Diagnosis of normal chest radiographs using an autonomous deep-learning algorithm

AIM

To evaluate the suitability of a deep-learning (DL) algorithm for identifying normality as a rule-out test for fully automated diagnosis in frontal adult chest radiographs (CXR) in an active clinical pathway.

MATERIALS AND METHODS

This multicentre study included 3,887 CXRs from four distinct NHS institutions. A convolutional neural network (CNN) was developed and trained prior to this study and was used to classify a subset of examinations with the lowest abnormality scores as high confidence normal (HCN). For each radiograph, the ground truth (GT) was established using two independent reviewers and an arbitrator in case of discrepancy.

RESULTS

The DL algorithm was able to classify 15% of all examinations as HCN, with a corresponding precision of 97.7%. There were 0.33% of examinations classified incorrectly as HCN, with 84.6% of these examinations identified as borderline cases by the radiologist GT process.

CONCLUSION

A DL algorithm can achieve a high level of precision as a fully automated diagnostic tool for reporting a subset of CXRs as normal. The removal of 15% of all CXRs has the potential to significantly reduce workload and focus radiology resources on more complex examinations. To optimise performance, site-specific deployment of algorithms should occur with robust feedback mechanisms for incorrect classifications.

Combined genome-wide linkage and targeted association analysis of head circumference in autism spectrum disorder families

Background

It has long been recognized that there is an association between enlarged head circumference (HC) and autism spectrum disorder (ASD), but the genetics of HC in ASD is not well understood. In order to investigate the genetic underpinning of HC in ASD, we undertook a genome-wide linkage study of HC followed by linkage signal targeted association among a sample of 67 extended pedigrees with ASD.

Methods

HC measurements on members of 67 multiplex ASD extended pedigrees were used as a quantitative trait in a genome-wide linkage analysis. The Illumina 6K SNP linkage panel was used, and analyses were carried out using the SOLAR implemented variance components model. Loci identified in this way formed the target for subsequent association analysis using the Illumina OmniExpress chip and imputed genotypes. A modification of the qTDT was used as implemented in SOLAR.

Results

We identified a linkage signal spanning 6p21.31 to 6p22.2 (maximum LOD = 3.4). Although targeted association did not find evidence of association with any SNP overall, in one family with the strongest evidence of linkage, there was evidence for association (rs17586672, p = 1.72E−07).

Conclusions

Although this region does not overlap with ASD linkage signals in these same samples, it has been associated with other psychiatric risk, including ADHD, developmental dyslexia, schizophrenia, specific language impairment, and juvenile bipolar disorder. The genome-wide significant linkage signal represents the first reported observation of a potential quantitative trait locus for HC in ASD and may be relevant in the context of complex multivariate risk likely leading to ASD.

Electronic supplementary material

The online version of this article (doi:10.1186/s11689-017-9187-8) contains supplementary material, which is available to authorized users.

Annealing Stability Study of Co20Fe60B20\MgO\ Co20Fe60B20 Perpendicular Magnetic Tunnel Junctions

A full Co20Fe60B20\MgO\ Co20Fe60B20 perpendicular magnetic tunnel junction (pMTJ) with (Co\Pt) multilayers as pinning layers and different functional multilayers stacks were made and annealed at different temperatures. The tunneling magnetoresistance ratio (TMR) and MgO barrier resistance-area product (RA) were measured and analyzed as a function of annealing temperature. The TMR of pMTJs dramatically declines with increasing annealing temperatures from 320 °C to 400 °C while the RA increases with temperature from 375 °C to 450 °C. The pMTJs and partial stacks were also measured in a vibrating sample magnetometer (VSM). We found that the (Co\Pt) multilayers are very stable and maintain a magnetization direction perpendicular to the film plane up to 450 °C. However, the magnetization direction of the CoFeB above and below the MgO barrier rotates from perpendicular to in-plane with increasing annealing temperature. Furthermore, the CoFeB layer influences the adjacent (Co\Pt) layers to rotate at the same time. The pMTJs' elemental depth profiles in the as deposited and annealed states were determined by Secondary Ion Mass Spectrometry (SIMS). We found that Boron and Tantalum migrate towards the top of the stack. The other elements (Platinum, Cobalt, Ruthenium, and Magnesium) are very stable and do not interdiffuse during annealing up to 450°C.

The self-reported oral health status and dental attendance of smokers and non-smokers

AIM

To report the oral health status and dental attendance of smokers and non-smokers.

METHODS

A postal survey enquiring about smoking status, stop smoking advice, dental attendance and perceptions of oral health was conducted in Yorkshire and the Humber, U.K., in 2008. To address potential biases data were weighted to account for variations in gender, age and deprivation. Data were analysed using descriptive statistics, chi-square tests and binary logistic regression.

RESULTS

A response rate of 43.1% was achieved (n=10,864). Across all deprivation quintiles, smokers (17.5% of respondents) were more likely than non-smokers to report fair, poor or very poor oral health (p<0.001). Smokers in the least deprived areas were more likely than non-smokers to attend the dentist symptomatically (p<0.001). Advice to quit was most frequently gained from GP services followed by NHS Stop Smoking Services and dental teams.

CONCLUSIONS

Smokers were more likely than non-smokers to have a poor self-rated oral health status and attend the dentist symptomatically, irrespective of deprivation.

OS049. Exome sequencing identifies likely functional variantsinfluencing preeclampsia and CVD risk

INTRODUCTION

Next-generation sequencing (NGS) in family-based study designs will be pivotal in unlocking the missing heritability of common complex diseases. Whilst our prior linkage- and association-based positional cloning studies in family- and population-based Australian cohorts, respectively, have discovered novel preeclampsia candidate genes (INHBB,ACVR2A,LCT,LRP1B,RND3,GCA,ERAP2,TNFSF13B), the full complement of causal genetic variation remains largely unknown. We have now sequenced the exomes of two Australian preeclampsia families in another step forward to unlocking preeclampsia's complex allelic architecture.

OBJECTIVES

Identify family-specific exon-centric loci segregating in preeclamptic women only.

METHODS

The exomes of 18 women (7 preeclamptics,11 controls) from two Australian families contributing to our chromosome 5q (Family 1) and 13q (Family 2) susceptibility loci, respectively, were sequenced using Illumina's TruSeq Exome Enrichment assay and NGS technology. Sequence alignments, quality control assessment and variant calling were conducted on our 8000 parallel processor compute server, MEDUSA. As a first pass, we prioritized exome sequence data to non-synonymous variants within the 1-LOD drop intervals of our 5q and 13q loci. Prioritized exonic variants were also genotyped in the Western Australian Pregnancy (Raine) Cohort to assess their significance against a plethora of cardiovascular disease (CVD) related traits.

RESULTS

In Family 1 we identified two missense SNPs and in Family 2 we identified one missense SNP to segregate in the preeclamptic women but not in the unaffected women. The first SNP in Family 1 (rs62375061) resides within the LYSMD3 gene, is predicted to "possibly" damage the focal protein and the only public record of this SNP is within the Watson genome. The second SNP in Family 1 (rs111033530) resides within the GPR98 gene, is predicted to "probably" damage the focal protein and is rare (1.7% population prevalence). The SNP in Family 2 (rs1805388) resides within the LIG4 gene, is predicted to be highly deleterious (F-SNP FSS=0.849) and is common (⩾17% population prevalence). In the Raine cohort the LIG4 SNP was also significantly associated with weight (p=0.0085), total cholesterol (p=0.0007), HDL cholesterol (p=0.0067) and LDL cholesterol (p=0.0324).

CONCLUSION

Our preliminary exome data documents the substantial potential to rapidly identify likely functional variants that influence preeclampsia risk. The GPR98 finding is of major interest to us as a recent genome-wide association study reported a significant association with diastolic blood pressure for a SNP at this same gene locus. Furthermore, our findings implicate LIG4 as a novel candidate susceptibility gene for CVD and add weight to the hypothesis of shared genetic risk factors for preeclampsia and CVD.

OS046. Genome-wide association scans identify novel maternalsusceptibility loci for preeclampsia

INTRODUCTION

We have successfully utilized a family-based study design to localize several positional candidate preeclampsia susceptibility genes to chromosomes 2q22(ACVR2A,LCT,LRP1B,RND3,GCA),5q (ERAP2) and 13q(TNFSF13B). We now report on our continued positional cloning efforts using an alternative genome-wide association (GWA) mapping strategy in large Caucasian case-control cohorts from Australia and Norway.

OBJECTIVES

To identify maternal genetic risk loci for preeclampsia.

METHODS

The unrelated Australian samples (545 cases,547 controls) were genotyped using Illumina BeadChip technology (700K loci) and have been analyzed using PLINK. All unrelated Norwegian samples were genotyped across several Illumina BeadChip substrates and consist of 847 cases (700K loci) and 638 controls. The Norwegian control samples originate from other HUNT studies pertaining to migraine (n=95,700K loci), lung cancer (n=89,370K loci) and normal brain pathology (n=454,2.5M loci). To analyze a concordant set of 2.5-3 million genotypes across all Norwegian samples we are currently using MaCH to impute those loci not directly genotyped. The Norwegian GWA data will be analyzed in SOLAR utilizing empirical kinship estimates to account for any distant relatedness.

RESULTS

1078 Australian samples (538 cases,540 controls) and 648, 175 SNPs passed our quality control metrics. Two SNP associations (rs7579169,p=3.6×10(-7); rs12711941,p=4.3×10(-7)) satisfied our genome-wide significant threshold (p<5.1×10(-7)). These SNPs reside less than 15kb downstream from the 3 terminus of the Inhibin, beta B (INHBB) gene on 2q14.2. Sequencing of the INHBB locus in our patient cohort identified a third intergenic SNP to significantly associate with preeclampsia (rs7576192,p=1.5×10(-7)). These three SNPs confer risk (OR>1.56) and are in strong linkage disequilibrium with each other (r(2)>0.9) but not with any other genotyped SNP ±200kb. The analysis of the Norwegian GWAS is underway.

CONCLUSION

The Australian GWAS has identified a novel preeclampsia risk locus on chromosome 2q. The INHBB gene closest to our SNP associations is a plausible positional candidate susceptibility gene. There is a substantive body of evidence implicating inhibins, activins and other members of the TGF-βsuperfamily to have a role in the development of preeclampsia. The biological connection between ACVR2A and INHBB leads us to speculate that our linkage-based and GWA-based study designs, respectively, have identified a key biological pathway involved in susceptibility to preeclampsia.

Genetics of adult body mass and maintenance of adult body mass in captive baboons (Papio hamadryas subspecies)

Adult body mass and changes in mass during an individual's life are important indicators of general health and reproductive fitness. Therefore, characterization of the factors that influence normal variation in body mass has important implications for colony management and husbandry. The main objective of this study was to quantify the genetic contribution to adult body mass and its maintenance in baboons. Intra-individual mean and variance in body mass were calculated from multiple weight measures available for each of 1,614 animals at least 10 years of age. Heritabilities were estimated using maximum likelihood methods. Mean adult body mass had a significant heritability (50%) as did variance in adult body mass (12%). The sexes differed in several respects: on average females were smaller than males and had greater variability in adult body mass; mean and variance in body mass increased with age in females only; and number of offspring showed a significant positive relationship with body mass in females only. There were significant differences between subspecies in body mass as well as ability to maintain body mass. These results indicate that there is a significant genetic influence on body mass and its maintenance, and suggest that different factors influence changes in body mass with age as well as body mass maintenance in male and female baboons.

Respiratory infection and coronary heart disease: progression of a paradigm

We have developed a previously published paradigm concerning causation of coronary heart disease, based on the probability that the fundamental cause is a microbe, probably Chlamydia pneumoniae, and that the progress of the disease is influenced by number of accelerating and inhibiting factors. We propose that cigarette smoking acts via respiratory infection, this itself being influenced by immunocompetence resulting from sunlight exposure. We also propose an immuno-enhancing effect of oestrogen and an anti-inflammatory effect of statin therapy. In respect of the geographical variation of coronary heart disease, we emphasize that this must be viewed as part of the bigger picture of a high mortality from all causes in countries of North-west Europe that have a particularly low level of sunlight exposure. Finally, we draw attention to the Albanian, French, Italian, Northern Ireland and Scottish paradoxes which should lead to a major review of the conventional wisdom concerning the aetiogenesis of coronary heart disease.

A prioritisation system for elective coronary angiography

OBJECTIVE

To devise a clinical prioritisation system for rationing access to a cardiac catheter waiting list and to describe its performance at predicting angiographic findings and selecting patients for angioplasty or coronary artery bypass graft surgery.

SETTING

Tertiary level cardiology centre.

METHODS

(1) 665 consecutive patients on an elective waiting list for coronary angiography were scored using a system derived from established clinical criteria for selecting patients for coronary surgery (New Zealand/Duke). The scores were compared with clinical outcome (referral for surgery, angioplasty, or medical management). (2) In a subset of 125 patients, scores derived from clinical criteria and exercise testing were compared with findings on coronary angiography. (3) Multivariate analysis was used in a new group of 178 patients to identify factors that would be better predictors of the angiographic score. (4) A new scoring system was devised based partly on the results of the multivariate analysis. It was applied to a new test group of 100 patients using clinical outcome and angiographic score as outcome measures.

RESULTS

(1) Using the established clinical score, similar proportions of patients were referred after angiography for medical management, angioplasty, or coronary bypass grafting, irrespective of their original score. The exceptions were patients with a score < 20, who were more likely to continue medical management. (2) There was poor correlation (r = 0.05) between the clinical score and the subsequent angiographic score. (3) Multivariate analysis identified age, male sex, previous myocardial infarction, high cholesterol, and diabetes as independent predictors of coronary score. (4) The modified scoring system, incorporating the predictors identified by multivariate analysis, performed better than the original scoring system in predicting coronary score when both were tested, but some patients had severe disease despite a low score.

CONCLUSIONS

Patients can be ranked using clinical and non-invasive criteria, and a rationing system implemented on this basis. With prioritisation by noninvasive criteria, the risk of missing serious coronary disease in patients with relatively mild symptoms must be accepted; this risk becomes greater the more stringently rationing is applied.

Ontogenetic changes in genetic regulation of fetal morphometrics in baboons (Papio hamadryas subspp.)

It is known that different genes are expressed during ontogeny; however, it is unclear how variation in that expression is associated with changes in growth patterns. The objective of this study is to assess how genetic variation in fetal morphology changes with ontogeny in baboons. Longitudinal measures of the head and femur (60 to 180 days gestation) were available for 892 pregnancies. We used a genetic model that allowed both the genetic and environmental variances (sigma 2G and sigma 2E) to change with age and estimated genetic and environmental correlations (rho G and rho E) between measurements at different ages. The results indicate a significant increase in the genetic variance for biparietal diameter and femur length but not for head circumference and fronto-occipital diameter. The rho G estimates for all measures decreased as the age between measures increased from 0 to 120 days, indicating that different groups of genes are expressed early in gestation and late in gestation. The rho E estimates dropped rapidly from 1 to 0 for all measures, indicating temporally localized environmental influences on fetal growth. Thus fetal morphometrics are significantly heritable and those genes that influence them show age-specific expression during ontogeny.

Sunlight, cholesterol and coronary heart disease

We investigated the relationship between geography and incidence of coronary heart disease, looking at deficiency of sunlight and thus of vitamin D as a factor that might influence susceptibility and thus disease incidence. Sunlight deficiency could increase blood cholesterol by allowing squalene metabolism to progress to cholesterol synthesis rather than to vitamin D synthesis as would occur with greater amounts of sunlight exposure, and the increased concentration of blood cholesterol during the winter months, confirmed in this study, may well be due to reduced sunlight exposure. We show evidence that outdoor activity (gardening) is associated with a lower concentration of blood cholesterol in the summer but not in the winter. We suggest that the geographical variation of coronary heart disease is not specific, but is seen in other diseases and sunlight influences susceptibility to a number of chronic diseases, of which coronary heart disease is one.

Interspecies cross-reactivity of a monoclonal antibody directed against wheat chloroplast fructose-1,6-bisphosphatase

The primary structure of several chloroplast fructose-1,6-bisphosphatase (CFBPase) was deduced from DNA sequences, but only spinach, pea and rapeseed enzymes have been characterized structurally. We analyzed whether CFBPases from different phylogenetic origin contain a common epitope. To this end a DNA fragment of 1200 base pairs encoding 338 amino acid residues of wheat CFBPase (38 kDa) was cloned in the expression plasmid pGEX-1 in frame with the gene coding for glutathione S-transferase (GT) of Schistosoma japonicun (26.5 kDa). Upon transformation of Escherichia coli and induction with isopropyl-beta-D-thiogalactopyranoside, centrifugation of the lysate partitioned 10% of the fusion protein in the supernatant fraction and the remaining 90% in the precipitate. The expected 65 kDa protein was purified from both the soluble and the particulate fraction by affinity chromatography on columns of glutathione-agarose. This fusion protein was successfully used to produce a monoclonal antibody that specifically recognized the CFBPase region of the fusion protein but not the GT moiety. Moreover, the monoclonal antibody immunoreacted not only with polypeptides (ca. 40 kDa) present in leaf crude extracts of other varieties of wheat (Triticum spelta, T. aestivum and T. durum), but also with homogeneous preparations of the spinach (Spinacia oleracea) and rapeseed (Brassica napus) enzymes. Thus, the cross reaction of this monoclonal antibody with counterparts from different plant species indicates the persistency of a common epitope through biological evolution.

Pseudohyperkalaemia associated with hereditary spherocytosis in four members of a family

Pseudohyperkalaemia was detected in four members of a family all of whom have hereditary spherocytosis with normal white blood cells and platelets counts. The degree of pseudohyperkalaemia was related to the time between sampling and cell separation, and inversely related to the temperature in which the sample was left to stand before cell separation. A fifth member of this family was free from both conditions. The association suggests linkage at a membrane level.

Chloroplast fructose-1,6-bisphosphatase: the product of a mosaic gene

We show here that light stimulates the expression of nuclear genes in wheat leaves for chloroplast fructose-1,6-bisphosphatase (FBPase) and describe a sequence of amino acids in this enzyme which may be responsible, via thioredoxin, for the light regulation of its activity. This data results from (a) our isolation and characterization of a cDNA of this enzyme which contains its entire coding sequence, and (b) our use of this cDNA as a probe to detect mRNA levels in wheat plants subjected to different light regimes. The similarity in amino acid sequence of the encoded enzyme from diverse sources suggests that the FBPase genes all had a common origin. However, their control sequences have been adjusted so that they are appropriately expressed and their coding sequences modified so that the enzymic activity of their products are suitably regulated in the particular cellular environment in which they must function. The light-activated regulatory sequences in the gene for the chloroplast protein have probably come together by a shuffling of DNA segments.

Chloroplast transfer RNAs and tRNA genes of wheat

Fractionation (by two-dimensional polyacrylamide gel electrophoresis) of total tRNA from wheat chloroplasts yields about 33 RNA spots. Of these, 30 have been identified by aminoacylation as containing tRNAs specific for 17 amino acids.Hybridization of labeled individual tRNAs to cloned chloroplast DNA fragments has revealed the location of at least nine pairs of tRNA genes in the segments of the inverted repeat, at least twelve tRNA genes in the large single copy region and one tRNA gene in the small single copy region.A comparison of this wheat chloroplast tRNA gene map to that of maize and of other higher plants suggests that gene rearrangements have occurred during evolution, even within cereal chloroplast DNA. These rearrangements have taken place within the inverted repeat, within the large single copy region and between the inverted repeat and the large single copy region.