Characterization of tissue types in basal cell carcinoma images via generative modeling and concept vectors

The promise of machine learning methods to act as decision support systems for pathologists continues to grow. However, central to their successful adoption must be interpretable implementations so that people can trust and learn from them effectively. Generative modeling, most notable in the form of adversarial generative models, is a naturally interpretable technique because the quality of the model is explicit from the quality of images it generates. Such a model can be further assessed by exploring its latent space, using human-meaningful concepts by defining concept vectors. Motivated by these ideas, we apply for the first time generative methods to histological images of basal cell carcinoma (BCC). By simultaneously learning to generate and encode realistic image patches, we extract feature rich latent vectors that correspond to various tissue morphologies, namely BCC, epidermis, keratin, papillary dermis and inflammation. We show that a logistic regression model trained on these latent vectors can achieve high classification accuracies across 6 binary tasks (86-98%). Further, by projecting the latent vectors onto learned concept vectors we can generate a score for the absence or degree of presence for a given concept, providing semantically accurate "conceptual summaries" of the various tissues types within a patch. This can be extended to generate multi-dimensional heat maps for whole-image specimens, which characterizes the tissue in a similar way to a pathologist. We additionally find that accurate concept vectors can be defined using a small labeled dataset.

Performance selection for Thoroughbreds racing in Hong Kong

REASONS FOR PERFORMING STUDY

Different indicators of racing performance are commonly used in the racing industry to assess the genetic superiority of racing Thoroughbreds. However, how well these indicators predict the performance of future progeny or siblings varies depending on the population and circumstances in which the indicators were recorded or achieved.

OBJECTIVES

To identify heritable indicators of racing performance for horses racing in Hong Kong.

STUDY DESIGN

Heritability analysis of racing performance traits.

METHODS

Performance data on the population of Thoroughbreds racing in Hong Kong between 3 September 2000 and 12 March 2011 (n = 4947) were acquired and used to estimate the heritabilities and probability values of fixed effects and covariates for a range of racing performance traits. Heritabilities for all performance traits were estimated using a single trait animal model. Each model included, as a minimum, the effects of sex, region of origin and trainer.

RESULTS

Heritability estimates for traits relating to finish position ranged from 0.01 to 0.06. Average handicap weight had a heritability of 0.07 ± 0.03. The effects of sex (fixed) and trainer (random) were significant (P<0.05) for all performance traits relating to earnings measures, handicap weights and finish positions. The heritability of win time at 1600 m was 0.52 ± 0.06 and was the only significant estimate of heritability for win time in the current study.

CONCLUSIONS

Although significantly affected by multiple environmental factors, certain indicators of Hong Kong racing performance can be reliably used to predict the performance of the individual's progeny or siblings. However, despite Hong Kong's controlled racing environment, these indicators appear to be no more heritable than in other less controlled racing environments.

Modelling cell turnover in a complex tissue during development

The growth of organs results from proliferation within distinct cellular compartments. Organ development also involves transitions between cell types and variations in cell cycle duration as development progresses, and is regulated by a balance between entry into the compartment, proliferation of cells within the compartment, acquisition of quiescence and exit from that cell state via differentiation or death. While it is important to understand how environmental or genetic alterations can perturb such development, most approaches employed to date are descriptive rather than quantitative. This is because the identification and quantification of such parameters, while tractable in vitro, is challenging in the context of a complex tissue in vivo. Here we present a new framework for determining cell turnover in developing organs in vivo that combines cumulative cell-labelling and quantification of distinct cell-cycle phases without assuming homogeneity of behaviour within that compartment. A mathematical model is given that allows the calculation of cell cycle length in the context of a specific biological example and assesses the uncertainty of this calculation due to incomplete knowledge of cell cycle dynamics. This includes the development of a two population model to quantify possible heterogeneity of cell cycle length within a compartment and estimate the aggregate proliferation rate. These models are demonstrated on data collected from a progenitor cell compartment within the developing mouse kidney, the cap mesenchyme. This tissue was labelled by cumulative infusion, volumetrically quantified across time, and temporally analysed for the proportion of cells undergoing proliferation. By combining the cell cycle length predicted by the model with measurements of total cell population and mitotic rate, this approach facilitates the quantification of exit from this compartment without the need for a direct marker of that event. As a method specifically designed with assumptions appropriate to developing organs we believe this approach will be applicable to a range of developmental systems, facilitating estimations of cell cycle length and compartment behaviour that extend beyond simple comparisons of mitotic rates between normal and perturbed states.

Recycling endosome-dependent and -independent mechanisms for IL-10 secretion in LPS-activated macrophages

IL-10 is a key anti-inflammatory cytokine secreted by activated macrophages as a feedback control mechanism to prevent excessive inflammatory responses. Here, we define multiple intracellular trafficking pathways involved in the secretion of newly synthesized IL-10 from macrophages following TLR4 activation with LPS, and show how this relates to the previously defined trafficking pathways for IL-6 and TNF in macrophages simultaneously producing these proinflammatory cytokines. IL-10 exits the Golgi in multiple tubular carriers, including those dependent on p230GRIP. Some of the IL-10 is then delivered to recycling endosomes, where cytokine sorting may occur prior to its release. Another portion of the IL-10 is delivered to the cell surface in distinct vesicles colabeled for apoE. Thus, we show at least two post-Golgi pathways via which IL-10 is trafficked, ensuring its secretion from activated macrophages under different physiological conditions.

Early biocompatibility of crosslinked and non-crosslinked biologic meshes in a porcine model of ventral hernia repair

PURPOSE

Biologic meshes have unique physical properties as a result of manufacturing techniques such as decellularization, crosslinking, and sterilization. The purpose of this study is to directly compare the biocompatibility profiles of five different biologic meshes, AlloDerm(®) (non-crosslinked human dermal matrix), PeriGuard(®) (crosslinked bovine pericardium), Permacol(®) (crosslinked porcine dermal matrix), Strattice(®) (non-crosslinked porcine dermal matrix), and Veritas(®) (non-crosslinked bovine pericardium), using a porcine model of ventral hernia repair.

METHODS

Full-thickness fascial defects were created in 20 Yucatan minipigs and repaired with the retromuscular placement of biologic mesh 3 weeks later. Animals were euthanized at 1 month and the repair sites were subjected to tensile testing and histologic analysis. Samples of unimplanted (de novo) meshes and native porcine abdominal wall were also analyzed for their mechanical properties.

RESULTS

There were no significant differences in the biomechanical characteristics between any of the mesh-repaired sites at 1 month postimplantation or between the native porcine abdominal wall without implanted mesh and the mesh-repaired sites (P > 0.05 for all comparisons). Histologically, non-crosslinked materials exhibited greater cellular infiltration, extracellular matrix (ECM) deposition, and neovascularization compared to crosslinked meshes.

CONCLUSIONS

While crosslinking differentiates biologic meshes with regard to cellular infiltration, ECM deposition, scaffold degradation, and neovascularization, the integrity and strength of the repair site at 1 month is not significantly impacted by crosslinking or by the de novo strength/stiffness of the mesh.

Correlates of distress in children at risk for affective disorder: exploring predictors in the offspring of depressed and nondepressed mothers

BACKGROUND

Efforts to understand the correlates of psychological distress in children frequently examine possible correlates in samples of children who are selected for high levels of distress. The propose of this study was to compare distress correlates in a sample with depressed mothers, and thus at high-risk for distress, to a low-risk sample.

METHODS

Examining data from part of a larger project, the association of children's depressive symptoms and internalizing and externalizing problems to maternal depression level, life stress, verbal ability, and the experience of a traumatic event were examined in a series of regression equations.

RESULTS

Results indicated that children's depressive symptoms, rather than internalizing and externalizing problems, tended to be most consistently related to maternal variables, and also suggested that any experience of maternal depressive symptoms was associated with child problems. It was also found that child depressive symptoms were correlated with life events, but only for nondepressed mothers, and that at-risk children with higher levels of verbal ability were significantly less likely to report experiencing depressive symptoms and internalizing problems than were those with lower levels of verbal ability.

LIMITATIONS

Because these data are preliminary, further research examining a broader array of variables is important.

CONCLUSIONS

These results suggest the need for different models of these processes in different populations of children.

Body fat distribution and hemodynamic stress responses in premenopausal obese women: a preliminary study

To examine the relationship between body fat distribution and hemodynamic stress responses, cardiovascular responses to a speech task and a forehead cold pressor task were evaluated with 24 premenopausal women classified a priori as either centrally or peripherally obese. Results showed that women with central adiposity exhibited greater stress-related increases in diastolic blood pressure and total peripheral resistance, whereas women with peripheral adiposity exhibited greater stress-related increases in cardiac output. Depression, self-consciousness, hostility, and mood scores did not explain significant variance in the stress response differences between regional adiposity groups. The findings suggest that central adiposity may increase the risk of cardiovascular disease in women at least in part by enhancing vascular responses to stress.

Body fat distribution and hemodynamic stress responses in premenopausal obese women: a preliminary study

To examine the relationship between body fat distribution and hemodynamic stress responses, cardiovascular responses to a speech task and a forehead cold pressor task were evaluated with 24 premenopausal women classified a priori as either centrally or peripherally obese. Results showed that women with central adiposity exhibited greater stress-related increases in diastolic blood pressure and total peripheral resistance, whereas women with peripheral adiposity exhibited greater stress-related increases in cardiac output. Depression, self-consciousness, hostility, and mood scores did not explain significant variance in the stress response differences between regional adiposity groups. The findings suggest that central adiposity may increase the risk of cardiovascular disease in women at least in part by enhancing vascular responses to stress.

Field research on the relationship between stress and disease activity in rheumatoid arthritis

We review empirical evidence from two field studies for the role of stressful life events in disease flare-ups among patients with rheumatoid arthritis (RA). RA patients were expected to be more vulnerable psychologically, and physiologically, to stressful events in their everyday lives than other arthritis patients without an autoimmune disease. Findings from two studies are reviewed both for their substantive contribution, but also to provide guidance on measurement issues in future field research of this kind. One study included 41 patients with RA, who were interviewed weekly and called to a clinic for blood work and joint examinations when their levels of interpersonal stress increased significantly. A second study used a similar design but included comparison samples of patients with osteoarthritis (OA) and healthy controls of similar age and the same gender as the RA sample. The findings provided support for the proposition that interpersonal stressors are predictive of increases in disease activity. Not all RA patients, however, showed these relationships, and there was evidence that some participants with OA who were depressed also showed higher disease activity following interpersonal stressors. Significant individual differences in the stress-disease relationship were uncovered that deserve greater attention in future studies. Important improvements in the assessment of stressful events and refinements in panel study design are also presented as guides to research on the role of stress in disease processes.

Development of ultrasonic vocalization responses in genetically heterogeneous National Institute of Health (N:NIH) rats. I. Influence of age, testing experience, and associated factors

The N:NIH strain of rats was developed by the National Institutes of Health to provide a maximally heterogeneous population as a base for selective breeding (Hansen & Spuhler, 1984). Using the N:NIH strain, this laboratory will selectively breed adult animals that exhibited extremes of high or low ultrasonic vocalization (USV) rates as infants. Because nothing was known about USV in N:NIH rats, we characterized the development of isolation-induced USV in the first generation of this strain born in our laboratory. In a longitudinal/cross-sectional study of pups tested at 3, 10, 15, and 18 days postnatally, N:NIH pups emitted their highest rates of USV at 3-4 days postnatally and calling remained high for 10 days before declining. USV rates were found to be a relatively environmentally stable behavioral trait in that repeated testing did not significantly affect the calling rates of either individuals or litters, and only at 3 days postnatal age did naturally occurring ambient temperature variations (6 degrees C range) significantly affect USV responses. Individual differences in USV responses emerged by 10 days of age that were not simply correlations of body weight or rectal temperature, and pups at that age showed isolation calling rates that were highly predictive of their response levels 5 days later.