quantify each type of pathologies in segmented GM/WM regions, respectively.

A unified 3D map of microscopic architecture and MRI of the human brain

We present the first three-dimensional (3D) concordance maps of cyto- and fiber architecture of the human brain, combining histology, immunohistochemistry, and 7-T quantitative magnetic resonance imaging (MRI), in two individual specimens. These 3D maps each integrate data from approximately 800 microscopy sections per brain, showing neuronal and glial cell bodies, nerve fibers, and interneuronal populations, as well as ultrahigh-field quantitative MRI, all coaligned at the 200-μm scale to the stacked blockface images obtained during sectioning. These unprecedented 3D multimodal datasets are shared without any restrictions and provide a unique resource for the joint study of cell and fiber architecture of the brain, detailed anatomical atlasing, or modeling of the microscopic underpinnings of MRI contrasts.

Molecular and Brain Volume Changes Following Aerobic Exercise, Cognitive and Combined Training in Physically Inactive Healthy Late-Middle-Aged Adults: The Projecte Moviment Randomized Controlled Trial

Behavioral interventions have shown promising neuroprotective effects, but the cascade of molecular, brain and behavioral changes involved in these benefits remains poorly understood. Projecte Moviment is a 12-week (5 days per week—45 min per day) multi-domain, single-blind, proof-of-concept randomized controlled trial examining the cognitive effect and underlying mechanisms of an aerobic exercise (AE), computerized cognitive training (CCT) and a combined (COMB) groups compared to a waitlist control group. Adherence was > 80% for 82/109 participants recruited (62% female; age = 58.38 ± 5.47). In this study we report intervention-related changes in plasma biomarkers (BDNF, TNF-α, HGF, ICAM-1, SDF1-α) and structural-MRI (brain volume) and how they related to changes in physical activity and individual variables (age and sex) and their potential role as mediators in the cognitive changes. Our results show that although there were no significant changes in molecular biomarker concentrations in any intervention group, changes in ICAM-1 and SDF1-α were negatively associated with changes in physical activity outcomes in AE and COMB groups. Brain volume changes were found in the CCT showing a significant increase in precuneus volume. Sex moderated the brain volume change in the AE and COMB groups, suggesting that men may benefit more than women. Changes in molecular biomarkers and brain volumes did not significantly mediate the cognitive-related benefits found previously for any group. This study shows crucial initial molecular and brain volume changes related to lifestyle interventions at early stages and highlights the value of examining activity parameters, individual difference characteristics and using a multi-level analysis approach to address these questions.

Combining plasma phospho-tau and accessible measures to evaluate progression to Alzheimer's dementia in mild cognitive impairment patients

Background

Up to now, there are no clinically available minimally invasive biomarkers to accurately identify mild cognitive impairment (MCI) patients who are at greater risk to progress to Alzheimer’s disease (AD) dementia. The recent advent of blood-based markers opens the door for more accessible biomarkers. We aimed to identify which combinations of AD related plasma biomarkers and other easily accessible assessments best predict progression to AD dementia in patients with mild cognitive impairment (MCI).

Methods

We included patients with amnestic MCI (n = 110) followed prospectively over 3 years to assess clinical status. Baseline plasma biomarkers (amyloid-β 42/40, phosphorylated tau217 [p-tau217], neurofilament light and glial fibrillary acidic protein), hippocampal volume, APOE genotype, and cognitive tests were available. Logistic regressions with conversion to amyloid-positive AD dementia within 3 years as outcome was used to evaluate the performance of different biomarkers measured at baseline, used alone or in combination. The first analyses included only the plasma biomarkers to determine the ones most related to AD dementia conversion. Second, hippocampal volume, APOE genotype and a brief cognitive composite score (mPACC) were combined with the best plasma biomarker.

Results

Of all plasma biomarker combinations, p-tau217 alone had the best performance for discriminating progression to AD dementia vs all other combinations (AUC 0.84, 95% CI 0.75–0.93). Next, combining p-tau217 with hippocampal volume, cognition, and APOE genotype provided the best discrimination between MCI progressors vs. non-progressors (AUC 0.89, 0.82–0.95). Across the few best models combining different markers, p-tau217 and cognition were consistently the main contributors. The most parsimonious model including p-tau217 and cognition had a similar model fit, but a slightly lower AUC (0.87, 0.79–0.95, p = 0.07).

Conclusion

We identified that combining plasma p-tau217 and a brief cognitive composite score was strongly related to greater risk of progression to AD dementia in MCI patients, suggesting that these measures could be key components of future prognostic algorithms for early AD.

Trial registration

NCT01028053, registered December 9, 2009.

Deep 3D Neural Network for Brain Structures Segmentation Using Self-Attention Modules in MRI Images

In recent years, the use of deep learning-based models for developing advanced healthcare systems has been growing due to the results they can achieve. However, the majority of the proposed deep learning-models largely use convolutional and pooling operations, causing a loss in valuable data and focusing on local information. In this paper, we propose a deep learning-based approach that uses global and local features which are of importance in the medical image segmentation process. In order to train the architecture, we used extracted three-dimensional (3D) blocks from the full magnetic resonance image resolution, which were sent through a set of successive convolutional neural network (CNN) layers free of pooling operations to extract local information. Later, we sent the resulting feature maps to successive layers of self-attention modules to obtain the global context, whose output was later dispatched to the decoder pipeline composed mostly of upsampling layers. The model was trained using the Mindboggle-101 dataset. The experimental results showed that the self-attention modules allow segmentation with a higher Mean Dice Score of 0.90 ± 0.036 compared with other UNet-based approaches. The average segmentation time was approximately 0.038 s per brain structure. The proposed model allows tackling the brain structure segmentation task properly. Exploiting the global context that the self-attention modules incorporate allows for more precise and faster segmentation. We segmented 37 brain structures and, to the best of our knowledge, it is the largest number of structures under a 3D approach using attention mechanisms.

Hippocampus Segmentation Using U-Net Convolutional Network from Brain Magnetic Resonance Imaging (MRI)

Hippocampus is a part of the limbic system in human brain that plays an important role in forming memories and dealing with intellectual abilities. In most of the neurological disorders related to dementia, such as, Alzheimer's disease, hippocampus is one of the earliest affected regions. Because there are no effective dementia drugs, an ambient assisted living approach may help to prevent or slow the progression of dementia. By segmenting and analyzing the size/shape of hippocampus, it may be possible to classify the early dementia stages. Because of complex structure, traditional image segmentation techniques can't segment hippocampus accurately. Machine learning (ML) is a well known tool in medical image processing that can predict and deliver the outcomes accurately by learning from it's previous results. Convolutional Neural Networks (CNN) is one of the most popular ML algorithms. In this work, a U-Net Convolutional Network based approach is used for hippocampus segmentation from 2D brain images. It is observed that, the original U-Net architecture can segment hippocampus with an average performance rate of 93.6%, which outperforms all other discussed state-of-arts. By using a filter size of [Formula: see text], the original U-Net architecture performs a sequence of convolutional processes. We tweaked the architecture further to extract more relevant features by replacing all [Formula: see text] kernels with three alternative kernels of sizes [Formula: see text], [Formula: see text], and [Formula: see text]. It is observed that, the modified architecture achieved an average performance rate of 96.5%, which outperforms the original U-Net model convincingly.

Ventricular features as reliable differentiators between bvFTD and other dementias

•

Our results showed a consistent pattern of ventricle enlargement in the bvFTD patients, particularly in the anterior parts of the frontal and temporal horns of the lateral ventricles.

•

The estimation of the proposed ventricular anteroposterior ratio (APR) resulted in statistically significant difference compared to all other groups.

•

Our study proposes an easy to obtain and generalizable ventricle-based feature (APR) from T1-weighted structural MRI (routinely acquired and available in the clinic) that can be used not only to differentiate bvFTD from normal subjects, but also from other FTD variants (SV and PNFA), MCI, and AD patients.

•

We have made our ventricle feature estimation and bvFTD diagnosis tool (VentRa) publicly available, allowing application of our model in other studies.

•

If validated in a prospective study, VentRa has the potential to aid bvFTD diagnosis, particularly in settings where access to specialized FTD care is limited.

Introduction

Lateral ventricles are reliable and sensitive indicators of brain atrophy and disease progression in behavioral variant frontotemporal dementia (bvFTD). We aimed to investigate whether an automated tool using ventricular features could improve diagnostic accuracy in bvFTD across neurodegenerative diseases.

Methods

Using 678 subjects −69 bvFTD, 38 semantic variant, 37 primary non-fluent aphasia, 218 amyloid + mild cognitive impairment, 74 amyloid + Alzheimer’s Dementia and 242 normal controls- with a total of 2750 timepoints, lateral ventricles were segmented and differences in ventricular features were assessed between bvFTD, normal controls and other dementia cohorts.

Results

Ventricular antero-posterior ratio (APR) was the only feature that was significantly different and increased faster in bvFTD compared to all other cohorts. We achieved a 10-fold cross-validation accuracy of 80% (77% sensitivity, 82% specificity) in differentiating bvFTD from all other cohorts with other ventricular features (i.e., total ventricular volume and left–right lateral ventricle ratios), and 76% accuracy using only the single APR feature.

Discussion

Ventricular features, particularly the APR, might be reliable and easy-to-implement markers for bvFTD diagnosis. We have made our ventricle feature estimation and bvFTD diagnostic tool publicly available, allowing application of our model in other studies.

A novel deep learning based hippocampus subfield segmentation method

The automatic assessment of hippocampus volume is an important tool in the study of several neurodegenerative diseases such as Alzheimer's disease. Specifically, the measurement of hippocampus subfields properties is of great interest since it can show earlier pathological changes in the brain. However, segmentation of these subfields is very difficult due to their complex structure and for the need of high-resolution magnetic resonance images manually labeled. In this work, we present a novel pipeline for automatic hippocampus subfield segmentation based on a deeply supervised convolutional neural network. Results of the proposed method are shown for two available hippocampus subfield delineation protocols. The method has been compared to other state-of-the-art methods showing improved results in terms of accuracy and execution time.

Automated Segmentation and Connectivity Analysis for Normal Pressure Hydrocephalus

Objective and Impact Statement. We propose an automated method of predicting Normal Pressure Hydrocephalus (NPH) from CT scans. A deep convolutional network segments regions of interest from the scans. These regions are then combined with MRI information to predict NPH. To our knowledge, this is the first method which automatically predicts NPH from CT scans and incorporates diffusion tractography information for prediction. Introduction. Due to their low cost and high versatility, CT scans are often used in NPH diagnosis. No well-defined and effective protocol currently exists for analysis of CT scans for NPH. Evans' index, an approximation of the ventricle to brain volume using one 2D image slice, has been proposed but is not robust. The proposed approach is an effective way to quantify regions of interest and offers a computational method for predicting NPH. Methods. We propose a novel method to predict NPH by combining regions of interest segmented from CT scans with connectome data to compute features which capture the impact of enlarged ventricles by excluding fiber tracts passing through these regions. The segmentation and network features are used to train a model for NPH prediction. Results. Our method outperforms the current state-of-the-art by 9 precision points and 29 recall points. Our segmentation model outperforms the current state-of-the-art in segmenting the ventricle, gray-white matter, and subarachnoid space in CT scans. Conclusion. Our experimental results demonstrate that fast and accurate volumetric segmentation of CT brain scans can help improve the NPH diagnosis process, and network properties can increase NPH prediction accuracy.

Sex-specific associations between maternal pregnancy-specific anxiety and newborn amygdalar volumes - preliminary findings from the FinnBrain Birth Cohort Study

Previous literature links maternal pregnancy-specific anxiety (PSA) with later difficulties in child emotional and social cognition as well as memory, functions closely related to the amygdala and the hippocampus. Some evidence also suggests that PSA affects child amygdalar volumes in a sex-dependent way. However, no studies investigating the associations between PSA and newborn amygdalar and hippocampal volumes have been reported. We investigated the associations between PSA and newborn amygdalar and hippocampal volumes and whether associations are sex-specific in 122 healthy newborns (68 males/54 females) scanned at 2-5 weeks postpartum. PSA was measured at gestational week 24 with the Pregnancy-Related Anxiety Questionnaire Revised 2 (PRAQ-R2). The associations were analyzed with linear regression controlling for confounding variables. PSA was associated positively with left amygdalar volume in girls, but no significant main effect was found in the whole group or in boys. No significant main or sex-specific effect was found for hippocampal volumes. Although this was an exploratory study, the findings suggest a sexually dimorphic association of mid-pregnancy PSA with newborn amygdalar volumes.

Automatic segmentation of Deep Grey Nuclei using a high-resolution 7T MRI Atlas - quantification of T1 values in healthy volunteers

We present a new consensus atlas of deep grey nuclei obtained by shape-based averaging of manual segmentation of two experienced neuroradiologists and optimized from 7T MP2RAGE images acquired at (0.6mm)³ in 60 healthy subjects. A group-wise normalization method was used to build a high-contrast and high-resolution T₁ -weighted brain template (0.5mm)³ using data from 30 out of the 60 controls. Delineation of 24 deep grey nuclei per hemisphere, including the claustrum and twelve thalamic nuclei, was then performed by two expert neuroradiologists and reviewed by a third neuroradiologist according to tissue contrast and external references based on the Morel atlas. Corresponding deep grey matter structures were also extracted from the Morel and CIT168 atlases. The data-derived, Morel and CIT168 atlases were all applied at the individual level using non-linear registration to fit the subject reference and to extract absolute mean quantitative T₁ values derived from the 3D-MP2RAGE volumes, after correction for residual B₁ ⁺ biases. Three metrics (The Dice and the volumetric similarity coefficients, and a novel Hausdorff distance) were used to estimate the inter-rater agreement of manual MRI segmentation and inter-atlas variability, and these metrics were measured to quantify biases due to image registration and their impact on the measurements of the quantitative T₁ values was highlighted. This represents a fully-automated segmentation process permitting the extraction of unbiased normative T₁ values in a population of young healthy controls as a reference for characterizing subtle structural alterations of deep grey nuclei relevant to a range of neurological diseases.

SOMA: Subject-, object-, and modality-adapted precision atlas approach for automatic anatomy recognition and delineation in medical images

PURPOSE

In the multi-atlas segmentation (MAS) method, a large enough atlas set, which can cover the complete spectrum of the whole population pattern of the target object will benefit the segmentation quality. However, the difficulty in obtaining and generating such a large set of atlases and the computational burden required in the segmentation procedure make this approach impractical. In this paper, we propose a method called SOMA to select subject-, object-, and modality-adapted precision atlases for automatic anatomy recognition in medical images with pathology, following the idea that different regions of the target object in a novel image can be recognized by different atlases with regionally best similarity, so that effective atlases have no need to be globally similar to the target subject and also have no need to be overall similar to the target object.

METHODS

The SOMA method consists of three main components: atlas building, object recognition, and object delineation. Considering the computational complexity, we utilize an all-to-template strategy to align all images to the same image space belonging to the root image determined by the minimum spanning tree (MST) strategy among a subset of radiologically near-normal images. The object recognition process is composed of two stages: rough recognition and refined recognition. In rough recognition, subimage matching is conducted between the test image and each image of the whole atlas set, and only the atlas corresponding to the best-matched subimage contributes to the recognition map regionally. The frequency of best match for each atlas is recorded by a counter, and the atlases with the highest frequencies are selected as the precision atlases. In refined recognition, only the precision atlases are examined, and the subimage matching is conducted in a nonlocal manner of searching to further increase the accuracy of boundary matching. Delineation is based on a U-net-based deep learning network, where the original gray scale image together with the fuzzy map from refined recognition compose a two-channel input to the network, and the output is a segmentation map of the target object.

RESULTS

Experiments are conducted on computed tomography (CT) images with different qualities in two body regions - head and neck (H&N) and thorax, from 298 subjects with nine objects and 241 subjects with six objects, respectively. Most objects achieve a localization error within two voxels after refined recognition, with marked improvement in localization accuracy from rough to refined recognition of 0.6-3 mm in H&N and 0.8-4.9 mm in thorax, and also in delineation accuracy (Dice coefficient) from refined recognition to delineation of 0.01-0.11 in H&N and 0.01-0.18 in thorax.

CONCLUSIONS

The SOMA method shows high accuracy and robustness in anatomy recognition and delineation. The improvements from rough to refined recognition and further to delineation, as well as immunity of recognition accuracy to varying image and object qualities, demonstrate the core principles of SOMA where segmentation accuracy increases with precision atlases and gradually refined object matching.

A sub+cortical fMRI-based surface parcellation

Both cortical and subcortical structures are organized into a large number of distinct areas reflecting functional and cytoarchitectonic differences. Mapping these areas is of fundamental importance to neuroscience. A central obstacle to this task is the inaccuracy associated with bringing results from individuals into a common space. The vast individual differences in morphology pose a serious problem for volumetric registration. Surface‐based approaches fare substantially better, but have thus far been used only for cortical parcellation, leaving subcortical parcellation in volumetric space. We extend the surface‐based approach to include also the subcortical deep gray‐matter structures, thus achieving a uniform representation across both cortex and subcortex, suitable for use with surface‐based metrics that span these structures, for example, white/gray contrast. Using data from the Enhanced Nathan Klein Institute—Rockland Sample, limited to individuals between 19 and 69 years of age, we generate a functional parcellation of both the cortical and subcortical surfaces. To assess this extended parcellation, we show that (a) our parcellation provides greater homogeneity of functional connectivity patterns than do arbitrary parcellations matching in the number and size of parcels; (b) our parcels align with known cortical and subcortical architecture; and (c) our extended functional parcellation provides an improved fit to the complexity of life‐span (6–85 years) changes in white/gray contrast data compared to arbitrary parcellations matching in the number and size of parcels, supporting its use with surface‐based measures. We provide our extended functional parcellation for the use of the neuroimaging community.

Deep Learning-Based Segmentation and Uncertainty Assessment for Automated Analysis of Myocardial Perfusion MRI Datasets Using Patch-Level Training and Advanced Data Augmentation

In this work, we develop a patch-level training approach and a task-driven intensity-based augmentation method for deep-learning-based segmentation of motion-corrected perfusion cardiac magnetic resonance imaging (MRI) datasets. Further, the proposed method generates an image-based uncertainty map thanks to a novel spatial sliding-window approach used during patch-level training, hence allowing for uncertainty quantification. Using the quantified uncertainty, we detect the out-of-distribution test data instances so that the end-user can be alerted that the test data is not suitable for the trained network. This feature has the potential to enable a more reliable integration of the proposed deep learning-based framework into clinical practice. We test our approach on external MRI data acquired using a different acquisition protocol to demonstrate the robustness of our performance to variations in pulse-sequence parameters. The presented results further demonstrate that our deep-learning image segmentation approach trained with the proposed data-augmentation technique incorporating spatiotemporal (2D+time) patches is superior to the state-of-the-art 2D approach in terms of generalization performance.

Automatic Prediction of Cognitive and Functional Decline Can Significantly Decrease the Number of Subjects Required for Clinical Trials in Early Alzheimer's Disease

Background:

While both cognitive and magnetic resonance imaging (MRI) data has been used to predict progression in Alzheimer’s disease, heterogeneity between patients makes it challenging to predict the rate of cognitive and functional decline for individual subjects.

Objective:

To investigate prognostic power of MRI-based biomarkers of medial temporal lobe atrophy and macroscopic tissue change to predict cognitive decline in individual patients in clinical trials of early Alzheimer’s disease.

Methods:

Data used in this study included 312 patients with mild cognitive impairment from the ADNI dataset with baseline MRI, cerebrospinal fluid amyloid-β, cognitive test scores, and a minimum of two-year follow-up information available. We built a prognostic model using baseline cognitive scores and MRI-based features to determine which subjects remain stable and which functionally decline over 2 and 3-year follow-up periods.

Results:

Combining both sets of features yields 77%accuracy (81%sensitivity and 75%specificity) to predict cognitive decline at 2 years (74%accuracy at 3 years with 75%sensitivity and 73%specificity). When used to select trial participants, this tool yields a 3.8-fold decrease in the required sample size for a 2-year study (2.8-fold decrease for a 3-year study) for a hypothesized 25%treatment effect to reduce cognitive decline.

Conclusion:

When used in clinical trials for cohort enrichment, this tool could accelerate development of new treatments by significantly increasing statistical power to detect differences in cognitive decline between arms. In addition, detection of future decline can help clinicians improve patient management strategies that will slow or delay symptom progression.

Hippocampus Segmentation Method Based on Subspace Patch-Sparsity Clustering in Noisy Brain MRI

Since the hippocampus is of small size, low contrast, and irregular shape, a novel hippocampus segmentation method based on subspace patch-sparsity clustering in brain MRI is proposed to improve the segmentation accuracy, which requires that the representation coefficients in different subspaces should be as sparse as possible, while the representation coefficients in the same subspace should be as average as possible. By restraining the coefficient matrix with the patch-sparse constraint, the coefficient matrix contains a patch-sparse structure, which is helpful to the hippocampus segmentation. The experimental results show that our proposed method is effective in the noisy brain MRI data, which can well deal with hippocampus segmentation problem.

Long-term global and focal cerebral atrophy in perimesencephalic subarachnoid hemorrhage-a case-control study

PURPOSE

Non-aneurysmal perimesencephalic subarachnoid hemorrhage (PmSAH) represents 6.8% of spontaneous subarachnoid hemorrhage, and usually has a benign clinical course. However, patients might have early cerebral ischemic lesions and long-term neurocognitive complaints. Cerebral atrophy has been described in patients after aneurysmal SAH, but not in PmSAH. We aimed to investigate if PmSAH associates with increased brain volume loss.

METHODS

In this prospective study, we included consecutive patients with PmSAH that performed MR in the first 10 days after hemorrhage, and follow-up MR 6-7 years later. Automated volumetric measurements of intracranial, white matter, gray matter, whole brain, lateral ventricles, hippocampus, and amygdala volumes were performed. Volumes were compared to a normal population, matched for age.

RESULTS

Eight patients with PmSAH were included, with a mean age of 51.5 (SE 3.6) at baseline. The control group included 22 patients with a mean age of 56.3 (SE 2.0). A relative reduction of all volumes was found in both groups; however, PmSAH patients had significant reductions in intracranial, white and gray matter, whole brain, and hippocampal volumes when compared to controls. These changes had a higher magnitude in whole brain volume, with a significant absolute decrease of 6.5% in PmSAH patients (versus 1.9% in controls), and a trend for an increase in lateral ventricle volume (absolute 21.3% increase, versus 3.9% in controls).

CONCLUSION

Our cohort of PmSAH patients showed significant long-term parenchymal atrophy, and higher global and focal parenchymal volume loss rates when compared to a non-SAH population.

A robust discriminative multi-atlas label fusion method for hippocampus segmentation from MR image

Accurate and automatic segmentation of the hippocampus plays a vital role in the diagnosis and treatment of nervous system diseases. However, due to the anatomical variability of different subjects, the registered atlas images are not always perfectly aligned with the target image. This makes the segmentation of the hippocampus still face great challenges. In this paper, we propose a robust discriminative label fusion method under the multi-atlas framework. It is a patch embedding label fusion method based on conditional random field (CRF) model that integrates the metric learning and the graph cuts by an integrated formulation. Unlike most current label fusion methods with fixed (non-learning) distance metrics, a novel distance metric learning is presented to enhance discriminative observation and embed it into the unary potential function. In particular, Bayesian inference is utilized to extend a classic distance metric learning, in which large margin constraints are instead of pairwise constraints to obtain a more robust distance metric. And the pairwise homogeneity is fully considered in the spatial prior term based on classification labels and voxel intensity. The resulting integrated formulation is globally minimized by the efficient graph cuts algorithm. Further, sparse patch based method is utilized to polish the obtained segmentation results in label space. The proposed method is evaluated on IABA dataset and ADNI dataset for hippocampus segmentation. The Dice scores achieved by our method are 87.2%, 87.8%, 88.2% and 88.9% on left and right hippocampus on both two datasets, while the best Dice scores obtained by other methods are 86.0%, 86.9%, 86.8% and 88.0% on IABA dataset and ADNI dataset respectively. Experiments show that our approach achieves higher accuracy than state-of-the-art methods. We hope the proposed model can be transferred to combine with other promising distance measurement algorithms.

Whole brain and deep gray matter structure segmentation: Quantitative comparison between MPRAGE and MP2RAGE sequences

Objective

T1-weighted MRI images are commonly used for volumetric assessment of brain structures. Magnetization prepared 2 rapid gradient echo (MP2RAGE) sequence offers superior gray (GM) and white matter (WM) contrast. This study aimed to quantitatively assess the agreement of whole brain tissue and deep GM (DGM) volumes obtained from MP2RAGE compared to the widely used MP-RAGE sequence.

Methods

Twenty-nine healthy participants were included in this study. All subjects underwent a 3T MRI scan acquiring high-resolution 3D MP-RAGE and MP2RAGE images. Twelve participants were re-scanned after one year. The whole brain, as well as DGM segmentation, was performed using CAT12, volBrain, and FSL-FAST automatic segmentation tools based on the acquired images. Finally, contrast-to-noise ratio between WM and GM (CNR_WG), the agreement between the obtained tissue volumes, as well as scan-rescan variability of both sequences were explored.

Results

Significantly higher CNR_WG was detected in MP2RAGE vs. MP-RAGE (Mean ± SD = 0.97 ± 0.04 vs. 0.8 ± 0.1 respectively; p<0.0001). Significantly higher total brain GM, and lower cerebrospinal fluid‏ volumes were obtained from MP2RAGE vs. MP-RAGE based on all segmentation methods (p<0.05 in all cases). Whole-brain voxel-wise comparisons revealed higher GM tissue probability in the thalamus, putamen, caudate, lingual gyrus, and precentral gyrus based on MP2RAGE compared with MP-RAGE. Moreover, significantly higher WM probability was observed in the cerebellum, corpus callosum, and frontal-and-temporal regions in MP2RAGE vs. MP-RAGE. Finally, MP2RAGE showed a higher mean percentage of change in total brain GM compared to MP-RAGE. On the other hand, MP-RAGE demonstrated a higher overtime percentage of change in WM and DGM volumes compared to MP2RAGE.

Conclusions

Due to its higher CNR, MP2RAGE resulted in reproducible brain tissue segmentation, and thus is a recommended method for volumetric imaging biomarkers for the monitoring of neurological diseases.

Simultaneous brain structure segmentation in magnetic resonance images using deep convolutional neural networks

In brain magnetic resonance imaging (MRI) examinations, rapidly acquired two-dimensional (2D) T1-weighted sagittal slices are typically used to confirm brainstem atrophy and the presence of signals in the posterior pituitary gland. Image segmentation is essential for the automatic evaluation of chronological changes in the brainstem and pituitary gland. Thus, the purpose of our study was to use deep learning to automatically segment internal organs (brainstem, corpus callosum, pituitary, cerebrum, and cerebellum) in midsagittal slices of 2D T1-weighted images. Deep learning for the automatic segmentation of seven regions in the images was accomplished using two different methods: patch-based segmentation and semantic segmentation. The networks used for patch-based segmentation were AlexNet, GoogLeNet, and ResNet50, whereas semantic segmentation was accomplished using SegNet, VGG16-weighted SegNet, and U-Net. The precision and Jaccard index were calculated, and the extraction accuracy of the six convolutional network (DCNN) systems was evaluated. The highest precision (0.974) was obtained with the VGG16-weighted SegNet, and the lowest precision (0.506) was obtained with ResNet50. Based on the data, calculation times, and Jaccard indices obtained in this study, segmentation on a 2D image may be considered a viable and effective approach. We found that the optimal automatic segmentation of organs (brainstem, corpus callosum, pituitary, cerebrum, and cerebellum) on brain sagittal T1-weighted images could be achieved using SegNet with VGG16.

The role of lateralisation and sex on insular cortex: 3D volumetric analysis

Background/aim

The insula has attracted the attention of many neuroimaging studies because of its key role between brain structures. However, the number of studies investigating the effect of sex and laterality on insular volume is insufficient. The aim of this study was to investigate the differences in insular volume between sexes and hemispheres.

Materials and methods

A total of 47 healthy participants [24 males (20.08 ± 1.44 years) and 23 females (19.57 ± 0.90 years)] underwent magnetic resonance imaging (MRI). Imaging was performed using the 3T MRI scanner. The insular volume was measured using the Individual Brain Atlases using Statistical Parametric Mapping (IBASPM); total intracranial, cerebral, grey and white matter volumes were measured using volBrain.

Results

The right insular volume was significantly higher than the left insular volume in the participants, and the left cerebral volume was significantly higher than the right cerebral volume (p < 0.05). The total brain, total cerebral, left and right insular, and cerebral volumes were significantly larger in males than in females (p < 0.001). Also, the ratios of the insular volume to total brain and cerebral volume were significantly higher in males than in females (p < 0.05).

Conclusion

This study shows that insular volume differs with laterality and sex. This outcome may be explained by the anatomical relationship between the insula and behavioural functions and emotional reactions and the fact that the right side of the brain is best at expressive and creative tasks.

Impaired cerebral microcirculation in isolated REM sleep behaviour disorder

During the prodromal period of Parkinson's disease and other α-synucleinopathy-related parkinsonisms, neurodegeneration is thought to progressively affect deep brain nuclei, such as the locus coeruleus, caudal raphe nucleus, substantia nigra, and the forebrain nucleus basalis of Meynert. Besides their involvement in the regulation of mood, sleep, behaviour, and memory functions, these nuclei also innervate parenchymal arterioles and capillaries throughout the cortex, possibly to ensure that oxygen supplies are adjusted according to the needs of neural activity. The aim of this study was to examine whether patients with isolated REM sleep behaviour disorder, a parasomnia considered to be a prodromal phenotype of α-synucleinopathies, reveal microvascular flow disturbances consistent with disrupted central blood flow control. We applied dynamic susceptibility contrast MRI to characterize the microscopic distribution of cerebral blood flow in the cortex of 20 polysomnographic-confirmed patients with isolated REM sleep behaviour disorder (17 males, age range: 54-77 years) and 25 healthy matched controls (25 males, age range: 58-76 years). Patients and controls were cognitively tested by Montreal Cognitive Assessment and Mini Mental State Examination. Results revealed profound hypoperfusion and microvascular flow disturbances throughout the cortex in patients compared to controls. In patients, the microvascular flow disturbances were seen in cortical areas associated with language comprehension, visual processing and recognition and were associated with impaired cognitive performance. We conclude that cortical blood flow abnormalities, possibly related to impaired neurogenic control, are present in patients with isolated REM sleep behaviour disorder and associated with cognitive dysfunction. We hypothesize that pharmacological restoration of perivascular neurotransmitter levels could help maintain cognitive function in patients with this prodromal phenotype of parkinsonism.

Cortical diffusion kurtosis imaging and thalamic volume are associated with cognitive and walking performance in relapsing-remitting multiple sclerosis

BACKGROUND

In multiple sclerosis (MS), pronounced neurodegeneration manifests itself as cerebral gray matter (GM) atrophy, which is associated with cognitive and physical impairments. Microstructural changes in GM estimated by diffusion kurtosis imaging (DKI) may reveal neurodegeneration that is undetectable by conventional structural MRI and thus serve as a more sensitive marker of disease progression.

OBJECTIVE

The primary objective was to investigate the relationships between morphological and diffusional properties in cerebral GM and physical and cognitive performance in relapsing-remitting MS (RRMS) patients. A secondary objective was to investigate the relationship between GM microstructure and white matter (WM) injury, estimated by the volume of WM lesions.

METHODS

Sixty-seven RRMS patients performed the brief repeatable battery of neuropsychological tests (BRB-N), the 6-minute walk test (6MWT), the six spot step test (SSST), and underwent MRI scans using structural and DKI protocols. GM volumetrics and DKI measurements were analyzed in the cortex and deep GM structures using a general linear model with demographics, physical- and cognitive performance as covariates.

RESULTS

Mean diffusivity (MD) in the cortex was associated with the SSST, 6MWT, information processing, global cognitive performance, and volume of WM lesions. In addition, thalamic volume was associated with SSST (r² = 0.21, 6MWT (r² = 0.18), information processing (r² = 0.21), and WM lesion volume (r² = 0.60).

CONCLUSION

Cortical diffusion and thalamic volume are associated with walking and cognitive performance in RRMS patients and are highly affected by the presence of WM lesions.