Relational Proxies: Fine-Grained Relationships as Zero-Shot Discriminators.

Visual categories that largely share the same set of local parts cannot be discriminated based on part information alone, as they mostly differ in the way the local parts relate to the overall global structure of the object. We propose Relational Proxies, a novel approach that leverages the relational information between the global and local views of an object for encoding its semantic label, even for categories it has not encountered during training. Starting with a rigorous formalization of the notion of distinguishability between categories that share attributes, we prove the necessary and sufficient conditions that a model must satisfy in order to learn the underlying decision boundaries to tell them apart. We design Relational Proxies based on our theoretical findings and evaluate it on seven challenging fine-grained benchmark datasets and achieve state-of-the-art results on all of them, surpassing the performance of all existing works with a margin exceeding 4% in some cases. We additionally show that Relational Proxies also generalizes to the zero-shot setting, where it can efficiently leverage emergent relationships among attributes and image views to generalize to unseen categories, surpassing current state-of-the-art in both the non-generative and generative settings. Implementation will be made public upon acceptance.

New perspectives on tracheal resection for COVID-19-related stenosis: A propensity score matching analysis.

OBJECTIVE: The large number of patients with COVID-19 subjected to prolonged invasive mechanical ventilation has been expected to result in a significant increase in tracheal stenosis in the next years. The aim of this study was to evaluate and compare postoperative outcomes of patients who survived COVID-19 critical illness and underwent tracheal resection for postintubation/posttracheostomy tracheal stenosis with those of non-COVID-19 patients. METHODS: It was single-center, retrospective study. All consecutive patients with post-intubation/posttracheostomy tracheal stenosis who underwent tracheal resection from February 2020 to March 2022 were enrolled. A total of 147 tracheal resections were performed: 24 were in post-COVID-19 patients and 123 were in non-COVID-19 patients. A 1:1 propensity score matching analysis was performed, considering age, gender, body mass index, and length of stenosis. After matching, 2 groups of 24 patients each were identified: a post-COVID-19 group and a non-COVID group. RESULTS: No mortality after surgery was registered. Posttracheostomy etiology of stenosis resulted more frequently in post-COVID-19 patients (n = 20 in the post-COVID-19 group vs n = 11 in the non-COVID-19 group; P = .03), as well as intensive care unit admissions during the postoperative period (16 vs 9 patients; P = .04). Need for postoperative reintubation for glottic edema and respiratory failure was higher in the post-COVID-19 group (7 vs 2 postoperative reintubation procedures; P = .04). Postoperative dysphonia was observed in 11 (46%) patients in the post-COVID-19 group versus 4 (16%) patients in the non-COVID-19 group (P = .03). CONCLUSIONS: Tracheal resection continues to be safe and effective in COVID-19-related tracheal stenosis scenarios. Intensive care unit admission rates and postoperative complications seem to be higher in post-COVID-19 patients who underwent tracheal resection compared with non-COVID-19 patients.

Prognostic Factors of Non-Predominant-Lepidic Lung Adenocarcinoma Presenting as Ground Glass Opacity: Results of a Multicenter Study.

This study aims to define the clinicopathological characteristics and prognosis of non-predominant lepidic invasive adenocarcinoma presenting as Ground Glass Opacity (GGO) nodules. The goal is to assess statistical relationships between histology, tumor size, location, and the incidence of relapse and lymph node dissemination. A retrospective multicenter study was conducted, including patients with GGO observed on CT scans between 2003 and 2021. Anamnestic, radiological, and histological data, as well as SUV values, lymphatic and vascular invasion, pathological stage, resection type, and adjuvant treatment, were analyzed. The primary endpoints were to evaluate prognostic factors for death and recurrence using Cox regression analysis. All 388 patients, including 277 with non-predominant lepidic invasive adenocarcinoma and 161 with lepidic adenocarcinoma, underwent curative anatomical resection. Non-predominant lepidic invasive adenocarcinoma demonstrated a worse prognosis than lepidic adenocarcinoma (p = 0.001). Independent prognostic factors for death and recurrence included lymph node involvement (p = 0.002) and vascular and lymphatic invasion (p < 0.001). In conclusion, non-predominant lepidic invasive adenocarcinoma and lymphatic and vascular invasion are prognostic factors for death and recurrence in GGO patients. Results suggest adjuvant treatment in the case of pN1-N2 disease, emphasizing the necessity of lymphadenectomy (sampling or systematic) for accurate staging and subsequent therapeutic procedures.

Thyroid transcription factor-1 expression in lung neuroendocrine tumours: a gender-related biomarker?

PURPOSE: Thyroid transcription factor-1 (TTF-1) assessed by immunohistochemistry (IHC) is a specific biomarker for lung adenocarcinoma, and is commonly used to confirm the pulmonary origin of neuroendocrine tumours (NET). The majority of the available data suggest that TTF-1 is favourable prognostic biomarker for lung adenocarcinomas, whereas its role is more conflicting for lung NET. The main aim of this multicenter retrospective study was to investigate the potentially relevant associations between TTF-1 biomarker and clinical and pathological features of the study population, as well as determine TTF-1 prognostic effect on the clinical outcome of the patients. METHODS: A multicentre retrospective study was conducted on 155 surgically-removed lung NET, with available IHC TTF-1 assessment. RESULTS: Median age was 59.5 years (range 13-86), 97 patients (62.6%) were females, 31 cases (20%) were atypical carcinoids, 4 (2.6%) had TNM stage IV. Mitotic count ≥2 per 10 high-power field was found in 35 (22.6%) subjects, whereas necrosis was detected in 20 patients (12.9%). TTF-1 was positive in 78 cases (50.3%). The median overall survival was 46.9 months (range 0.6-323) and the median progression-free survival was 39.1 months (range 0.6-323). Statistically significant associations were found between (1) TTF-1 positivity and female sex (p = 0.007); and among (2) TTF-1 positivity and the absence of necrosis (p = 0.018). CONCLUSIONS: This study highlights that TTF-1 positivity differs according to sex in lung NET, with a more common TTF-1 positive staining in female. Moreover, TTF-1 positivity correlated with the absence of necrosis. These data suggest that TTF-1 could potentially represent a gender-related biomarker for lung NET.

Histologic Analysis of Idiopathic Pulmonary Fibrosis by Morphometric and Fractal Analysis.

Idiopathic pulmonary fibrosis (IPF) is a chronic, progressive fibrotic lung disorder, ultimately leading to respiratory failure and death. Despite great research advances in understanding the mechanisms underlying the disease, its diagnosis, and its treatment, IPF still remains idiopathic without known biological or histological markers able to predict disease progression or response to treatment. The histologic hallmark of IPF is usual interstitial pneumonia (UIP), with its intricate architectural distortion and temporal inhomogeneity. We hypothesize that normal lung alveolar architecture can be compared to fractals, such as the Pythagoras tree with its fractal dimension (Df), and every pathological insult, distorting the normal lung structure, could result in Df variations. In this study, we aimed to assess the UIP histologic fractal dimension in relationship to other morphometric parameters in newly diagnosed IPF patients and its possible role in the prognostic stratification of the disease. Clinical data and lung tissue specimens were obtained from twelve patients with IPF, twelve patients with non-specific interstitial pneumonia (NSIP), and age-matched "healthy" control lung tissue from patients undergoing lung surgery for other causes. Histology and histomorphometry were performed to evaluate Df and lacunarity measures, using the box counting method on the FracLac ImageJ plugin. The results showed that Df was significantly higher in IPF patients compared to controls and fibrotic NSIP patients, indicating greater architectural distortion in IPF. Additionally, high Df values were associated with higher fibroblastic foci density and worse prognostic outcomes in IPF, suggesting that Df may serve as a potential novel prognostic marker for IPF. The scalability of Df measurements was demonstrated through repeated measurements on smaller portions from the same surgical biopsies, which were selected to mimic a cryobiopsy. Our study provides further evidence to support the use of fractal morphometry as a tool for quantifying and determining lung tissue remodeling in IPF, and we demonstrated a significant correlation between histological and radiological Df in UIP pattern, as well as a significant association between Df and FF density. Furthermore, our study demonstrates the scalability and self-similarity of Df measurements across different biopsy types, including surgical and smaller specimens.

CD36 regulates substrates utilisation in brown adipose tissue of spontaneously hypertensive rats: In vitro study.

Thermogenesis in brown adipose tissue (BAT) uses intracellular triglycerides, circulating free fatty acids and glucose as the main substrates. The objective of the current study was to analyse the role of CD36 fatty acid translocase in regulation of glucose and fatty acid utilisation in BAT. BAT isolated from spontaneously hypertensive rat (SHR) with mutant Cd36 gene and SHR-Cd36 transgenic rats with wild type variant was incubated in media containing labeled glucose and palmitate to measure substrate incorporation and oxidation. SHR-Cd36 versus SHR rats showed significantly increased glucose incorporation into intracellular lipids associated with reduced glycogen synthase kinase 3ß (GSK-3ß) protein expression and phosphorylation and increased oxidation of exogenous palmitate. It can be concluded that CD36 enhances glucose transport for lipogenesis in BAT by suppressing GSK-3ß and promotes direct palmitate oxidation.

B4GALT1 as a New Biomarker of Idiopathic Pulmonary Fibrosis.

Idiopathic pulmonary fibrosis (IPF) is a disease characterized by progressive scarring of the lung that involves the pulmonary interstitium. The disease may rapidly progress, leading to respiratory failure, and the long-term survival is poor. There are no accurate biomarkers available so far. Our aim was to evaluate the expression of the B4GALT1 in patients with IPF. Analysis of B4GALT1 gene expression was performed in silico on two gene sets, retrieved from the Gene Expression Omnibus database. Expression of B4GALT1 was then evaluated, both at the mRNA and protein levels, on lung specimens obtained from lung biopsies of 4 IPF patients, on one IPF-derived human primary cell and on 11 cases of IPF associated with cancer. In silico re-analysis demonstrated that the B4GALT1 gene was overexpressed in patients and human cell cultures with IPF (p = 0.03). Network analysis demonstrated that B4GALT1 upregulation was correlated with genes belonging to the EMT pathway (p = 0.01). The overexpression of B4GALT1 was observed, both at mRNA and protein levels, in lung biopsies of our four IPF patients and in the IPF-derived human primary cell, in other fibrotic non-lung tissues, and in IPF associated with cancer. In conclusion, our results indicate that B4GALT1 is overexpressed in IPF and could represent a novel marker of this disease.

Primary Leiomyosarcoma of the Mediastinum: A Rare and Challenging Diagnosis?

(1) Introduction: Leiomyosarcomas are highly aggressive mesenchymal neoplasm derived from smooth muscle cells which, in the mediastinum, are present in various primary organs; To our knowledge, less than 10 cases of primary mediastinal leiomyosarcoma have been described. Here, we report a compelling case of primary mediastinal leiomyosarcoma. (2) Case presentation: A 79-year-old woman was admitted to the Thoracic Surgery Unit of S. Andrea University Hospital for persisting cough, exertional dyspnea, and sternal pain. After multidisciplinary consultation, a CT-guided core needle biopsy of the mass was performed, resulting in a provisional diagnosis of mesenchymal neoplasm with smooth muscle differentiation without apparent signs of atypia. The patient underwent surgery that revealed a large irregularly shaped mass with a whorled pattern cut surface, showing admixed yellowish areas of necrosis and areas of hemorrhage. Histologic examination showed a smooth muscle neoplasm with atypia and necrosis, and a grade 2 primary mediastinal leiomyosarcoma diagnosis was given. (3) Conclusions: Soft tissue sarcomas represent a challenging diagnostic group of tumors due to their location, morphologic spectrum, and unique molecular background. Our case of primary mediastinal leiomyosarcoma shows how tumor heterogeneity and limited tissue sampling impact diagnosis. Further studies are needed to shed light on the disease by finding an appropriate molecular signature for each leiomyosarcoma subgroup, providing a more precise diagnosis and the correct background for tailored therapy.

Angioside: The role of Angiogenesis and Hypoxia in Lung Neuroendocrine Tumours According to Primary Tumour Location in Left or Right Parenchyma.

Well-differentiated lung neuroendocrine tumours (Lu-NETs), classified as typical (TC) and atypical (AC) carcinoids, represent 30% of NETs. Angiogenesis plays an essential role in NET development and progression. A higher vascular network is a marker of differentiation, with positive prognostic implications. Materials and Methods: We retrospectively evaluated microvessel density (MVD) by CD34 immunohistochemical (IHC) staining and hypoxia by IHC staining for Hypoxia-inducible factor 1α (HIF-1α), comparing right- and left-lung parenchyma in 53 lung NETs. Results: The median age was 66 years (39−81), 56.6% males, 24.5% AC, 40.5% left-sided tumours and 69.8% TNM stage I. The mitotic count was <2/10 per 10 HPF in 79.2%, and the absence of necrosis in 81.1%, 39.6% with Ki67, was ≤2%. The MVD, the number of vessels and the average vessel area median values were significantly higher in the right than the left parenchyma (p: 0.025, p: 0.019, p: 0.016, respectively). Hypoxia resulted present in 14/19 (73.6%) left tumours and in 10/20 (50%) right tumours in the parenchyma (p: 0.129). Conclusions: This study suggests a biological rationale for a different angiogenesis and hypoxia according to the Lu-NETs' location. In our study, left primary tumours were less vascularized and most likely to present hypoxia than right primary tumours. This finding could have potentially useful prognostic and predictive implications for Lu-NETs.

Semi-Supervised and Unsupervised Deep Visual Learning: A Survey.

State-of-the-art deep learning models are often trained with a large amount of costly labeled training data. However, requiring exhaustive manual annotations may degrade the model's generalizability in the limited-label regime.Semi-supervised learning and unsupervised learning offer promising paradigms to learn from an abundance of unlabeled visual data. Recent progress in these paradigms has indicated the strong benefits of leveraging unlabeled data to improve model generalization and provide better model initialization. In this survey, we review the recent advanced deep learning algorithms on semi-supervised learning (SSL) and unsupervised learning (UL) for visual recognition from a unified perspective. To offer a holistic understanding of the state-of-the-art in these areas, we propose a unified taxonomy. We categorize existing representative SSL and UL with comprehensive and insightful analysis to highlight their design rationales in different learning scenarios and applications in different computer vision tasks. Lastly, we discuss the emerging trends and open challenges in SSL and UL to shed light on future critical research directions.

Learning Graph Embeddings for Open World Compositional Zero-Shot Learning.

Compositional Zero-Shot learning (CZSL) aims to recognize unseen compositions of state and object visual primitives seen during training. A problem with standard CZSL is the assumption of knowing which unseen compositions will be available at test time. In this work, we overcome this assumption operating on the open world setting, where no limit is imposed on the compositional space at test time, and the search space contains a large number of unseen compositions. To address this problem, we propose a new approach, Compositional Cosine Graph Embedding (Co-CGE), based on two principles. First, Co-CGE models the dependency between states, objects and their compositions through a graph convolutional neural network. The graph propagates information from seen to unseen concepts, improving their representations. Second, since not all unseen compositions are equally feasible, and less feasible ones may damage the learned representations, Co-CGE estimates a feasibility score for each unseen composition, using the scores as margins in a cosine similarity-based loss and as weights in the adjacency matrix of the graphs. Experiments show that our approach achieves state-of-the-art performances in standard CZSL while outperforming previous methods in the open world scenario.

Prognostic significance of laterality in lung neuroendocrine tumors.

PURPOSE: Well-differentiated lung neuroendocrine tumors (Lu-NET) are classified as typical (TC) and atypical (AC) carcinoids, based on mitotic counts and necrosis. However, prognostic factors, other than tumor node metastasis (TNM) stage and the histopathological diagnosis, are still lacking. The current study is aimed to identify potential prognostic factors to better stratify lung NET, thus, improving patients' treatment strategy and follow-up. METHODS: A multicentric retrospective study, including 300 Lung NET, all surgically removed, from Italian and Spanish Institutions. RESULTS: Median age 61 years (13-86), 37.7% were males, 25.0% were AC, 42.0% were located in the lung left parenchyma, 80.3% presented a TNM stage I-II. Mitotic count was ≥2 per 10 high-power field (HPF) in 24.7%, necrosis in 13.0%. Median overall survival (OS) was 46.1 months (0.6-323), median progression-free survival (PFS) was 36.0 months (0.3-323). Female sex correlated with a more indolent disease (T1; N0; lower Ki67; lower mitotic count and the absence of necrosis). Left-sided primary tumors were associated with higher mitotic count and necrosis. At Cox-multivariate regression model, age, left-sided tumors, nodal (N) positive status and the diagnosis of AC resulted independent negative prognostic factors for PFS and OS. CONCLUSIONS: This study highlights that laterality is an independent prognostic factors in Lu-NETs, with left tumors being less frequent but showing a worse prognosis than right ones. A wider spectrum of clinical and pathological prognostic factors, including TNM stage, age and laterality is suggested. These parameters could help clinicians to personalize the management of Lu-NET.

Modeling the Background for Incremental and Weakly-Supervised Semantic Segmentation.

Deep neural networks have enabled major progresses in semantic segmentation. However, even the most advanced neural architectures suffer from important limitations. First, they are vulnerable to catastrophic forgetting, i.e., they perform poorly when they are required to incrementally update their model as new classes are available. Second, they rely on large amount of pixel-level annotations to produce accurate segmentation maps. To tackle these issues, we introduce a novel incremental class learning approach for semantic segmentation taking into account a peculiar aspect of this task: since each training step provides annotation only for a subset of all possible classes, pixels of the background class exhibit a semantic shift. Therefore, we revisit the traditional distillation paradigm by designing novel loss terms which explicitly account for the background shift. Additionally, we introduce a novel strategy to initialize classifier's parameters at each step in order to prevent biased predictions toward the background class. Finally, we demonstrate that our approach can be extended to point- and scribble-based weakly supervised segmentation, modeling the partial annotations to create priors for unlabeled pixels. We demonstrate the effectiveness of our approach with an extensive evaluation on the Pascal-VOC, ADE20K, and Cityscapes datasets, significantly outperforming state-of-the-art methods.

Pleural Involvement in IgG4-Related Disease: Case Report and Review of the Literature.

Diagnostic work-up of IgG4-related disease (IgG4-RD) pleural involvement is a complex task, as there is a broad spectrum of differential diagnoses to consider. We report the case of a patient presenting with relapsing pleural effusion, discussing the main challenges for achievement of a definite diagnosis. A 63-year-old man was admitted for pleural effusion prevalent on the ride side, initially labeled as idiopathic non-specific pleuritis, based on tissue evaluation after a medical thoracoscopy. He was started on steroids with initial improvement, but a later CT scan showed a relapse of pleural effusion associated with diffuse pleural thickening; a subsequent surgical pleural biopsy revealed features suggestive for IgG4-RD, with a marked increase of IgG4 positive plasma cells. High IgG4 serum levels were also found. The present case underlines the importance of increasing awareness of this potential condition among physicians in order to properly guide the diagnostic work-up, as it is likely that IgG4-RD accounts for a proportion of patients with pleural effusions, labeled as idiopathic. In particular, in patients with unexplained pleural effusion, IgG4-RD should be included among differential diagnoses when lymphoplasmacytic infiltration is observed, and a multidisciplinary interaction between clinicians and pathologists appears crucial for an accurate diagnosis and an appropriate management.

Inferring Latent Domains for Unsupervised Deep Domain Adaptation.

Unsupervised Domain Adaptation (UDA) refers to the problem of learning a model in a target domain where labeled data are not available by leveraging information from annotated data in a source domain. Most deep UDA approaches operate in a single-source, single-target scenario, i.e., they assume that the source and the target samples arise from a single distribution. However, in practice most datasets can be regarded as mixtures of multiple domains. In these cases, exploiting traditional single-source, single-target methods for learning classification models may lead to poor results. Furthermore, it is often difficult to provide the domain labels for all data points, i.e. latent domains should be automatically discovered. This paper introduces a novel deep architecture which addresses the problem of UDA by automatically discovering latent domains in visual datasets and exploiting this information to learn robust target classifiers. Specifically, our architecture is based on two main components, i.e. a side branch that automatically computes the assignment of each sample to its latent domain and novel layers that exploit domain membership information to appropriately align the distribution of the CNN internal feature representations to a reference distribution. We evaluate our approach on publicly available benchmarks, showing that it outperforms state-of-the-art domain adaptation methods.

Early histologic findings of pulmonary SARS-CoV-2 infection detected in a surgical specimen.

Despite the current pandemic season, reports on pathologic features of coronavirus disease 19 (Covid-19) are exceedingly rare at the present time. Here we describe the pathologic features of early lung involvement by Covid-19 in a surgical sample resected for carcinoma from a patient who developed SARS-CoV-2 infection soon after surgery. The main histologic findings observed were pneumocyte damage, alveolar hemorrhages with clustering of macrophages, prominent and diffuse neutrophilic margination within septal vessels, and interstitial inflammatory infiltrates, mainly represented by CD8+ T lymphocytes. These features are similar to those previously described in SARS-CoV-1 infection. Subtle histologic changes suggestive pulmonary involvement by Covid-19 may be accidentally encountered in routine pathology practice, especially when extensive sampling is performed for histology. These findings should be carefully interpreted in light of the clinical context of the patient and could prompt a pharyngeal swab PCR test to rule out the possibility of SARS-CoV-2 infection in asymptomatic patients.

Author Correction: WWP2 regulates pathological cardiac fibrosis by modulating SMAD2 signaling.

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

WWP2 regulates pathological cardiac fibrosis by modulating SMAD2 signaling.

Cardiac fibrosis is a final common pathology in inherited and acquired heart diseases that causes cardiac electrical and pump failure. Here, we use systems genetics to identify a pro-fibrotic gene network in the diseased heart and show that this network is regulated by the E3 ubiquitin ligase WWP2, specifically by the WWP2-N terminal isoform. Importantly, the WWP2-regulated pro-fibrotic gene network is conserved across different cardiac diseases characterized by fibrosis: human and murine dilated cardiomyopathy and repaired tetralogy of Fallot. Transgenic mice lacking the N-terminal region of the WWP2 protein show improved cardiac function and reduced myocardial fibrosis in response to pressure overload or myocardial infarction. In primary cardiac fibroblasts, WWP2 positively regulates the expression of pro-fibrotic markers and extracellular matrix genes. TGFß1 stimulation promotes nuclear translocation of the WWP2 isoforms containing the N-terminal region and their interaction with SMAD2. WWP2 mediates the TGFß1-induced nucleocytoplasmic shuttling and transcriptional activity of SMAD2.