A CT-Based Multiparametric Imaging Biomarker for Assessment of Pediatric Interstitial Lung Disease

项目摘要 儿童（弥漫性）间质性肺病影响婴儿、儿童和青少年， 表现为呼吸困难、低氧血症和呼吸损害，导致高 发病率和死亡率或幸存者的终身后遗症。尽管在这方面有所改善， 过去15年对小儿弥漫性肺病的认识和干预 年，通过以下方法确认和表征儿童的现行护理标准 薄层胸部计算机断层扫描（CT）通过视觉的进展有限， 评估本身是主观的，并受到读者之间的差异， 定义特定的特征发现。这可能会限制早期诊断的准确性。 和进行性疾病，并转化为更耗时和负担 临床评价一种自动化和客观的方法来量化常见的 在ILD中观察到的放射学肺部CT模式，特别是在儿科中， 通过标准化和简化图像，获得更可靠的信息并降低成本- 基于评估。 在这项拨款提案中，英比奥公司，在开发、商业化和 实现成像生物标志物软件的监管批准，建议开发一个 用于量化肺部CT纹理的全自动软件应用程序。协作 利用洛杉矶儿童医院的临床和研究专长， （CHLA）是一家大型的安全网儿科医院。具体目标是：（1） 开发儿科弥漫性肺受试者胸部CT扫描数据库 疾病和对照受试者沿着放射学纹理的专家注释，以及2） 开发并验证基于深度学习的集成儿科/成人算法， 量化放射性肺纹理（即，DeepLTA）。在完成目标后， 最终的算法将具有广泛的适用性和普适性，以检测实质 成人和儿童ILD的CT纹理。这将使第二阶段提交和 集成该软件，以增强儿科获得最先进的定量 用于改善预后、诊断和治疗反应的医学成像分析 评估小儿间质性肺疾病。

PROJECT SUMMARY Childhood (diffuse) interstitial lung diseases affects infants, children, and teens and manifests as dyspnea, hypoxemia, and respiratory compromise resulting in high morbidity and mortality or life-long sequelae for survivors. Despite improvements in the understanding of and interventions for pediatric diffuse lung disease in the past 15 years, the current standard-of-care for confirmation and characterization of chILD by thin-section chest computed tomography (CT) has had limited progress through visual assessments that are inherently subjective and suffer from inter-reader variability in defining the specific characteristic findings. This can limit accurate diagnosis of early and progressive disease and translates to a more time-consuming and burdensome clinical evaluation. An automated and objective approach to quantify common radiological lung CT patterns observed in ILDs specifically in pediatrics would provide more reliable information and reduce costs by standardizing and streamlining image- based assessment. In this grant proposal, Imbio Inc., an industry leader in developing, commercializing, and achieving regulatory approval of imaging biomarker software, proposes to develop a fully-automated software application for quantifying lung CT textures. The collaboration leverages the clinical and research expertise at the Children’s Hospital Los Angeles (CHLA) which is a large, safety-net pediatric hospital. The specific aims are 1) to develop a data repository of chest CT scans in subjects with pediatric diffuse lung disease and control subjects along with expert annotations of radiologic textures and 2) develop and validate an integrated pediatric/adult deep-learning-based algorithm to quantify radiological lung textures (i.e., DeepLTA). Upon completion of the aims, the final algorithm will have broad applicability and generalizability to detect parenchymal CT textures in both adult and pediatric ILDs. This will enable a Phase II submission and integration of this software to enhance pediatric access to state-of-the-art quantitative medical imaging analysis for improved prognosis, diagnosis, and therapy response assessment in pediatric interstitial lung disease.