Generalizability and quality control of deep learning-based 2D echocardiography segmentation models in a large clinical dataset.

Use of machine learning (ML) for automated annotation of heart structures from echocardiographic videos is an active research area, but understanding of comparative, generalizable performance among models is lacking. This study aimed to (1) assess the generalizability of five state-of-the-art ML-based echocardiography segmentation models within a large Geisinger clinical dataset, and (2) test the hypothesis that a quality control (QC) method based on segmentation uncertainty can further improve segmentation results. Five models were applied to 47,431 echocardiography studies that were independent from any training samples. Chamber volume and mass from model segmentations were compared to clinically-reported values. The median absolute errors (MAE) in left ventricular (LV) volumes and ejection fraction exhibited by all five models were comparable to reported inter-observer errors (IOE). MAE for left atrial volume and LV mass were similarly favorable to respective IOE for models trained for those tasks. A single model consistently exhibited the lowest MAE in all five clinically-reported measures. We leveraged the tenfold cross-validation training scheme of this best-performing model to quantify segmentation uncertainty. We observed that removing segmentations with high uncertainty from 14 to 71% studies reduced volume/mass MAE by 6-10%. The addition of convexity filters improved specificity, efficiently removing < 10% studies with large MAE (16-40%). In conclusion, five previously published echocardiography segmentation models generalized to a large, independent clinical dataset-segmenting one or multiple cardiac structures with overall accuracy comparable to manual analyses-with variable performance. Convexity-reinforced uncertainty QC efficiently improved segmentation performance and may further facilitate the translation of such models.

The Initial Gut Microbiota and Response to Antibiotic Perturbation Influence Clostridioides difficile Clearance in Mice.

The gut microbiota has a key role in determining susceptibility to Clostridioides difficile infections (CDIs). However, much of the mechanistic work examining CDIs in mouse models uses animals obtained from a single source. We treated mice from 6 sources (2 University of Michigan colonies and 4 commercial vendors) with clindamycin, followed by a C. difficile challenge, and then measured C. difficile colonization levels throughout the infection. The microbiota were profiled via 16S rRNA gene sequencing to examine the variation across sources and alterations due to clindamycin treatment and C. difficile challenge. While all mice were colonized 1 day postinfection, variation emerged from days 3 to 7 postinfection with animals from some sources colonized with C. difficile for longer and at higher levels. We identified bacteria that varied in relative abundance across sources and throughout the experiment. Some bacteria were consistently impacted by clindamycin treatment in all sources of mice, including Lachnospiraceae, Ruminococcaceae, and Enterobacteriaceae To identify bacteria that were most important to colonization regardless of the source, we created logistic regression models that successfully classified mice based on whether they cleared C. difficile by 7 days postinfection using community composition data at baseline, post-clindamycin treatment, and 1 day postinfection. With these models, we identified 4 bacterial taxa that were predictive of whether C. difficile cleared. They varied across sources (Bacteroides) or were altered by clindamycin (Porphyromonadaceae) or both (Enterobacteriaceae and Enterococcus). Allowing for microbiota variation across sources better emulates human interindividual variation and can help identify bacterial drivers of phenotypic variation in the context of CDIs.IMPORTANCEClostridioides difficile is a leading nosocomial infection. Although perturbation to the gut microbiota is an established risk, there is variation in who becomes asymptomatically colonized, develops an infection, or has adverse infection outcomes. Mouse models of C. difficile infection (CDI) are widely used to answer a variety of C. difficile pathogenesis questions. However, the interindividual variation between mice from the same breeding facility is less than what is observed in humans. Therefore, we challenged mice from 6 different breeding colonies with C. difficile We found that the starting microbial community structures and C. difficile persistence varied by the source of mice. Interestingly, a subset of the bacteria that varied across sources were associated with how long C. difficile was able to colonize. By increasing the interindividual diversity of the starting communities, we were able to better model human diversity. This provided a more nuanced perspective of C. difficile pathogenesis.

Prediction of mortality from 12-lead electrocardiogram voltage data using a deep neural network.

The electrocardiogram (ECG) is a widely used medical test, consisting of voltage versus time traces collected from surface recordings over the heart1. Here we hypothesized that a deep neural network (DNN) can predict an important future clinical event, 1-year all-cause mortality, from ECG voltage-time traces. By using ECGs collected over a 34-year period in a large regional health system, we trained a DNN with 1,169,662 12-lead resting ECGs obtained from 253,397 patients, in which 99,371 events occurred. The model achieved an area under the curve (AUC) of 0.88 on a held-out test set of 168,914 patients, in which 14,207 events occurred. Even within the large subset of patients (n = 45,285) with ECGs interpreted as 'normal' by a physician, the performance of the model in predicting 1-year mortality remained high (AUC = 0.85). A blinded survey of cardiologists demonstrated that many of the discriminating features of these normal ECGs were not apparent to expert reviewers. Finally, a Cox proportional-hazard model revealed a hazard ratio of 9.5 (P < 0.005) for the two predicted groups (dead versus alive 1 year after ECG) over a 25-year follow-up period. These results show that deep learning can add substantial prognostic information to the interpretation of 12-lead resting ECGs, even in cases that are interpreted as normal by physicians.

A Machine Learning Approach to Management of Heart Failure Populations.

BACKGROUND: Heart failure is a prevalent, costly disease for which new value-based payment models demand optimized population management strategies. OBJECTIVES: This study sought to generate a strategy for managing populations of patients with heart failure by leveraging large clinical datasets and machine learning. METHODS: Geisinger electronic health record data were used to train machine learning models to predict 1-year all-cause mortality in 26,971 patients with heart failure who underwent 276,819 clinical episodes. There were 26 clinical variables (demographics, laboratory test results, medications), 90 diagnostic codes, 41 electrocardiogram measurements and patterns, 44 echocardiographic measurements, and 8 evidence-based "care gaps": flu vaccine, blood pressure of <130/80 mm Hg, A1c of <8%, cardiac resynchronization therapy, and active medications (active angiotensin-converting enzyme inhibitor/angiotensin II receptor blocker/angiotensin receptor-neprilysin inhibitor, aldosterone receptor antagonist, hydralazine, and evidence-based beta-blocker) were collected. Care gaps represented actionable variables for which associations with all-cause mortality were modeled from retrospective data and then used to predict the benefit of prospective interventions in 13,238 currently living patients. RESULTS: Machine learning models achieved areas under the receiver-operating characteristic curve (AUCs) of 0.74 to 0.77 in a split-by-year training/test scheme, with the nonlinear XGBoost model (AUC: 0.77) outperforming linear logistic regression (AUC: 0.74). Out of 13,238 currently living patients, 2,844 were predicted to die within a year, and closing all care gaps was predicted to save 231 of these lives. Prioritizing patients for intervention by using the predicted reduction in 1-year mortality risk outperformed all other priority rankings (e.g., random selection or Seattle Heart Failure risk score). CONCLUSIONS: Machine learning can be used to priority-rank patients most likely to benefit from interventions to optimize evidence-based therapies. This approach may prove useful for optimizing heart failure population health management teams within value-based payment models.

Coding-Complete RNA Virus Genomes Assembled from Murine Cecal Metatranscriptomes.

Efforts to catalog viral diversity in the gut microbiome have largely focused on DNA viruses, while RNA viruses remain understudied. To address this, we screened assemblies of previously published mouse gut metatranscriptomes for the presence of RNA viruses. We identified the coding-complete genomes of an astrovirus and five mitovirus-like viruses.

Metatranscriptomic Analyses of Diel Metabolic Functions During a Microcystis Bloom in Western Lake Erie (United States).

This study examined diel shifts in metabolic functions of Microcystis spp. during a 48-h Lagrangian survey of a toxin-producing cyanobacterial bloom in western Lake Erie in the aftermath of the 2014 Toledo Water Crisis. Transcripts mapped to the genomes of recently sequenced lower Great Lakes Microcystis isolates showed distinct patterns of gene expression between samples collected across day (10:00 h, 16:00 h) and night (22:00 h, 04:00 h). Daytime transcripts were enriched in functions related to Photosystem II (e.g., psbA), nitrogen and phosphate acquisition, cell division (ftsHZ), heat shock response (dnaK, groEL), and uptake of inorganic carbon (rbc, bicA). Genes transcribed during nighttime included those involved in phycobilisome protein synthesis and Photosystem I core subunits. Hierarchical clustering and principal component analysis (PCA) showed a tightly clustered group of nighttime expressed genes, whereas daytime transcripts were separated from each other over the 48-h duration. Lack of uniform clustering within the daytime transcripts suggested that the partitioning of gene expression in Microcystis is dependent on both circadian regulation and physicochemical changes within the environment.

Corrigendum: Genome and Environmental Activity of a Chrysochromulina parva Virus and Its Virophages.

[This corrects the article DOI: 10.3389/fmicb.2019.00703.].

Genome and Environmental Activity of a Chrysochromulina parva Virus and Its Virophages.

Some giant viruses are ecological agents that are predicted to be involved in the top-down control of single-celled eukaryotic algae populations in aquatic ecosystems. Despite an increased interest in giant viruses since the discovery and characterization of Mimivirus and other viral giants, little is known about their physiology and ecology. In this study, we characterized the genome and functional potential of a giant virus that infects the freshwater haptophyte Chrysochromulina parva, originally isolated from Lake Ontario. This virus, CpV-BQ2, is a member of the nucleo-cytoplasmic large DNA virus (NCLDV) group and possesses a 437 kb genome encoding 503 ORFs with a GC content of 25%. Phylogenetic analyses of core NCLDV genes place CpV-BQ2 amongst the emerging group of algae-infecting Mimiviruses informally referred to as the "extended Mimiviridae," making it the first virus of this group to be isolated from a freshwater ecosystem. During genome analyses, we also captured and described the genomes of three distinct virophages that co-occurred with CpV-BQ2 and likely exploit CpV for their own replication. These virophages belong to the polinton-like viruses (PLV) group and encompass 19-23 predicted genes, including all of the core PLV genes as well as several genes implicated in genome modifications. We used the CpV-BQ2 and virophage reference sequences to recruit reads from available environmental metatranscriptomic data to estimate their activity in fresh waters. We observed moderate recruitment of both virus and virophage transcripts in samples obtained during Microcystis aeruginosa blooms in Lake Erie and Lake Tai, China in 2013, with a spike in activity in one sample. Virophage transcript abundance for two of the three isolates strongly correlated with that of the CpV-BQ2. Together, the results highlight the importance of giant viruses in the environment and establish a foundation for future research on the physiology and ecology CpV-BQ2 as a model system for algal Mimivirus dynamics in freshwaters.

Diversity of Active Viral Infections within the Sphagnum Microbiome.

Sphagnum-dominated peatlands play an important role in global carbon storage and represent significant sources of economic and ecological value. While recent efforts to describe microbial diversity and metabolic potential of the Sphagnum microbiome have demonstrated the importance of its microbial community, little is known about the viral constituents. We used metatranscriptomics to describe the diversity and activity of viruses infecting microbes within the Sphagnum peat bog. The vegetative portions of six Sphagnum plants were obtained from a peatland in northern Minnesota, and the total RNA was extracted and sequenced. Metatranscriptomes were assembled and contigs were screened for the presence of conserved virus marker genes. Using bacteriophage capsid protein gp23 as a marker for phage diversity, we identified 33 contigs representing undocumented phages that were active in the community at the time of sampling. Similarly, RNA-dependent RNA polymerase and the nucleocytoplasmic large DNA virus (NCLDV) major capsid protein were used as markers for single-stranded RNA (ssRNA) viruses and NCLDV, respectively. In total, 114 contigs were identified as originating from undescribed ssRNA viruses, 22 of which represent nearly complete genomes. An additional 64 contigs were identified as being from NCLDVs. Finally, 7 contigs were identified as putative virophage or polinton-like viruses. We developed co-occurrence networks with these markers in relation to the expression of potential-host housekeeping gene rpb1 to predict virus-host relationships, identifying 13 groups. Together, our approach offers new tools for the identification of virus diversity and interactions in understudied clades and suggests that viruses may play a considerable role in the ecology of the Sphagnum microbiome.IMPORTANCESphagnum-dominated peatlands play an important role in maintaining atmospheric carbon dioxide levels by modifying conditions in the surrounding soil to favor the growth of Sphagnum over that of other plant species. This lowers the rate of decomposition and facilitates the accumulation of fixed carbon in the form of partially decomposed biomass. The unique environment produced by Sphagnum enriches for the growth of a diverse microbial consortia that benefit from and support the moss's growth, while also maintaining the hostile soil conditions. While a growing body of research has begun to characterize the microbial groups that colonize Sphagnum, little is currently known about the ecological factors that constrain community structure and define ecosystem function. Top-down population control by viruses is almost completely undescribed. This study provides insight into the significant viral influence on the Sphagnum microbiome and identifies new potential model systems to study virus-host interactions in the peatland ecosystem.

Seasonal Gene Expression and the Ecophysiological Implications of Toxic Microcystis aeruginosa Blooms in Lake Taihu.

Harmful cyanobacterial blooms represent an increasing threat to freshwater resources globally. Despite increased research, the physiological basis of how the dominant bloom-forming cyanobacteria, Microcystis spp., proliferate and then maintain high population densities through changing environmental conditions is poorly understood. In this study, we examined the transcriptional profiles of the microbial community in Lake Taihu, China at 9 stations sampled monthly from June to October in 2014. To target Microcystis populations, we collected metatranscriptomic data and mapped reads to the M. aeruginosa NIES 843 genome. Our results revealed significant temporal gene expression patterns, with many genes separating into either early or late bloom clusters. About one-third of genes observed from M. aeruginosa were differentially expressed between these two clusters. Conductivity and nutrient availability appeared to be the environmental factors most strongly associated with these temporal gene expression shifts. Compared with the early bloom season (June and July), genes involved in N and P transport, energy metabolism, translation, and amino acid biosynthesis were down-regulated during the later season (August to October). In parallel, genes involved in regulatory functions as well as transposases and the production of microcystin and extracellular polysaccharides were up-regulated in the later season. Our observation indicates an eco-physiological shift occurs within the Microcystis spp. transcriptome as cells move from the rapid growth of early summer to bloom maintenance in late summer and autumn.

Molecular prediction of lytic vs lysogenic states for Microcystis phage: Metatranscriptomic evidence of lysogeny during large bloom events.

Microcystis aeruginosa is a freshwater bloom-forming cyanobacterium capable of producing the potent hepatotoxin, microcystin. Despite increased interest in this organism, little is known about the viruses that infect it and drive nutrient mobilization and transfer of genetic material between organisms. The genomic complement of sequenced phage suggests these viruses are capable of integrating into the host genome, though this activity has not been observed in the laboratory. While analyzing RNA-sequence data obtained from Microcystis blooms in Lake Tai (Taihu, China), we observed that a series of lysogeny-associated genes were highly expressed when genes involved in lytic infection were down-regulated. This pattern was consistent, though not always statistically significant, across multiple spatial and temporally distinct samples. For example, samples from Lake Tai (2014) showed a predominance of lytic virus activity from late July through October, while genes associated with lysogeny were strongly expressed in the early months (June-July) and toward the end of bloom season (October). Analyses of whole phage genome expression shows that transcription patterns are shared across sampling locations and that genes consistently clustered by co-expression into lytic and lysogenic groups. Expression of lytic-cycle associated genes was positively correlated to total dissolved nitrogen, ammonium concentration, and salinity. Lysogeny-associated gene expression was positively correlated with pH and total dissolved phosphorous. Our results suggest that lysogeny may be prevalent in Microcystis blooms and support the hypothesis that environmental conditions drive switching between temperate and lytic life cycles during bloom proliferation.

Thalamus segmentation using multi-modal feature classification: Validation and pilot study of an age-matched cohort.

Automatic segmentation of the thalamus can be used to measure differences and track changes in thalamic volume that may occur due to disease, injury or normal aging. An automatic thalamus segmentation algorithm incorporating features from diffusion tensor imaging (DTI) and thalamus priors constructed from multiple atlases is proposed. Multiple atlases with corresponding manual thalamus segmentations are registered to the target image and averaged to generate the thalamus prior. At each voxel in a region of interest around the thalamus, a multidimensional feature vector that includes the thalamus prior as well as a set of DTI features, including fractional anisotropy, mean diffusivity, and fiber orientation is formed. A random forest is trained to classify each voxel as belonging to the thalamus or background within the region of interest. Using a leave-one-out cross-validation on nine subjects, the proposed algorithm achieves a mean Dice score of 0.878 and 0.890 for the left and right thalami, respectively, which are higher Dice scores than the three state-of-art methods we compared to. We demonstrate the utility of the method with a pilot study exploring the difference in the thalamus fraction between 21 multiple sclerosis (MS) patients and 21 age-matched healthy controls. The left and right thalamic volumes (normalized by intracranial volumes) are larger in healthy controls by 7.6% and 7.3% respectively, compared to MS patients (though neither result is statistically significant).

Ecophysiological Examination of the Lake Erie Microcystis Bloom in 2014: Linkages between Biology and the Water Supply Shutdown of Toledo, OH.

Annual cyanobacterial blooms dominated by Microcystis have occurred in western Lake Erie (U.S./Canada) during summer months since 1995. The production of toxins by bloom-forming cyanobacteria can lead to drinking water crises, such as the one experienced by the city of Toledo in August of 2014, when the city was rendered without drinking water for >2 days. It is important to understand the conditions and environmental cues that were driving this specific bloom to provide a scientific framework for management of future bloom events. To this end, samples were collected and metatranscriptomes generated coincident with the collection of environmental metrics for eight sites located in the western basin of Lake Erie, including a station proximal to the water intake for the city of Toledo. These data were used to generate a basin-wide ecophysiological fingerprint of Lake Erie Microcystis populations in August 2014 for comparison to previous bloom communities. Our observations and analyses indicate that, at the time of sample collection, Microcystis populations were under dual nitrogen (N) and phosphorus (P) stress, as genes involved in scavenging of these nutrients were being actively transcribed. Targeted analysis of urea transport and hydrolysis suggests a potentially important role for exogenous urea as a nitrogen source during the 2014 event. Finally, simulation data suggest a wind event caused microcystin-rich water from Maumee Bay to be transported east along the southern shoreline past the Toledo water intake. Coupled with a significant cyanophage infection, these results reveal that a combination of biological and environmental factors led to the disruption of the Toledo water supply. This scenario was not atypical of reoccurring Lake Erie blooms and thus may reoccur in the future.

A Student's Guide to Giant Viruses Infecting Small Eukaryotes: From Acanthamoeba to Zooxanthellae.

The discovery of infectious particles that challenge conventional thoughts concerning "what is a virus" has led to the evolution a new field of study in the past decade. Here, we review knowledge and information concerning "giant viruses", with a focus not only on some of the best studied systems, but also provide an effort to illuminate systems yet to be better resolved. We conclude by demonstrating that there is an abundance of new host-virus systems that fall into this "giant" category, demonstrating that this field of inquiry presents great opportunities for future research.

Functional Characteristics of the Gut Microbiome in C57BL/6 Mice Differentially Susceptible to Plasmodium yoelii.

C57BL/6 mice are widely used for in vivo studies of immune function and metabolism in mammals. In a previous study, it was observed that when C57BL/6 mice purchased from different vendors were infected with Plasmodium yoelii, a causative agent of murine malaria, they exhibited both differential immune responses and significantly different parasite burdens: these patterns were reproducible when gut contents were transplanted into gnotobiotic mice. To gain insight into the mechanism of resistance, we removed whole ceca from mice purchased from two vendors, Taconic Biosciences (low parasitemia) and Charles River Laboratories (high parasitemia), to determine the combined host and microflora metabolome and metatranscriptome. With the exception of two Charles River samples, we observed ≥90% similarity in overall bacterial gene expression within vendors and ≤80% similarity between vendors. In total 33 bacterial genes were differentially expressed in Charles River mice (p-value < 0.05) relative to the mice purchased from Taconic. Included among these, fliC, ureABC, and six members of the nuo gene family were overrepresented in microbiomes susceptible to more severe malaria. Moreover, 38 mouse genes were differentially expressed in these purported genetically identical mice. Differentially expressed genes included basigin, a cell surface receptor required for P. falciparum invasion of red blood cells. Differences in metabolite pools were detected, though their relevance to malaria infection, microbial community activity, or host response is not yet understood. Our data have provided new targets that may connect gut microbial activity to malaria resistance and susceptibility phenotypes in the C57BL/6 model organism.

Thalamus parcellation using multi-modal feature classification and thalamic nuclei priors.

Segmentation of the thalamus and thalamic nuclei is useful to quantify volumetric changes from neurodegenerative diseases. Most thalamus segmentation algorithms only use T1-weighted magnetic resonance images and current thalamic parcellation methods require manual interaction. Smaller nuclei, such as the lateral and medial geniculates, are challenging to locate due to their small size. We propose an automated segmentation algorithm using a set of features derived from diffusion tensor image (DTI) and thalamic nuclei location priors. After extracting features, a hierarchical random forest classifier is trained to locate the thalamus. A second random forest classifies thalamus voxels as belonging to one of six thalamic nuclei classes. The proposed algorithm was tested using a leave-one-out cross validation scheme and compared with state-of-the-art algorithms. The proposed algorithm has a higher Dice score compared to other methods for the whole thalamus and several nuclei.

Effect of Clostridium perfringens infection and antibiotic administration on microbiota in the small intestine of broiler chickens.

The etiological agent of necrotic enteritis (NE) is Clostridium perfringens (CP), which is an economically significant problem for broiler chicken producers worldwide. Traditional use of in-feed antibiotic growth promoters to control NE disease have resulted in the emergence of antibiotic resistance in CP strains. Identification of probiotic bacteria strains as an alternative to antibiotics for the control of intestinal CP colonization is crucial. Two experiments were conducted to determine changes in intestinal bacterial assemblages in response to CP infection and in-feed bacitracin methylene disalicylate (BMD) in broiler chickens. In each experiment conducted in battery-cage or floor-pen housing, chicks were randomly assigned to the following treatment groups: 1) BMD-supplemented diet with no CP challenge (CM), 2) BMD-free control diet with no CP challenge (CX), 3) BMD-supplemented diet with CP challenge (PCM), or 4) BMD-free control diet with CP challenge (PCX). The establishment of CP infection was confirmed, with the treatment groups exposed to CP having a 1.5- to 2-fold higher CP levels (P < 0.05) compared to the non-exposed groups. Next-generation sequencing of PCR amplified 16S rRNA genes, was used to perform intestinal bacterial diversity analyses pre-challenge, and at 1, 7, and 21 d post-challenge. The results indicated that the intestinal bacterial assemblage was dominated by members of the phylum Firmicutes in all treatments before and after CP challenge, especially the Lactobacillaceae and Clostridiales families. In addition, we observed post-challenge emergence of members of the Enterobacteriaceae and Streptococcaceae in the non-medicated PCX treatment, and emergence of the Enterococcaceae in the medicated PCM treatment. This study highlights the bacterial interactions that could be important in suppressing or eliminating CP infection within the chicken intestine. Future studies should explore the potential to use commensal strains of unknown Clostridiales, Lactobacillaceae, Enterobacteriaceae, Streptococcaceae, and Enterococcaceae in effective probiotic formulations for the control of CP and NE disease.

Automatic method for thalamus parcellation using multi-modal feature classification.

Segmentation and parcellation of the thalamus is an important step in providing volumetric assessment of the impact of disease n brain structures. Conventionally, segmentation is carried out on T1-weighted magnetic resonance (MR) images and nuclear parcellation using diffusion weighted MR images. We present the first fully automatic method that incorporates both tissue contrasts and several derived fea-fractional anisotrophy, fiber orientation from the 5D Knutsson representation of the principal eigenvectors, and connectivity between the thalamus and the cortical lobes, as features. Combining these multiple information sources allows us to identify discriminating dimensions and thus parcellate the thalamic nuclei. A hierarchical random forest framework with a multidimensional feature per voxel, first distinguishes thalamus from background, and then separates each group of thalamic nuclei. Using a leave one out cross-validation on 12 subjects we have a mean Dice score of 0.805 and 0.799 for the left and right thalami, respectively. We also report overlap for the thalamic nuclear groups.