On the Importance of Domain Awareness in Classifier Interpretations in Medical Imaging.

Translating the success of deep learning-based computer-assisted classification into clinical adaptation hinges on the ability to explain a prediction's causality. Post-hoc interpretability approaches, especially counterfactual techniques, have shown both technical and psychological potential. Nevertheless, currently dominant approaches utilize heuristic, unvalidated methodology. Thereby, they potentially operate the underlying networks outside their validated domain, adding doubt in the predictor's abilities instead of generating knowledge and trust. In this work, we investigate this out-of-distribution problem for medical image pathology classifiers and propose marginalization techniques and evaluation procedures to overcome it. Furthermore, we propose a complete domain-aware pipeline for radiology environments. Its validity is demonstrated on a synthetic and two publicly available image datasets. Specifically, we evaluate using the CBIS-DDSM/DDSM mammography collection and the Chest X-ray14 radiographs. Our solution shows, both quantitatively and qualitatively, a significant reduction of localization ambiguity and clearer conveying results.

Employing similarity to highlight differences: On the impact of anatomical assumptions in chest X-ray registration methods.

To facilitate both the detection and the interpretation of findings in chest X-rays, comparison with a previous image of the same patient is very valuable to radiologists. Today, the most common approach for deep learning methods to automatically inspect chest X-rays disregards the patient history and classifies only single images as normal or abnormal. Nevertheless, several methods for assisting in the task of comparison through image registration have been proposed in the past. However, as we illustrate, they tend to miss specific types of pathological changes like cardiomegaly and effusion. Due to assumptions on fixed anatomical structures or their measurements of registration quality, they produce unnaturally deformed warp fields impacting visualization of differences between moving and fixed images. We aim to overcome these limitations, through a new paradigm based on individual rib pair segmentation for anatomy penalized registration. Our method proves to be a natural way to limit the folding percentage of the warp field to 1/6 of the state of the art while increasing the overlap of ribs by more than 25%, implying difference images showing pathological changes overlooked by other methods. We develop an anatomically penalized convolutional multi-stage solution on the National Institutes of Health (NIH) data set, starting from less than 25 fully and 50 partly labeled training images, employing sequential instance memory segmentation with hole dropout, weak labeling, coarse-to-fine refinement and Gaussian mixture model histogram matching. We statistically evaluate the benefits of our method and highlight the limits of currently used metrics for registration of chest X-rays.

Lung Epithelial CYP1 Activity Regulates Aryl Hydrocarbon Receptor Dependent Allergic Airway Inflammation.

The lung epithelial barrier serves as a guardian towards environmental insults and responds to allergen encounter with a cascade of immune reactions that can possibly lead to inflammation. Whether the environmental sensor aryl hydrocarbon receptor (AhR) together with its downstream targets cytochrome P450 (CYP1) family members contribute to the regulation of allergic airway inflammation remains unexplored. By employing knockout mice for AhR and for single CYP1 family members, we found that AhR-/- and CYP1B1-/- but not CYP1A1-/- or CYP1A2-/- animals display enhanced allergic airway inflammation compared to WT. Expression analysis, immunofluorescence staining of murine and human lung sections and bone marrow chimeras suggest an important role of CYP1B1 in non-hematopoietic lung epithelial cells to prevent exacerbation of allergic airway inflammation. Transcriptional analysis of murine and human lung epithelial cells indicates a functional link of AhR to barrier protection/inflammatory mediator signaling upon allergen challenge. In contrast, CYP1B1 deficiency leads to enhanced expression and activity of CYP1A1 in lung epithelial cells and to an increased availability of the AhR ligand kynurenic acid following allergen challenge. Thus, differential CYP1 family member expression and signaling via the AhR in epithelial cells represents an immunoregulatory layer protecting the lung from exacerbation of allergic airway inflammation.

Multi-Task Fusion for Improving Mammography Screening Data Classification.

Machine learning and deep learning methods have become essential for computer-assisted prediction in medicine, with a growing number of applications also in the field of mammography. Typically these algorithms are trained for a specific task, e.g., the classification of lesions or the prediction of a mammogram's pathology status. To obtain a comprehensive view of a patient, models which were all trained for the same task(s) are subsequently ensembled or combined. In this work, we propose a pipeline approach, where we first train a set of individual, task-specific models and subsequently investigate the fusion thereof, which is in contrast to the standard model ensembling strategy. We fuse model predictions and high-level features from deep learning models with hybrid patient models to build stronger predictors on patient level. To this end, we propose a multi-branch deep learning model which efficiently fuses features across different tasks and mammograms to obtain a comprehensive patient-level prediction. We train and evaluate our full pipeline on public mammography data, i.e., DDSM and its curated version CBIS-DDSM, and report an AUC score of 0.962 for predicting the presence of any lesion and 0.791 for predicting the presence of malignant lesions on patient level. Overall, our fusion approaches improve AUC scores significantly by up to 0.04 compared to standard model ensembling. Moreover, by providing not only global patient-level predictions but also task-specific model results that are related to radiological features, our pipeline aims to closely support the reading workflow of radiologists.

Ragweed plants grown under elevated CO2 levels produce pollen which elicit stronger allergic lung inflammation.

BACKGROUND: Common ragweed has been spreading as a neophyte in Europe. Elevated CO2 levels, a hallmark of global climate change, have been shown to increase ragweed pollen production, but their effects on pollen allergenicity remain to be elucidated. METHODS: Ragweed was grown in climate-controlled chambers under normal (380 ppm, control) or elevated (700 ppm, based on RCP4.5 scenario) CO2 levels. Aqueous pollen extracts (RWE) from control- or CO2 -pollen were administered in vivo in a mouse model for allergic disease (daily for 3-11 days, n = 5) and employed in human in vitro systems of nasal epithelial cells (HNECs), monocyte-derived dendritic cells (DCs), and HNEC-DC co-cultures. Additionally, adjuvant factors and metabolites in control- and CO2 -RWE were investigated using ELISA and untargeted metabolomics. RESULTS: In vivo, CO2 -RWE induced stronger allergic lung inflammation compared to control-RWE, as indicated by lung inflammatory cell infiltrate and mediators, mucus hypersecretion, and serum total IgE. In vitro, HNECs stimulated with RWE increased indistinctively the production of pro-inflammatory cytokines (IL-8, IL-1ß, and IL-6). In contrast, supernatants from CO2 -RWE-stimulated HNECs, compared to control-RWE-stimulated HNECS, significantly increased TNF and decreased IL-10 production in DCs. Comparable results were obtained by stimulating DCs directly with RWEs. The metabolome analysis revealed differential expression of secondary plant metabolites in control- vs CO2 -RWE. Mixes of these metabolites elicited similar responses in DCs as compared to respective RWEs. CONCLUSION: Our results indicate that elevated ambient CO2 levels elicit a stronger RWE-induced allergic response in vivo and in vitro and that RWE increased allergenicity depends on the interplay of multiple metabolites.

Increased estrogen to androgen ratio enhances immunoglobulin levels and impairs B cell function in male mice.

Sex steroids, such as estrogens and androgens, are important regulators of the humoral immune response. Studies in female mice have demonstrated that alteration of circulating estrogen concentration regulates antibody-mediated immunity. As males have normally little endogenous estrogen, we hypothesized that in males high estrogens and low androgens affect the immune system and enhance the allergic inflammatory response. Here, we studied transgenic male mice expressing human aromatase (AROM+). These animals have a high circulating estrogen to androgen ratio (E/A), causing female traits such as gynecomastia. We found that AROM+ male mice had significantly higher plasma immunoglobulin levels, particularly IgE. Flow cytometry analyses of splenocytes revealed changes in mature/immature B cell ratio together with a transcriptional upregulation of the Igh locus. Furthermore, higher proliferation rate and increased IgE synthesis after IgE class-switching was found. Subsequently, we utilized an ovalbumin airway challenge model to test the allergic response in AROM+ male mice. In line with above observations, an increase in IgE levels was measured, albeit no impact on immune cell infiltration into the lungs was detected. Together, our findings suggest that high circulating E/A in males significantly alters B cell function without any significant enhancement in allergic inflammation.

Pollen exposure weakens innate defense against respiratory viruses.

BACKGROUND: Hundreds of plant species release their pollen into the air every year during early spring. During that period, pollen allergic as well as non-allergic patients frequently present to doctors with severe respiratory tract infections. Our objective was therefore to assess whether pollen may interfere with antiviral immunity. METHODS: We combined data from real-life human exposure cohorts, a mouse model and human cell culture to test our hypothesis. RESULTS: Pollen significantly diminished interferon-λ and pro-inflammatory chemokine responses of airway epithelia to rhinovirus and viral mimics and decreased nuclear translocation of interferon regulatory factors. In mice infected with respiratory syncytial virus, co-exposure to pollen caused attenuated antiviral gene expression and increased pulmonary viral titers. In non-allergic human volunteers, nasal symptoms were positively correlated with airborne birch pollen abundance, and nasal birch pollen challenge led to downregulation of type I and -III interferons in nasal mucosa. In a large patient cohort, numbers of rhinoviruspositive cases were correlated with airborne birch pollen concentrations. CONCLUSION: The ability of pollen to suppress innate antiviral immunity, independent of allergy, suggests that high-risk population groups should avoid extensive outdoor activities when pollen and respiratory virus seasons coincide.

Deep Sequential Segmentation of Organs in Volumetric Medical Scans.

Segmentation in 3-D scans is playing an increasingly important role in current clinical practice supporting diagnosis, tissue quantification, or treatment planning. The current 3-D approaches based on convolutional neural networks usually suffer from at least three main issues caused predominantly by implementation constraints-first, they require resizing the volume to the lower-resolutional reference dimensions, and second, the capacity of such approaches is very limited due to memory restrictions, and third, all slices of volumes have to be available at any given training or testing time. We address these problems by a U-Net-like architecture consisting of bidirectional convolutional long short-term memory and convolutional, pooling, upsampling, and concatenation layers enclosed into time-distributed wrappers. Our network can either process the full volumes in a sequential manner or segment slabs of slices on demand. We demonstrate performance of our architecture on vertebrae and liver segmentation tasks in 3-D computed tomography scans.

Fully automatic cross-modality localization and labeling of vertebral bodies and intervertebral discs in 3D spinal images.

PURPOSE: We present a cross-modality and fully automatic pipeline for labeling of intervertebral discs and vertebrae in volumetric data of the lumbar and thoracolumbar spine. The main goal is to provide an algorithm that is applicable to a wide range of different sequences and acquisition protocols, like T1- and T2- weighted MR scans, MR Dixon data, and CT scans. This requires that the learned models generalize without retraining to modalities and scans with unseen image contrasts. METHODS: We address this challenge by automatically localizing the sacral region combining local entropy-optimized texture models with convolutional neural networks. For subsequent labeling, local three-disc entropy models are matched iteratively to the spinal column. Every model-matched position is further refined by an intensity-based template-matching approach, based solely on the reduced intensity scale provided by the entropy models. RESULTS: We evaluated our method on 161 publicly available scans, acquired on various scanners. We showed that our method can deal with a wide range of different MR protocols as well as with CT data. We achieved a sacrum detection rate of 93.6%. Mean center accuracies ranged from 2.5 ± 1.5 to 5.7 ± 3.8 mm for the different sets of scans. CONCLUSION: We present a novel spine labeling framework that is applicable to a highly heterogeneous set of scans without retraining of the method. Our approach achieves high sacrum localization accuracy and shows promising labeling results. To the best of our knowledge, an algorithm able to deal with such a diverse set of MR and CT scans has not yet been presented in the literature.

Fully Convolutional Architectures for Multiclass Segmentation in Chest Radiographs.

The success of deep convolutional neural networks (NNs) on image classification and recognition tasks has led to new applications in very diversified contexts, including the field of medical imaging. In this paper, we investigate and propose NN architectures for automated multiclass segmentation of anatomical organs in chest radiographs (CXRs), namely for lungs, clavicles, and heart. We address several open challenges including model overfitting, reducing number of parameters, and handling of severely imbalanced data in CXR by fusing recent concepts in convolutional networks and adapting them to the segmentation problem task in CXR. We demonstrate that our architecture combining delayed subsampling, exponential linear units, highly restrictive regularization, and a large number of high-resolution low-level abstract features outperforms state-of-the-art methods on all considered organs, as well as the human observer on lungs and heart. The models use a multiclass configuration with three target classes and are trained and tested on the publicly available Japanese Society of Radiological Technology database, consisting of 247 X-ray images the ground-truth masks for which are available in the segmentation in CXR database. Our best performing model, trained with the loss function based on the Dice coefficient, reached mean Jaccard overlap scores of 95% for lungs, 86.8% for clavicles, and 88.2% for heart. This architecture outperformed the human observer results for lungs and heart.

Pro-Inflammatory versus Immunomodulatory Effects of Silver Nanoparticles in the Lung: The Critical Role of Dose, Size and Surface Modification.

The growing use of silver nanoparticles (Ag-NPs) in consumer products raises concerns about their toxicological potential. The purpose of the study was to investigate the size- and coating-dependent pulmonary toxicity of Ag-NPs in vitro and in vivo, using an ovalbumin (OVA)-mouse allergy model. Supernatants from (5.6-45 µg/mL) Ag50-PVP, Ag200-PVP or Ag50-citrate-treated NR8383 alveolar macrophages were tested for lactate dehydrogenase and glucuronidase activity, tumor necrosis factor (TNF)-α release and reactive oxygen species (ROS) production. For the in vivo study, NPs were intratracheally instilled in non-sensitized (NS) and OVA-sensitized (S) mice (1-50 µg/mouse) prior to OVA-challenge and bronchoalveolar lavage fluid (BALF) inflammatory infiltrate was evaluated five days after challenge. In vitro results showed a dose-dependent cytotoxicity of Ag-NPs, which was highest for Ag50-polyvinilpyrrolidone (PVP), followed by Ag50-citrate, and lowest for Ag200-PVP. In vivo 10-50 µg Ag50-PVP triggered a dose-dependent pulmonary inflammatory milieu in NS and S mice, which was significantly higher in S mice and was dampened upon instillation of Ag200-PVP. Surprisingly, instillation of 1 µg Ag50-PVP significantly reduced OVA-induced inflammatory infiltrate in S mice and had no adverse effect in NS mice. Ag50-citrate showed similar beneficial effects at low concentrations and attenuated pro-inflammatory effects at high concentrations. The lung microbiome was altered by NPs instillation dependent on coating and/or mouse batch, showing the most pronounced effects upon instillation of 50 µg Ag50-citrate, which caused an increased abundance of operational taxonomic units assigned to Actinobacteria, Bacteroidetes, Firmicutes and Proteobacteria. However, no correlation with the biphasic effect of low and high Ag-NPs dose was found. Altogether, both in vitro and in vivo data on the pulmonary effects of Ag-NPs suggest the critical role of the size, dose and surface functionalization of Ag-NPs, especially in susceptible allergic individuals. From the perspective of occupational health, care should be taken by the production of Ag-NPs-containing consumer products.

Automated Anatomy-Based Tracking of Systemic Arteries in Arbitrary Field-of-View CTA Scans.

We propose an automated pipeline for vessel centerline extraction in 3-D computed tomography angiography (CTA) scans with arbitrary fields of view. The principal steps of the pipeline are body part detection, candidate seed selection, segment tracking, which includes centerline extraction, and vessel tree growing. The final tree-growing step can be instantiated in either a semi- or fully automated fashion. The fully automated initialization is carried out using a vessel position regression algorithm. Both semi-and fully automated methods were evaluated on 30 CTA scans comprising neck, abdominal, and leg arteries in multiple fields of view. High detection rates and centerline accuracy values for 38 distinct vessels demonstrate the effectiveness of our approach.

CD40-signalling abrogates induction of RORγt+ Treg cells by intestinal CD103+ DCs and causes fatal colitis.

Immune homeostasis in intestinal tissues depends on the generation of regulatory T (Treg) cells. CD103+ dendritic cells (DCs) acquire microbiota-derived material from the gut lumen for transport to draining lymph nodes and generation of receptor-related orphan γt+ (RORγt+) Helios--induced Treg (iTreg) cells. Here we show CD40-signalling as a microbe-independent signal that can induce migration of CD103+ DCs from the lamina propria (LP) to the mesenteric lymph nodes. Transgenic mice with constitutive CD11c-specific CD40-signalling have reduced numbers of CD103+ DCs in LP and a low frequency of RORγt+Helios- iTreg cells, exacerbated inflammatory Th1/Th17 responses, high titres of microbiota-specific immunoglobulins, dysbiosis and fatal colitis, but no pathology is detected in other tissues. Our data demonstrate a CD40-dependent mechanism capable of abrogating iTreg cell induction by DCs, and suggest that the CD40L/CD40-signalling axis might be able to intervene in the generation of new iTreg cells in order to counter-regulate immune suppression to enhance immunity.

Evaluation and comparison of 3D intervertebral disc localization and segmentation methods for 3D T2 MR data: A grand challenge.

The evaluation of changes in Intervertebral Discs (IVDs) with 3D Magnetic Resonance (MR) Imaging (MRI) can be of interest for many clinical applications. This paper presents the evaluation of both IVD localization and IVD segmentation methods submitted to the Automatic 3D MRI IVD Localization and Segmentation challenge, held at the 2015 International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI2015) with an on-site competition. With the construction of a manually annotated reference data set composed of 25 3D T2-weighted MR images acquired from two different studies and the establishment of a standard validation framework, quantitative evaluation was performed to compare the results of methods submitted to the challenge. Experimental results show that overall the best localization method achieves a mean localization distance of 0.8 mm and the best segmentation method achieves a mean Dice of 91.8%, a mean average absolute distance of 1.1 mm and a mean Hausdorff distance of 4.3 mm, respectively. The strengths and drawbacks of each method are discussed, which provides insights into the performance of different IVD localization and segmentation methods.

Ambrosia artemisiifolia (ragweed) in Germany - current presence, allergological relevance and containment procedures.

Ambrosia artemisiifolia (ragweed) is a neophyte in Europe and Germany, which originated from the United States of America. In the USA the rate of sensitization against ragweed equals that of grass pollen, and without containment the rate of allergic sensitizations against ragweed pollen will clearly increase. Currently, the most frequent sensitizations in Germany are against grass pollen, followed by sensitizations against house dust mite and birch pollen. Ragweed pollen evokes symptoms at about 10 pollen/m3, grass pollen at about 15 pollen/m3. These concentrations of ragweed pollen are only reached on limited occasions in Germany. Ragweed cross-reacts with mugwort (Artemisia vulgaris) and a correct diagnosis is only feasible with the ragweed specific allergen Amb a 1. Due to cross reactivity with mugwort, new sensitizations against ragweed pollen are not needed to evoke allergic symptoms. The neophyte encounters an already mugwort-sensitized population, extends the pollen season and may provoke new sensitizations. Ragweed sensitizations are characterized by an increased tendency to also affect the lower airways, which is less with mugwort sensitizations. Thus containment of ragweed is needed. Ragweed seeds are imported or spread by contaminated bird feed, the transport of ragweed contaminated soil (also in tyre treads) and agricultural products from infested areas. States bordering on ragweed positive areas, like Brandenburg and Bavaria, are especially at risk and invasion is already underway. Ragweed seeds survive up to 40 years in soil, and so extended timescales for eradication and observations are needed. Germany is, compared to other countries like France (Rhone-Valley), Italy (Po-Valley), Ukraine and Hungary, limited in respect to ragweed infestation. Conditions in Germany are therefore favourable for the containment of ragweed. Switzerland implemented legislation against birdseed contamination by ragweed early during the plants expansion, and obligatory ragweed registration- and eradication showed that ragweed containment is possible. Without counter measures ragweed expansion in Germany will take place, resulting in more allergic disease. Considering the increasing number of allergic individuals, even without ragweed invasion, containment of the neophyte should be actively persued. Unfortunately, time is running out.

Vascular dysfunction after coarctation repair is related to the age at surgery.

BACKGROUND: Despite repair of aortic coarctation, hypertension is frequent in adults and premature coronary and cerebrovascular disease remain of concern. Persistent impairment of arterial dilation has been suspected to contribute to abnormal blood pressure regulation. We tested the hypothesis that arterial reactivity is more likely to be impaired in patients corrected at older age. METHODS: We studied changes in brachial artery diameter in response to reactive hyperemia (FMD) and to nitroglycerin (NMD) in 36 patients and 25 controls. Depending on their age at surgery, patients were divided in group A (surgery <9 years) and group B (surgery > or =9 years). RESULTS: Cholesterol levels and percentage of smokers were similar in patients and controls, but 16 patients had arterial hypertension compared to none of the controls. Endothelium-dependent vasodilation, FMD, and endothelium-independent vasodilation, NMD, were significantly impaired in patients vs. controls (8.2+/-6.2% vs. 13.0+/-5.1%, p<0.001 and 12.9+/-8.0% vs. 18.8+/-9.2%, p<0.01, respectively), both, in hypertensives (8.3+/-6.0%, p<0.01 and 11.8+/-6.0%, p<0.05) and in normotensives (8.1+/-6.5% p<0.01 and 13.8+/-9.3%, p<0.05). However, FMD and NMD in patients of group A did not significantly differ from that in controls (10.0+/-6.7% n.s. and 15.0+/-7.6% n.s.), whereas they were lowest in patients of group B (5.5+/-4.3%, p<0.0001 and 9.6+/-7.7% p<0.001). CONCLUSIONS: Persistent impairment of FMD and NMD after repair of coarctation is more likely to be present in patients corrected at older age. It may be an important contributor to abnormal blood pressure regulation and late morbidity and mortality.

Long-term outcome of patients with ventricular septal defect considered not to require surgical closure during childhood.

OBJECTIVES: The purpose of the study was to assess the long-term outcome of patients with small ventricular septal defects (VSDs) considered not to require surgical closure during childhood. BACKGROUND: Although patients with small VSDs have generally been considered not to require surgery, more recent data suggest that a significant percentage of these patients develop serious problems during adult life. METHODS: A total of 229 consecutive patients (115 females) with a VSD considered too small to require surgery during childhood as defined by normal pulmonary artery pressure, less than 50% shunt, pulmonary vascular resistance < or =200 dynes x s cm(-5), no VSD-related aortic regurgitation (AR), and no symptoms and who had no additional hemodynamically relevant heart defect were followed in an adult congenital heart disease program. Physical examination, electrocardiography, and echocardiography were performed in all patients in one- to three-year intervals; exercise tests and Holter monitoring were performed in 140 and 127 patients, respectively. RESULTS: Follow-up was completed in 222 patients (97%). Mean age at last visit was 30 +/- 10 years. Spontaneous VSD closure was observed in 14 patients (6%). No patients died, four patients (1.8%) had an episode of endocarditis, of whom two required aortic valve replacement, and one additional patient (0.4%) had surgical closure for hemodynamic reasons. For 118 patients who entered the study between 1993 and 1996 and were prospectively followed for 7.4 +/- 1.2 years, event-free survival with end points defined as death, endocarditis or heart surgery was 99.1 +/- 0.8% at three years, 96.5 +/- 1.7% at six years and 95.5 +/- 1.9% at eight years. At last visit, 94.6% of all patients studied were symptom free. Left ventricular (LV) size by echocardiography was normal in 198 (89%) patients, borderline in 23 patients and definitely enlarged in only one patient. None had systolic LV dysfunction, and pulmonary artery pressure (PAP) was normal in all patients. Mean exercise capacity was 92 +/- 21% of expected, and 87% of patients had no arrhythmias on Holter monitoring, with the remainder showing benign rhythm disorders. CONCLUSIONS: Outcome in well-selected patients with a small VSD is good. Surgical closure does not appear to be required during childhood as long as left-to-right shunt is <50% and signs of LV volume overload are absent, when PAP is not elevated, and no VSD-related AR or symptoms are present.