Quantification and automated characterization of mucus plug pathology in asthmatics

项目摘要 粘液堵塞长期以来一直与严重哮喘的急性和致命的呼吸道事件有关，但我们已经 最近发现慢性粘液阻塞在哮喘患者中很常见，并且是机械性的。 与气流受损和病情恶化有关。特别是在基线分析中 在哮喘患者的CT肺扫描中，我们发现呼吸道粘液塞高度堵塞。 流行，持续多年，经常发生，没有咳嗽和痰的症状，并且很强烈 与气流阻塞有关。然而，粘液的放射学特征尚不清楚。 堵塞会导致严重的气流阻塞，部分原因是CT上粘液堵塞的详细特征 扫描是非常劳动密集型的，需要训练有素的胸科放射科医生进行评估。在目标1中 在这一应用中，我们建议检验假设，根据我们的初步数据，即三种射线照片 粘液塞子的特征-粘液塞子体积、近端塞子的数量和阻塞的呼吸道树的分数 -所有人都预测气流阻塞会恶化。此外，我们还建议将气道树转换为 一种由电阻元件组成的网络，在该网络中，整个树的有效电阻分别在有无的情况下进行计算 可以确定粘液塞对呼吸道阻力的相对贡献。在目标2中， 我们的目标是通过开发一种新的方法来显著降低CT扫描中粘液堵塞的量化障碍 基于卷积神经网络的粘液塞自动分割算法。我们相信我们的 这些发现将使我们能够识别出一大批慢性重度“粘液高”哮喘患者。 提出新的粘液靶向治疗的可能性，以改善气流和其他疾病结局 患者的子集。

Project Summary Mucus plugging has long been implicated in acute and fatal respiratory events in severe asthma, but we have recently shown that chronic mucus plugging is common in asthmatic patients and appears mechanistically linked with both impaired airflow and worsening disease severity. In particular, in analyses of baseline computed tomography (CT) lung scans in asthmatic patients, we found that airway mucus plugs are highly prevalent, persist for many years, frequently occur without cough and sputum symptoms, and are strongly associated with airflow obstruction. However, it is unknown what radiographic characteristics of mucus plugging cause severe airflow obstruction, in part because detailed characterization of mucus plugs on CT scan is extremely labor intensive and requires highly trained thoracic radiologists for assessment. In Aim 1 of this application, we propose to test the hypotheses, informed by our preliminary data, that three radiographic features of mucus plugs— mucus plug volume, number of proximal plugs, and fraction of airway tree occluded —all predict worsening airflow obstruction. We additionally propose that the airway tree can be converted into a network of resistive elements in which the effective resistance of the entire tree is computed with and without mucus plugs, and the relative contribution of mucus plugs to airway resistance can be determined. In Aim 2, we aim to substantially lower the barrier to quantification of mucus plugging on CT scans by developing an automated, convolutional neural network-based algorithm for mucus plug segmentation. We believe that our findings will allow the identification of a large subset of patients with chronic severe “mucushigh” asthma and raise possibilities for novel mucus-targeted treatments to improve airflow and other disease outcomes in this subset of patients.