Graph-Based Multi-Surface Segmentation of OCT Data Using Trained Hard and Soft Constraints

Optical coherence tomography (OCT) is a well-established image modality in ophthalmology and used daily in the clinic. Automatic evaluation of such datasets requires an accurate segmentation of the retinal cell layers. However, due to the naturally low signal to noise ratio and the resulting bad image quality, this task remains challenging. We propose an automatic graph-based multi-surface segmentation algorithm that internally uses soft constraints to add prior information from a learned model. This improves the accuracy of the segmentation and increase the robustness to noise. Furthermore, we show that the graph size can be greatly reduced by applying a smart segmentation scheme. This allows the segmentation to be computed in seconds instead of minutes, without deteriorating the segmentation accuracy, making it ideal for a clinical setup. An extensive evaluation on 20 OCT datasets of healthy eyes was performed and showed a mean unsigned segmentation error of 3.05 ±0.54 μm over all datasets when compared to the average observer, which is lower than the inter-observer variability. Similar performance was measured for the task of drusen segmentation, demonstrating the usefulness of using soft constraints as a tool to deal with pathologies.

First insights into the phylogenetic diversity of Mycobacterium tuberculosis in Nepal

Background Tuberculosis (TB) is a major public health problem in Nepal. Strain variation in Mycobacterium tuberculosis may influence the outcome of TB infection and disease. To date, the phylogenetic diversity of M. tuberculosis in Nepal is unknown. Methods and Findings We analyzed 261 M. tuberculosis isolates recovered from pulmonary TB patients recruited between August 2009 and August 2010 in Nepal. M. tuberculosis lineages were determined by single nucleotide polymorphisms (SNP) typing and spoligotyping. Drug resistance was determined by sequencing the hot spot regions of the relevant target genes. Overall, 164 (62.8%) TB patients were new, and 97 (37.2%) were previously treated. Any drug resistance was detected in 50 (19.2%) isolates, and 16 (6.1%) were multidrug-resistant. The most frequent M. tuberculosis lineage was Lineage 3 (CAS/Delhi) with 106 isolates (40.6%), followed by Lineage 2 (East-Asian lineage, includes Beijing genotype) with 84 isolates (32.2%), Lineage 4 (Euro-American lineage) with 41 (15.7%) isolates, and Lineage 1 (Indo-Oceanic lineage) with 30 isolates (11.5%). Based on spoligotyping, we found 45 different spoligotyping patterns that were previously described. The Beijing (83 isolates, 31.8%) and CAS spoligotype (52, 19.9%) were the dominant spoligotypes. A total of 36 (13.8%) isolates could not be assigned to any known spoligotyping pattern. Lineage 2 was associated with female sex (adjusted odds ratio [aOR] 2.58, 95% confidence interval [95% CI] 1.42–4.67, p = 0.002), and any drug resistance (aOR 2.79; 95% CI 1.43–5.45; p = 0.002). We found no evidence for an association of Lineage 2 with age or BCG vaccination status. Conclusions We found a large genetic diversity of M. tuberculosis in Nepal with representation of all four major lineages. Lineages 3 and 2 were dominating. Lineage 2 was associated with clinical characteristics. This study fills an important gap on the map of the M. tuberculosis genetic diversity in the Asian region.

Two new rapid SNP-typing methods for classifying Mycobacterium tuberculosis complex into the main phylogenetic lineages

There is increasing evidence that strain variation in Mycobacterium tuberculosis complex (MTBC) might influence the outcome of tuberculosis infection and disease. To assess genotype-phenotype associations, phylogenetically robust molecular markers and appropriate genotyping tools are required. Most current genotyping methods for MTBC are based on mobile or repetitive DNA elements. Because these elements are prone to convergent evolution, the corresponding genotyping techniques are suboptimal for phylogenetic studies and strain classification. By contrast, single nucleotide polymorphisms (SNP) are ideal markers for classifying MTBC into phylogenetic lineages, as they exhibit very low degrees of homoplasy. In this study, we developed two complementary SNP-based genotyping methods to classify strains into the six main human-associated lineages of MTBC, the "Beijing" sublineage, and the clade comprising Mycobacterium bovis and Mycobacterium caprae. Phylogenetically informative SNPs were obtained from 22 MTBC whole-genome sequences. The first assay, referred to as MOL-PCR, is a ligation-dependent PCR with signal detection by fluorescent microspheres and a Luminex flow cytometer, which simultaneously interrogates eight SNPs. The second assay is based on six individual TaqMan real-time PCR assays for singleplex SNP-typing. We compared MOL-PCR and TaqMan results in two panels of clinical MTBC isolates. Both methods agreed fully when assigning 36 well-characterized strains into the main phylogenetic lineages. The sensitivity in allele-calling was 98.6% and 98.8% for MOL-PCR and TaqMan, respectively. Typing of an additional panel of 78 unknown clinical isolates revealed 99.2% and 100% sensitivity in allele-calling, respectively, and 100% agreement in lineage assignment between both methods. While MOL-PCR and TaqMan are both highly sensitive and specific, MOL-PCR is ideal for classification of isolates with no previous information, whereas TaqMan is faster for confirmation. Furthermore, both methods are rapid, flexible and comparably inexpensive.

Phylogenetic analysis of the gag region encoding the matrix protein of small ruminant lentiviruses: comparative analysis and molecular epidemiological applications

Little sequence information exists on the matrix-protein (MA) encoding region of small ruminant lentiviruses (SRLV). Fifty-two novel sequences were established and permitted a first phylogenetic analysis of this region of the SRLV genome. The variability of the MA encoding region is higher compared to the gag region encoding the capsid protein and surprisingly close to that reported for the env gene. In contrast to primate lentiviruses, the deduced amino acid sequences of the N- and C-terminal domains of MA are variable. This permitted to pinpoint a basic domain in the N-terminal domain that is conserved in all lentiviruses and likely to play an important functional role. Additionally, a seven amino acid insertion was detected in all MVV strains, which may be used to differentiate CAEV and MVV isolates. A molecular epidemiology analysis based on these sequences indicates that the Italian lentivirus strains are closely related to each other and to the CAEV-CO strain, a prototypic strain isolated three decades ago in the US. This suggests a common origin of the SRLV circulating in the monitored flocks, possibly related to the introduction of infected goats in a negative population. Finally, this study shows that the MA region is suitable for phylogenetic studies and may be applied to monitor SRLV eradication programs.