Search

The U-net is a deep-learning network model that has been used to solve a number of inverse problems. In this work, the concatenation of two-element U-nets, termed the W-net, operating in k-space (K) and image (I) domains, were evaluated for multi-channel magnetic resonance (MR) image reconstruction. The two-element network combinations were evaluated for the four possible image-k-space domain configurations: a) W-net II, b) W-net KK, c) W-net IK, and d) W-net KI. Selected four element (WW-nets) and six element (WWW-nets) networks were also examined. Two configurations of each network were compared: 1) each coil channel was processed independently, and 2) all channels were processed simultaneously. One hundred and eleven volumetric, T1-weighted, 12-channel coil k-space datasets were used in the experiments. Normalized root mean squared error, peak signal-to-noise ratio and visual information fidelity were used to assess the reconstructed images against the fully sampled reference images. Our results indicated that networks that operate solely in the image domain were better when independently processing individual channels of multi-channel data. Dual-domain methods were better when simultaneously reconstructing all channels of multi-channel data. In addition, the best cascade of U-nets performed better (p 0.01) than the previously published, state-of-the-art Deep Cascade and Hybrid Cascade models in three out of four experiments.

Finding a clinically useful neuroimaging biomarker that can predict treatment response in patients with major depressive disorder (MDD) is challenging, in part because of poor reproducibility and generalizability of findings across studies. Previous work has suggested that posterior hippocampal volumes in depressed patients may be associated with antidepressant treatment outcomes. The primary purpose of this investigation was to examine further whether posterior hippocampal volumes predict remission following antidepressant treatment. Magnetic resonance imaging (MRI) scans from 196 patients with MDD and 110 healthy participants were obtained as part of the first study in the Canadian Biomarker Integration Network in Depression program (CAN-BIND 1) in which patients were treated for 16 weeks with open-label medication. Hippocampal volumes were measured using both a manual segmentation protocol and FreeSurfer 6.0. Baseline hippocampal tail (Ht) volumes were significantly smaller in patients with depression compared to healthy participants. Larger baseline Ht volumes were positively associated with remission status at weeks 8 and 16. Participants who achieved early sustained remission had significantly greater Ht volumes compared to those who did not achieve remission by week 16. Ht volume is a prognostic biomarker for antidepressant treatment outcomes in patients with MDD.

Carotid-artery atherosclerosis (CA) contributes significantly to overall morbidity and mortality in ischemic stroke. We propose a machine learning technique to automatically identify subjects with CA from a heterogeneous cohort of magnetic resonance brain images. The cohort includes 190 subjects with CA, white mater hyperintensites of presumed vascular origin or multiple sclerosis, as well as 211 presumed healthy subjects. We determined a set of handcrafted and convolutional discriminant features to perform this task. A support vector machine (SVM) was used to perform this four-class classification task. Our approach had an accuracy rate of 97.5% (higher than chance accuracy of 52.6% for guessing majority class), sensitivity of 96.4% and specificity of 97.9% in identifying subjects with CA, suggesting that the proposed combination of features may be used as an imaging biomarker for characterizing atherosclerotic disease on brain imaging.

Subtle changes in hippocampal volumes may occur during both physiological and pathophysiological processes in the human brain. Assessing hippocampal volumes manually is a time-consuming procedure, however, creating a need for automated segmentation methods that are both fast and reliable over time. Segmentation algorithms that employ deep convolutional neural networks (CNN) have emerged as a promising solution for large longitudinal neuroimaging studies. However, for these novel algorithms to be useful in clinical studies, the accuracy and reproducibility should be established on independent datasets. Here, we evaluate the performance of a CNN-based hippocampal segmentation algorithm that was developed by Thyreau and colleagues – Hippodeep. We compared its segmentation outputs to manual segmentation and FreeSurfer 6.0 in a sample of 200 healthy participants scanned repeatedly at seven sites across Canada, as part of the Canadian Biomarker Integration Network in Depression consortium. The algorithm demonstrated high levels of stability and reproducibility of volumetric measures across all time points compared to the other two techniques. Although more rigorous testing in clinical populations is necessary, this approach holds promise as a viable option for tracking volumetric changes in longitudinal neuroimaging studies.

To describe the clinical response and side effects of radiofrequency treatment in patients with urinary incontinence after radical prostatectomy. This is a phase 1 clinical trial with 10 men up to 65 years of age who had urinary incontinence after radical prostatectomy, post void residual volume

The impact of HIV co-infection on the plasma immunological biomarker profile of HTLV-1 infected patients was evaluated. The plasma levels of leukotrienes and chemokines/cytokines were quantified by ELISA and Cytometric Bead Array. A total of 138 volunteers were enrolled and divided into two subgroups ("HTLV-1(+)HIV(-)" and "HTLV-1(+)(HIV(+)"), which were categorized according to the HTLV-1-associated neurological disease (AS, pHAM and HAM). Reference controls were BD and HIV mono-infected patients. HAM(+) exhibited higher CD4+ T-cell counts as compared to HIV+ mono-infected patients and lower HTLV-1 proviral load as compared to mono-infected HAM(-) patients. AS(+) exhibited higher levels of CysLT, CXCL8/IL-8 and lower levels of CCL5/RANTES as compared to AS(-). Increased levels of IL-6 and TNF with reduced levels of CXCL10/IP10 and CCL5/RANTES were observed in co-infected pHAM(+) as compared to mono-infected pHAM(-). HAM(+) patients revealed an increase in CXCL8/IL-8, CCL2/MCP-1, CXCL-10/IP-10, TNF and a decrease in IL-2 as compared to HAM(-) subgroup.

In patients with Mansoni schistosomiasis, it is fundamental to evaluate the disease morbidity, which is reflected by the severity of periportal fibrosis (PPF) and parameters of portal hypertension, as analyzed by ultrasonography (US). This study aimed to evaluate the morbidity of schistosomiasis by hepatic and splenic point shear-wave elastography (pSWE) and relate this to US parameters. The PPF pattern, the diameter of the portal and splenic veins and the size of the spleen were evaluated by US. Then, liver and spleen pSWEs were assessed in 74 patients using the same equipment. As the PPF pattern progressed, the splenic pSWE values significantly increased. Significant correlations between splenic pSWE, the longitudinal and transverse lengths of the spleen and the diameters of the portal and splenic veins were observed. These findings, however, were not observed through hepatic pSWE. In conclusion, the splenic pSWE has the potential for assessing morbidity in schistosomiasis mansoni.

This paper presents an open, multi-vendor, multi-field strength magnetic resonance (MR) T1-weighted volumetric brain imaging dataset, named Calgary-Campinas-359 (CC-359). The dataset is composed of images of older healthy adults (29–80 years) acquired on scanners from three vendors (Siemens, Philips and General Electric) at both 1.5 T and 3 T. CC-359 is comprised of 359 datasets, approximately 60 subjects per vendor and magnetic field strength. The dataset is approximately age and gender balanced, subject to the constraints of the available images. It provides consensus brain extraction masks for all volumes generated using supervised classification. Manual segmentation results for twelve randomly selected subjects performed by an expert are also provided. The CC-359 dataset allows investigation of 1) the influences of both vendor and magnetic field strength on quantitative analysis of brain MR; 2) parameter optimization for automatic segmentation methods; and potentially 3) machine learning classifiers with big data, specifically those based on deep learning methods, as these approaches require a large amount of data. To illustrate the utility of this dataset, we compared to the results of a supervised classifier, the results of eight publicly available skull stripping methods and one publicly available consensus algorithm. A linear mixed effects model analysis indicated that vendor ( p − v a l u e 0.001 ) and magnetic field strength ( p − v a l u e 0.001 ) have statistically significant impacts on skull stripping results.

Yogurt is one of the most popular and consumed foods worldwide, being widely used in gastronomy. This work aimed to evaluate how the addition of skim milk powder (SMP) and dairy cream (DC) can influence the fermentative process of milk and the physicochemical and sensory properties of concentrated Greek-style yogurt. The acidification profile of milk was evaluated by the decrease of pH values during the fermentation process. The quality of the yogurts was investigated by physicochemical, microbiological, and sensory analyses. The addition of SMP affected the fermentation process and resulted in yogurts with high protein content (6.38–6.61%) and low-fat content (1.37–1.38%) in relation to T0 (control), which presented 5.96% of proteins and 1.59% of lipids. The elaborated yogurts did not present contamination, demonstrating satisfactory hygienic-sanitary conditions sufficient on the processing which ensure their microbiological quality. The sensory attributes were influenced by the addition of SMP and DC, contributing to the choices made by the panelists. The yogurt with 10% of SMP presented the best contents of protein and lipids and was the most accepted by the panelists in the sensorial tests. Thus, it is possible to obtain high added-value yogurts with physicochemical and nutritional properties, and sensorial acceptance better than traditional yogurts from the concentration and addition of SMP and DC.

Training convolutional neural networks (CNNs) for medical image segmentation often requires large and representative sets of images and their corresponding annotations. Obtaining annotated images usually requires manual intervention, which is expensive and time consuming, as it typically requires a specialist. An alternative approach is to leverage existing automatic segmentation tools and combine them to create consensus-based "silver-standards" annotations. A drawback to this approach is that silver-standards are usually smooth and this smoothness is transmitted to the output segmentation of the network. Our proposal is to use a two-staged approach. First, silver-standard datasets are used to generate a large set of annotated images in order to train the brain extraction network from scratch. Second, fine-tuning is performed using much smaller amounts of manually annotated data so that the network can learn the finer details that are not preserved in the silver-standard data. As an example, our two-staged brain extraction approach has been shown to outperform seven state-of-the-art techniques across three different public datasets. Our results also suggest that CNNs can potentially capture inter-rater annotation variability between experts who annotate the same set of images following the same guidelines, and also adapt to different annotation guidelines.

Decreasing magnetic resonance (MR) image acquisition times can potentially reduce procedural cost and make MR examinations more accessible. Compressed sensing (CS)-based image reconstruction methods, for example, decrease MR acquisition time by reconstructing high-quality images from data that were originally sampled at rates inferior to the Nyquist-Shannon sampling theorem. In this work we propose a hybrid architecture that works both in the k-space (or frequency-domain) and the image (or spatial) domains. Our network is composed of a complex-valued residual U-net in the k-space domain, an inverse Fast Fourier Transform (iFFT) operation, and a real-valued U-net in the image domain. Our experiments demonstrated, using MR raw k-space data, that the proposed hybrid approach can potentially improve CS reconstruction compared to deep-learning networks that operate only in the image domain. In this study we compare our method with four previously published deep neural networks and examine their ability to reconstruct images that are subsequently used to generate regional volume estimates. We evaluated undersampling ratios of 75% and 80%. Our technique was ranked second in the quantitative analysis, but qualitative analysis indicated that our reconstruction performed the best in hard to reconstruct regions, such as the cerebellum. All images reconstructed with our method were successfully post-processed, and showed good volumetry agreement compared with the fully sampled reconstruction measures., Comment: 8 pages, 7 figures

Aim Alterations in limbic structures may be present before the onset of serious mental illness, but whether subfield-specific limbic brain changes parallel stages in clinical risk is unknown. To address this gap, we compared the hippocampus, amygdala, and thalamus subfield-specific volumes in adolescents at various stages of risk for mental illness. Methods MRI scans were obtained from 182 participants (aged 12-25 years) from the Canadian Psychiatric Risk and Outcome study. The sample comprised of four groups: asymptomatic youth at risk due to family history of mental illness (Stage 0, n = 32); youth with early symptoms of distress (Stage 1a, n = 41); youth with subthreshold psychotic symptoms (Stage 1b, n = 72); and healthy comparison participants with no family history of serious mental illness (n = 37). Analyses included between-group comparisons of brain measurements and correlational analyses that aimed to identify significant associations between neuroimaging and clinical measurements. A machine-learning technique examined the discriminative properties of the clinical staging model. Results Subfield-specific limbic volume deficits were detected at every stage of risk for mental illness. A machine-learning classifier identified volume deficits within the body of the hippocampus, left amygdala nuclei, and medial-lateral nuclei of the thalamus that were most informative in differentiating between risk stages. Conclusion Aberrant subfield-specific changes within the limbic system may serve as biological evidence to support transdiagnostic clinical staging in mental illness. Differential patterns of volume deficits characterize those at risk for mental illness and may be indicative of a risk-stage progression.

Enterobacter cloacae is a Gram-negative bacterium associated with high morbidity and mortality in intensive care patients due to its resistance to multiple antibiotics. Currently, therapy against multi-resistant bacteria consists of using colistin, in spite of its toxic effects at higher concentrations. In this context, colistin-resistant E. cloacae strains were challenged with lower levels of colistin combined with other antibiotics to reduce colistin-associated side effects. Colistin-resistant E. cloacae (ATCC 49141) strains were generated by serial propagation in subinhibitory colistin concentrations. After this, three colistin-resistant and three nonresistant replicates were isolated. The identity of all the strains was confirmed by MALDI-TOF MS, VITEK 2 and MicroScan analysis. Furthermore, cross-resistance to other antibiotics was checked by disk diffusion and automated systems. The synergistic effects of the combined use of colistin and chloramphenicol were observed via the broth microdilution checkerboard method. First, data here reported showed that all strains presented intrinsic resistance to penicillin, cephalosporin (except fourth generation), monobactam, and some associations of penicillin and β-lactamase inhibitors. Moreover, a chloramphenicol and colistin combination was capable of inhibiting the induced colistin-resistant strains as well as two colistin-resistant clinical strains. Furthermore, no cytotoxic effect was observed by using such concentrations. In summary, the data reported here showed for the first time the possible therapeutic use of colistin-chloramphenicol for infections caused by colistin-resistant E. cloacae.

Magnetic resonance (MR) as well as other imaging modalities have been used in a large number of clinical and research studies for the analysis and quantification of important structures and the detection of abnormalities. In this context, machine learning is playing an increasingly important role in the development of automated tools for aiding in image quantification, patient diagnosis and follow-up. Normally, these techniques require large, heterogeneous datasets to provide accurate and generalizable results. Large, multi-center studies, for example, can provide such data. Images acquired at different centers, however, can present varying characteristics due to differences in acquisition parameters, site procedures and scanners configuration. While variability in the dataset is required to develop robust, generalizable studies (i.e., independent of the acquisition parameters or center), like all studies there is also a need to ensure overall data quality by prospectively identifying and removing poor-quality data samples that should not be included, e.g., outliers. We wish to keep image samples that are representative of the underlying population (so called inliers), yet removing those samples that are not. We propose a framework to analyze data variability and identify samples that should be removed in order to have more representative, reliable and robust datasets. Our example case study is based on a public dataset containing T1-weighted volumetric head images data acquired at six different centers, using three different scanner vendors and at two commonly used magnetic fields strengths. We propose an algorithm for assessing data robustness and finding the optimal data for study occlusion (i.e., the data size that presents with lowest variability while maintaining generalizability (i.e., using samples from all sites)).

Manual annotation is considered to be the “gold standard” in medical imaging analysis. However, medical imaging datasets that include expert manual segmentation are scarce as this step is time-consuming, and therefore expensive. Moreover, single-rater manual annotation is most often used in data-driven approaches making the network biased to only that single expert. In this work, we propose a CNN for brain extraction in magnetic resonance (MR) imaging, that is fully trained with what we refer to as “silver standard” masks. Therefore, eliminating the cost associated with manual annotation. Silver standard masks are generated by forming the consensus from a set of eight, public, non-deep-learning-based brain extraction methods using the Simultaneous Truth and Performance Level Estimation (STAPLE) algorithm. Our method consists of 1) developing a dataset with “silver standard” masks as input, and implementing 2) a tri-planar method using parallel 2D U-Net-based convolutional neural networks (CNNs) (referred to as CONSNet). This term refers to our integrated approach, i.e., training with silver standard masks and using a 2D U-Net-based architecture. We conducted our analysis using three public datasets: the Calgary-Campinas-359 (CC-359), the LONI Probabilistic Brain Atlas (LPBA40), and the Open Access Series of Imaging Studies (OASIS). Five performance metrics were used in our experiments: Dice coefficient, sensitivity, specificity, Hausdorff distance, and symmetric surface-to-surface mean distance. Our results showed that we outperformed (i.e., larger Dice coefficients) the current state-of-the-art skull-stripping methods without using gold standard annotation for the CNNs training stage. CONSNet is the first deep learning approach that is fully trained using silver standard data and is, thus, more generalizable. Using these masks, we eliminate the cost of manual annotation, decreased inter-/intra-rater variability, and avoided CNN segmentation overfitting towards one specific manual annotation guideline that can occur when gold standard masks are used. Moreover, once trained, our method takes few seconds to process a typical brain image volume using modern a high-end GPU. In contrast, many of the other competitive methods have processing times in the order of minutes.

This Grand Challenge at MICCAI 2017 aims to directly compare methods for the automatic segmentation of White Matter Hyperintensities (WMH) of presumed vascular origin. Our method automatically segment WMH by using texture-based classification of pixels within the brain white matter. It uses no a priori information about the WMH size, contrast or location. The main goal is to compute the probability of each pixel being normal or WMH tissue, by generating a probability map. Based on this probability map, we can automatically segment the WMHs.

A better understanding of normal human brain aging is required to better study age-related neurodegeneration including cognitive impairment. We propose an automatic deep-learning method to analyze the predictive ability of magnetic resonance images with respect to age, sex and the presence of an age-related pathology (white matter hyperintensity, WMH). Experiments performed in a large dataset, containing 200 normal subjects, resulted in average accuracy rates to predict subject age (82.0%), sex (79.5%), and WMH occurrence (72.5%) when combining handcrafted texture and convolutional features. Positive and negative correlations between other extracted features and the subject characteristics (age, sex and WMH occurrence) were also observed. Even though human brain variability due to age, sex and WMH occurrence in structural magnetic resonance imaging may be subtle (and often not observable by human specialists), our results demonstrate that MR images alone contain relevant information that can better characterize the aging process and some demographic information of the population.

Atherosclerosis is a disease responsible for millions of deaths each year, primarily due to heart attack and stroke. Magnetic resonance (MR) imaging is a non-invasive method that can be used to analyze the carotid artery and detect signs of atherosclerosis. Most MR methods acquire high contrast, static images. These methods, however, are sensitive to artifacts from cardiac motion, produce time-averaged images, and do not allow for carotid distensibility analysis. Carotid distensibility is an important, systematic measure of vascular health. Cine fast spin echo (FSE) is a new MR imaging that can obtain dynamic MR data (i.e., cardiac phase-resolved datasets). Dynamic imaging, however, comes at the cost of lower spatial resolution and signal-to-noise ratio, making these data potentially more difficult to segment. This paper introduces a semi-automated segmentation method that segments the common carotid artery (CCA) lumen across the cardiac cycle from dynamic MR images. To the best of our knowledge, this work is the first proposed technique for segmenting cardiac cycle-resolved cine FSE images. It combines a priori knowledge about the size and shape of the CCA, with the max-tree data structure, the tie-zone watershed transform (using identified internal and external markers) and supervised classification, to segment the carotid artery wall-lumen boundary. The user has to select only a seed point (centred in the carotid artery lumen). Technique performance was assessed using forty-five cine FSE data sets, each consisting of images reconstructed at sixteen temporal bins across the cardiac cycle. The automatic segmentation results were compared against the consensus of three different manual segmentation results. Our technique achieved an average Dice coefficient, sensitivity and false positive rate of 0.928±0.031 (mean ± standard deviation), 0.915 ± 0.037 and 0.056 ± 0.049, respectively. Our method achieved higher agreement versus the consensus of the three manual segmentations than the individual manual segmentations versus the consensus.

The objective of this study is to determine the main neuroimaging findings of microcephalic newborns with possible Zika virus (ZIKV) intrauterine infection using transfontanellar cranial ultrasound.We performed a retrospective study to describe the main neuroimaging findings in newborns with microcephaly and possible association with congenital ZIKV infection. Microcephaly was defined in the postnatal period using transfontanellar cranial examination which was performed using both two- (2D) and three-dimensional (3D) ultrasound.One hundred and fifty newborns with microcephaly were identified during the study period. The mean ± (standard deviation - SD) of cephalic perimeter was 28.5 ± 4.2 cm (range, 25-38 cm). Transfontanellar neuroimaging patterns detected cerebral calcifications, neuronal migrational abnormalities, dysgenesis of the corpus callosum, and cerebellar atrophy in 34.9%, 31.1%, 26%, and 16.2%, respectively. Hydrocephalus was seen in 28% of overall newborns. A history of maculopapular rash was present in almost half of the mothers (46.1%).Neuroimaging patterns by means of transfontanellar ultrasound are accurate and diagnostic investigations of brain pathology in newborns affected by microcephaly and possible intrauterine ZIKV infection.

The bottleneck of convolutional neural networks (CNN) for medical imaging is the large number of annotated data required for training. Manual segmentation is considered to be the "gold-standard". However, medical imaging datasets with expert manual segmentation are scarce as this step is time-consuming and expensive. We propose in this work the use of what we refer to as silver standard masks for data augmentation in deep-learning-based skull-stripping also known as brain extraction. We generated the silver standard masks using the consensus algorithm Simultaneous Truth and Performance Level Estimation (STAPLE). We evaluated CNN models generated by the silver and gold standard masks. Then, we validated the silver standard masks for CNNs training in one dataset, and showed its generalization to two other datasets. Our results indicated that models generated with silver standard masks are comparable to models generated with gold standard masks and have better generalizability. Moreover, our results also indicate that silver standard masks could be used to augment the input dataset at training stage, reducing the need for manual segmentation at this step.

RATIONALE Total absence of superior vena cava (ASVC) is a very rare anomaly, and the patient usually suffers from superior vena cava syndrome (SVCS) or conduction disturbances. PATIENT CONCERNS We report a 36-year-old white male, born and living in Brazil, without comorbidities presented to hematologist thrombotic episodes even under anticoagulant therapy. On his first hematologic appointment, he had no active complaints except by the fullness after meals, and his physical examination presented remarkable collateral circulation in the chest. DIAGNOSES Congenital ASVC associated with factor V Leiden mutation. OUTCOMES In his magnetic resonance angiography of the thorax, a great amount of collateral circulation and communication of the azygos and hemiazygos veins with inferior vena cava were evident, as well as the absence of the upper cava vein. Furthermore, heterozygous genetic mutation was found for Leiden factor V. LESSONS This case gives us the lesson that we need to include ASVC in the differential diagnosis of SVCS. The importance of the V-Leiden factor as a joint risk with this congenital defect for venous thromboembolism episodes was also highlighted.

INTRODUCTION: Thyroid gland exposures to radiation induce nuclear chromosomal alteration. Objective: To evaluate the DNA content of thyroid gland submitted to radiation. MATERIALS AND METHODS: We radiated 75 rats while 25 were not radiated to be used as control group. Exposure was conducted by the use of Cobalt-60 radioactive source in the right anterior cervical region in a field of 3-30cm, comprising the second and the sixth tracheal rings with 600-centigray (cGY) doses. The DNA content (ploidy) was obtained with Feulgen-thionin stain and was quantified with CAS 200 quantitative measurement equipment. RESULTS: Diploid pattern was obtained in 88 cases (95.7%), independently of time of exposure: on the other hand, aneuploidy was observed in four cases (4.3%) only in the group sacrificed to the 33 days. Eight cases were excluded due to technical reasons. CONCLUSION: The early aneuploid pattern found in our study certainly corroborated that radiation affects thyroid gland with important consequences in terms of disorders. INTRODUÇÃO: A exposição da tireóide à irradiação está associada à alteração do componente cromossômico nessa glândula. OBJETIVO: Avaliar o conteúdo de DNA (ploidia) em glândula tireóide submetida à radiação. MATERIAL E MÉTODOS: Foram irradiadas tireóides de 75 ratos; 25 foram usados como grupo controle e não sofreram irradiação. A exposição à irradiação foi realizada com o uso de Cobalt-60 na região cervical anterior direita com espaço de 3-30cm, com anel traqueal de 2º e 6º e dose de 600 centigrays (cGY). O conteúdo de DNA (ploidia) foi obtido com o método de Feulgen-Thionin e quantificado com o aparelho CAS 200 de citometria estática. RESULTADO: Dos 92 casos, 88 (95,7%) foram diplóides, independente do tempo de exposição; aneuploidia foi observada em quatro casos (4,3%), somente no grupo sacrificado aos 33 dias. Oito casos foram excluídos por problemas técnicos. CONCLUSÃO: A irradiação afeta muito precocemente o conteúdo de DNA nuclear levando a aneuploidia.