A deep convolutional neural network architecture for interstitial lung disease pattern classification.
- 作者列表："Huang S","Lee F","Miao R","Si Q","Lu C","Chen Q
:Interstitial lung disease (ILD) refers to a group of various abnormal inflammations of lung tissues and early diagnosis of these disease patterns is crucial for the treatment. Yet it is difficult to make an accurate diagnosis due to the similarity among the clinical manifestations of these diseases. In order to assist the radiologists, computer-aided diagnosis systems have been developed. Besides, the potential of deep convolutional neural networks (CNNs) is also expected to exert on the medical image analysis in recent years. In this paper, we design a new deep convolutional neural network (CNN) architecture to achieve the classification task of ILD patterns. Furthermore, we also propose a novel two-stage transfer learning (TSTL) method to deal with the problem of the lack of training data, which leverages the knowledge learned from sufficient textural source data and auxiliary unlabeled lung CT data to the target domain. We adopt the unsupervised manner to learn the unlabeled data, by which the objective function composed of the prediction confidence and mutual information are optimized. The experimental results show that our proposed CNN architecture achieves desirable performance and outperforms most of the state-of-the-art ones. The comparative analysis demonstrates the promising feasibility and advantages of the proposed two-stage transfer learning strategy as well as the potential of the knowledge learning from lung CT data. Graphical Abstract The framework of the proposed two-stage transfer learning method.
间质性肺疾病 (ILD) 是指一组各种异常的肺组织炎症，早期诊断这些疾病模式对治疗至关重要。然而，由于这些疾病的临床表现相似，很难做出准确的诊断。为了帮助放射科医生，开发了计算机辅助诊断系统。此外，深度卷积神经网络 (CNNs) 的潜力也有望在近年来的医学图像分析中发挥作用。在本文中，我们设计了一种新的深度卷积神经网络 (CNN) 架构来实现 ILD 模式的分类任务。此外，我们还提出了一种新的两阶段迁移学习 (TSTL) 方法来解决训练数据缺乏的问题, 它利用从足够的纹理源数据和辅助未标记的肺 CT 数据中学到的知识到目标域。我们采用无监督的方式学习未标记数据，通过这种方式优化了由预测置信度和互信息组成的目标函数。实验结果表明，我们提出的 CNN 架构达到了理想的性能，优于大多数最先进的架构。对比分析证明了所提出的两阶段迁移学习策略的可行性和优势，以及从肺 CT 数据中学习知识的潜力。图形抽象提出的两阶段迁移学习方法的框架。
METHODS:OBJECTIVES:To asses the clinical course in RA-related interstitial lung disease (RA-ILD) patients with and without rituximab (RTX). The influence of other variables was also evaluated. METHODS:A longitudinal multicentre study was conducted in RA diagnosed with ILD from 2007 until 2018 in Madrid. Patients were included in a registry [pNEumology RhEumatology Autoinmune diseases (NEREA)] from the time of ILD diagnosis. The main endpoint was functional respiratory impairment (FI), when there was a decline ≥5% in the predicted forced vital capacity compared with the previous one. Pulmonary function was measured at baseline and in follow-up visits every 6-12 months. The independent variable was therapy with RTX. Covariables included sociodemographic, clinical, radiological and other therapies. Survival techniques were used to estimate the incidence rate (IR) and 95% CI of functional impairment, expressed per 100 patient-semesters. Cox multivariate regression models were run to examine the influence of RTX and other covariates on FI. Results were expressed as the hazard ratio (HR) and CI. RESULTS:A total of 68 patients were included. FI occurred in 42 patients [IR 23.5 (95% CI 19, 29.1)] and 50% of them had FI within 1.75 years of an ILD diagnosis. A multivariate analysis showed that RTX exposure resulted in a lower risk of FI compared with non-exposure [HR 0.51 (95% CI 0.31, 0.85)]. Interstitial pneumonia, glucocorticoids, disease activity and duration also influenced FI. CONCLUSION:RA-ILD patients deteriorate over time, with the median time free of impairment being <2 years. Patients exposed to RTX had a higher probability of remaining free of FI compared with other therapies. Other factors have also been identified. Key words: rheumatoid arthritis, interstitial lung disease, observational study, rituximab and prognosis
METHODS:The safety of anti-programmed cell death 1 (PD-1) antibody for patients with preexisting interstitial lung disease (ILD) remains unknown. The aim of this study was to evaluate the dependence of preexisting ILD on anti-PD-1 antibody-induced pneumonitis in non-small cell lung cancer (NSCLC) patients. We retrospectively reviewed the association of preexisting ILD with the incidence, radiographic pattern, and outcome of pneumonitis in NSCLC patients receiving anti-PD-1 antibody. A total of 331 patients were included in this study. Of these patients, 17 had preexisting ILD. The incidence of pneumonitis was higher among the patients with preexisting ILD than among those without preexisting ILD (29% vs. 10%, P = 0.027). The distributions of the CT appearances at the onset of anti-PD-1 antibody-induced pneumonitis were as follows: for the patients with preexisting ILD, two patients (40%) had diffuse alveolar damage (DAD), one patient each with organizing pneumonia-like (OP), hypersensitivity pneumonitis (HP), and other patterns (20% each); for the patients without preexisting ILD, 19 patients (61%) had OP, 8 (26%) had HP, 3 (10%) had DAD, and 1 (3.2%) had other patterns. The median onset time from the initiation of anti-PD-1 antibody treatment until the development of pneumonitis was 1.3 months (range 0.3–2.1 months) for the patients with preexisting ILD and 2.3 months (range 0.2–14.6 months) for the patients without preexisting ILD. Careful attention to the development of pneumonitis is needed, especially within the first 3 months after the start of anti-PD-1 antibody treatment, when using anti-PD-1 antibody to treat patients with preexisting ILD.
METHODS::Bacteria of the Burkholderia cepacia complex (Bcc) are ubiquitous multidrug resistant organisms and opportunistic pathogens capable of causing life threatening lung infections among cystic fibrosis (CF) patients. No effective therapies are currently available to eradicate Bcc bacteria from CF patients, as these organisms are inherently resistant to the majority of clinically available antimicrobials. An immunoproteomics approach was used to identify Bcc proteins that stimulate the humoral immune response of the CF host, using bacterial cells grown under conditions mimicking the CF lung environment and serum samples from CF patients with a clinical record of Bcc infection. 24 proteins of the Bcc strain B. cenocepacia J2315 were identified as immunoreactive, 19 here reported as immunogenic for the first time. Ten proteins were predicted as extracytoplasmic, 9 of them being conserved in Bcc genomes. The immunogenic Bcc extracytoplasmic proteins are potential targets for development of novel therapeutic strategies and diagnostic tools to protect patients against the onset of chronic Bcc lung infections.