Search

Brain extraction (masking of extra-cerebral tissues) and alignment are fundamental first steps of most neuroimage analysis pipelines. The lack of automated solutions for 3D ultrasound (US) has therefore limited its potential as a neuroimaging modality for studying fetal brain development using routinely acquired scans. In this work, we propose a convolutional neural network (CNN) that accurately and consistently aligns and extracts the fetal brain from minimally pre-processed 3D US scans. Our multi-task CNN, Brain Extraction and Alignment Network (BEAN), consists of two independent branches: 1) a fully-convolutional encoder-decoder branch for brain extraction of unaligned scans, and 2) a two-step regression-based branch for similarity alignment of the brain to a common coordinate space. BEAN was tested on 356 fetal head 3D scans spanning the gestational range of 14 to 30 weeks, significantly outperforming all current alternatives for fetal brain extraction and alignment. BEAN achieved state-of-the-art performance for both tasks, with a mean Dice Similarity Coefficient (DSC) of 0.94 for the brain extraction masks, and a mean DSC of 0.93 for the alignment of the target brain masks. The presented experimental results show that brain structures such as the thalamus, choroid plexus, cavum septum pellucidum, and Sylvian fissure, are consistently aligned throughout the dataset and remain clearly visible when the scans are averaged together. The BEAN implementation and related code can be found under www.github.com/felipemoser/kelluwen.

The quantification of subcortical volume development from 3D fetal ultrasound can provide important diagnostic information during pregnancy monitoring. However, manual segmentation of subcortical structures in ultrasound volumes is time-consuming and challenging due to low soft tissue contrast, speckle and shadowing artifacts. For this reason, we developed a convolutional neural network (CNN) for the automated segmentation of the choroid plexus (CP), lateral posterior ventricle horns (LPVH), cavum septum pellucidum et vergae (CSPV), and cerebellum (CB) from 3D ultrasound. As ground-truth labels are scarce and expensive to obtain, we applied few-shot learning, in which only a small number of manual annotations (n = 9) are used to train a CNN. We compared training a CNN with only a few individually annotated volumes versus many weakly labelled volumes obtained from atlas-based segmentations. This showed that segmentation performance close to intra-observer variability can be obtained with only a handful of manual annotations. Finally, the trained models were applied to a large number (n = 278) of ultrasound image volumes of a diverse, healthy population, obtaining novel US-specific growth curves of the respective structures during the second trimester of gestation.

Longevity has important impact on lifetime daily milk yields and therefore ecological and economic efficiency of dairy cows. However, actual productive lifespan of dairy cows in Europe is less than four lactations and thus far below the physiological potential. This holds also true for cows in organic systems. Besides management tools, selection criteria at the farmers’ hands are needed to improve longevity on farm level. We used data of culled cows from Swiss organic dairy farms to develop prediction models for the number of lactations completed. A dataset comprising 298 cows, which had been in 1st lactation in 2009, and were culled before June 2019, served to fit 32 prediction models initially based on lactation curve parameters (LCP), somatic cell count, age at first calving, calving interval and veterinary events against the binary trait of completing minimum four lactations. The lowest tercile regarding 1st lactation milk yield was excluded in order to separate economic culling decisions from biological reasons. Full and reduced generalised linear models (GLM), generalised mixed effect models, and linear discriminant analyses, respectively with and without inclusion of veterinary events and LCP were applied; all models were based on 1st lactation as well as on 2nd lactation data. In order to rank the models, the resulting scores were applied to a dataset of 7,742 culled cows for predicting their total number of completed lactations as well as lifetime milk yield. Predictability was higher when based on 2nd-compared to 1st lactation models, and it clearly increased if LCP were excluded. The best prediction for productive lifespan with Rho=14% was reached with a GLM model, excluding LCP. Thus, overall predictability of the best models was still poor. However, for lactation numbers above seven and lifetime yields above 50 tons the rate of cows being false classified by low scores decreased to less than 10%. We conclude that the chosen approach needs to be optimised by larger data for the model development, but might become suitable to support longevity-targeted culling decisions in low-input systems, reliably identifying the poorest percentile.

Significance Lassa virus is a highly pathogenic arenavirus that causes severe hemorrhagic fever in humans. Currently, there are no efficacious vaccines or treatments available to combat this pathogen. An important component of any vaccine candidate against Lassa virus will likely include the highly glycosylated glycoprotein complex presented on the virion surface. Here, we determine the composition of the Lassa virus glycome, revealing that the virus presents an abundance of glycans that are not biosynthetically processed to full maturity. Such underprocessed glycans form spatially distinct clusters, which shield the proteinous surface of the Lassa virus glycoprotein spike from the humoral immune response. These data are integral for the development of humoral-based vaccines that mimic the mature Lassa virion., Lassa virus is an Old World arenavirus endemic to West Africa that causes severe hemorrhagic fever. Vaccine development has focused on the envelope glycoprotein complex (GPC) that extends from the virion envelope. The often inadequate antibody immune response elicited by both vaccine and natural infection has been, in part, attributed to the abundance of N-linked glycosylation on the GPC. Here, using a virus-like−particle system that presents Lassa virus GPC in a native-like context, we determine the composite population of each of the N-linked glycosylation sites presented on the trimeric GPC spike. Our analysis reveals the presence of underprocessed oligomannose-type glycans, which form punctuated clusters that obscure the proteinous surface of both the GP1 attachment and GP2 fusion glycoprotein subunits of the Lassa virus GPC. These oligomannose clusters are seemingly derived as a result of sterically reduced accessibility to glycan processing enzymes, and limited amino acid diversification around these sites supports their role protecting against the humoral immune response. Combined, our data provide a structure-based blueprint for understanding how glycans render the glycoprotein spikes of Lassa virus and other Old World arenaviruses immunologically resistant targets.