利用電腦斷層導出肺臟含氣特徵之慢性阻塞性肺病肺功能指標計算與病變區域偵測

Pulmonary functional index calculation and pathological area detection for chronic obstructive pulmonary disease using computed tomography image-derived air content characteristics of lung

戴柔爾(Jou-Erh Tai)

指導教授 ： 莊濬超

共同指導教授 ： 施政廷(Cheng-Ting Shih)

中山醫學大學/醫學科技學院/醫學影像暨放射科學系/碩士(2021年)

https://doi.org/10.6834/csmu202100132

摘要

隨著科技進步，工廠與車輛排放之廢氣與日俱增，以及菸草的高度盛行，均使得肺臟在呼吸的過程中吸入過多有害物質並積聚於肺內而無法排出，最終導致慢性阻塞性肺病（Chronic obstructive pulmonary disease, COPD）的產生。對於COPD有一定嚴重程度的患者而言，由於無法良好配合肺功能檢查（pulmonary function test, PFT）的流程，使得檢查結果失去客觀評估的能力，此外PFT亦無法提供肺臟受損區域分布。因此本研究希望透過將胸部電腦斷層影像轉換為肺部含氣量分布，根據含氣特徵預測肺功能參數，並標示出肺臟內正常及異常區域。
本研究從美國國家衛生研究院（National Institutes of Health, NIH）資助所建立肺組織研究聯合會（Lung tissue research consortium, LTRC）的組織切片暨影像資料庫，篩選正常人與COPD患者共167名，其中有38名具有雙相位CT影像。並從CT影像當中圈選出空氣與軟組織有興趣區域(region of interest, ROI)，搭配雙組成模型（two-compartment model, TCM）計算各體素之空氣體積分率（air volume fraction, AV/TV），並依據相同的步進計算相對空氣體積直方圖。接著透過LASSO（least absolute shrinkage and selection operator）回歸進行含氣特徵篩選，並使用篩選到的含氣特徵對肺功能參數進行預測，以及標示出肺內正常及異常區域。
結果顯示即使在不同閾值中，使用含氣特徵預測肺功能在吸氣相中，用力呼氣一秒量（forced expiratory volume in one second, FEV1）相關係數最佳可達0.75以上。而流速容積比（forced expiratory fraction in one second, FEV1/FVC）與用力呼氣一秒量預測值（percentage of predicted forced expiratory volume in 1 second, FEV1(%predicted)）則可達0.82以上。如使用吐氣相FEV1與FEV1預測值相關係數最佳可達0.83，而FEV1/FVC則可達0.9。而當使用雙相位進行預測時，FEV1與FEV1預測值相關係數最佳可達0.85以上，而FEV1/FVC一樣可達0.9。且本研究針對LASSO從吸氣相篩選到的含氣特徵發現，特徵主要分布在90%與98%左右，因此本研究亦將其視為可代表為正常與異常的肺部區域。
我們認為使用含氣特徵能夠良好藉由單相位預測患者肺功能參數，並可從CT影像中標示出肺臟含氣異常之區域。

關鍵字

慢性阻塞性肺病 ； 肺功能檢查 ； 電腦斷層 ； 雙組成模型

並列摘要

With the advancement of technology, the exhaust emissions from factories and vehicles are increasing day by day, as well as the prevalence of tobacco. The lung inhale too much harmful substances during breathing and accumulate in the lung, which can’t be discharged, and eventually lead to the occurrence of chronic obstructive pulmonary disease (COPD). For patients with severe COPD, because they can’t cooperate with the process of pulmonary function test (PFT), the test results lose the ability of objective evaluation. In addition, PFT can’t provide the distribution of damaged areas of the lung. Therefore, the purpose of this study is to convert chest computed tomography (CT) images into lung air content distribution, predict lung function parameters according to air content features, and mark normal and abnormal areas in the lung.
In this study, 167 normal subjects and COPD patients were screened from the tissue section and imaging database of the lung tissue research consortium established by the National Institutes of Health. Among them, 38 patients had dual phase CT images. The region of interest of air and soft tissue was selected from the CT image, and the air volume fraction of each voxel was calculated with the two-compartment model, and the relative air volume histogram was calculated according to the same step. Then use least absolute shrinkage and selection operator (LASSO) regression to select air content features. Then use air content features to mark normal and abnormal areas.
The results showed that even in different thresholds, the best correlation coefficient of forced expiratory volume in one second (FEV1) was more than 0.75 in inspiratory phase, and the forced expiratory fraction in one second (FEV1/FVC) and percentage of predicted forced expiratory volume in 1 second (FEV1(%predicted)) was more than 0.82. The best correlation coefficient between FEV1 and predictive value of predicted FEV1 value by exhalation phase can reach 0.83, while FEV1/FVC can reach 0.9. When using Dual-phases prediction, the best correlation coefficient between FEV1 and predicted FEV1 value is more than 0.85, and FEV1/FVC is as high as 0.9.
In addition, this study finds that the air content features selected by LASSO from the inspiratory phase are mainly distributed around 90% and 98%. Therefore, this study also regards it as a normal and abnormal lung area.
We concluded that air content features can be used to predict lung function parameters by single phase, and the abnormal areas of lung can be identified from CT images.

並列關鍵字

chronic obstructive pulmonary disease ； lung function test ； computed tomography ； two-compartment model

參考文獻

[1] Brook, R. D., Rajagopalan, S., Pope III, C. A., Brook, J. R., Bhatnagar, A., Diez-Roux, A. V., ... Kaufman, J. D. (2010). Particulate matter air pollution and cardiovascular disease: an update to the scientific statement from the American Heart Association. Circulation, 121(21), 2331-2378.

[2] Fabbri, L. M., Rabe, K. F. (2007). From COPD to chronic systemic inflammatory syndrome?. The Lancet, 370(9589), 797-799.

[3] Mannino, D. M., Watt, G., Hole, D., Gillis, C., Hart, C., McConnachie, A., ... Vestbo, J. (2006). The natural history of chronic obstructive pulmonary disease. European Respiratory Journal, 27(3), 627-643.

[4] Celli, B. R., Cote, C. G., Marin, J. M., Casanova, C., Montes de Oca, M., Mendez, R. A., ... Cabral, H. J. (2004). The body-mass index, airflow obstruction, dyspnea, and exercise capacity index in chronic obstructive pulmonary disease. New England Journal of Medicine, 350(10), 1005-1012.

[5] Sherman, C. B. (1992). The health consequences of cigarette smoking: pulmonary diseases. Medical Clinics of North America, 76(2), 355-375.