適用本國人肺氣腫電腦輔助偵測系統之開發

Development of Computer Aided Detection System for Emphysema in Native People

杜業明(YE-MING DU)

指導教授 ： 蘇振隆

中原大學/工學院/生物醫學工程研究所/碩士(2019年)

https://doi.org/10.6840/cycu201900406

摘要

目前肺氣腫診斷的標準仍然是肺功能檢查，但其存在許多不足。隨著多切式電腦斷層的出現，電腦斷層的功能成像變得更加方便與準確，可以在反映病變形態、結構的同時反映其生理變化。如何利用低劑量肺部電腦斷層掃描評判與分類有無肺氣腫疾病是一個值得探討的問題。國內尚無類似研究，故本論文主要是開發一適用本國人肺氣腫電腦輔助偵測系統供臨床之需。
本研究為探討低劑量肺部電腦斷層癌症篩檢影像，藉由開發之程式擷取肺部DICOM影像並進行影像二值化、閥值設定、連通元件偵測等切割出肺部區域，並計算出肺部異常值十五項參數與傳統肺功能檢查相比較，以獲取適合本國人肺氣腫嚴重程度相對應的區間參數。研究樣本以臨床180位受檢者同時接受肺功能檢查與低劑量肺部電腦斷層癌症篩檢，分二階段進行，第一階段抽出20個肺功能樣本分四個族群分別為正常、輕度、中度、嚴重各五個當作訓練樣本，與開發之程式產出的15項參數藉由相關係數分析與一致性係數分析找出最佳區間參數。第二階段藉由第一階段分析出的14項最佳區間參數套用至測試組160個樣本，其中肺功能檢查結果分別為正常組群126個，輕度 COPD患者族群13個，中度COPD患者族群18個，重度COPD患者族群3個，並探討其一致性。
第一階段研究結果顯示程式計算出吸氣與吐氣肺氣腫占全肺比率、吸氣與吐氣肺部密度、吸氣15、20、50、70、75百分位比像素指數與吐氣15、20、50、70、75百分位比像素指數等15項參數，與肺功能檢查FVC、FEV1、FEV1/FVC、FEF25-75四個結果參數，在Pearson相關係數正相關分別為0.21~0.36、0.45~0.76、0.49~0.82、0.59~0.87，負相關則分別為-0.33~-0.44、-0.66~-0.78、-0.70~-0.85、-0.65~-0.70，在Kappa值分析值為0.667~0.933皆屬於高度吻合。但在吐氣70百分位相關係數0.21~0.48為中低度相關， Kappa值為0.26屬於低度吻合度。故此參數位於第二階段納入。將第一階段高度吻合之14項參數及其區間臨界值，帶入測試組160個樣本分析出Kappa值為0.618~0.966皆屬於高度吻合。此外，將國外文獻15百分位數值區間套入本國180組樣本進行一致性分析Kappa值0.13為極低的吻合度。本研究在吸氣肺氣腫占全肺比率Pearson相關係數(r=-0.85)皆高於其他國外文獻。
綜合以上分析結果，臨床醫師可依照本系統平台產出14項參數結果，進行肺氣腫程度的參考判斷。因此在做低劑量肺部電腦斷層篩檢時，可雙向藉由此系統平台進行肺氣腫的檢測以利及早發現疾病進行相關治療。

關鍵字

影像處理 ； 低劑量肺部電腦斷層 ； 肺氣腫 ； 慢性阻塞性肺部疾病

並列摘要

Currently, pulmonary function examination (PFE) is standard for diagnosis of emphysema, but exist many deficiencies. With the advent of multi-slice technology, CT images are more convenient and accurate, can reflect morphological lesions and physiological changes. How to use low-dose lung CT (LDLCT) scan to evaluate and classify the presence or absence of emphysema disease is a question worth to study.
In this study, a low dose CT lung cancer screening system is developed based on image processing methods. By capture and bilevel DICOM formatted lung CT image, edge detection, and lung region segementation. And then calculate the value of fifteen lung abnormalities parameters which are used to compare with conventional pulmonary function tests to classify people with emphysema severity based on corresponding parameters. 180 subjects with both PFE test and receiving LDLCT lung cancer screening are employed thru two stages (training and testing). During training stage, 20 subjects divided into four groups (normal, mildly, moderate, and serious COPD, 5 of each). 4 parameters (FVC, FEV1, FEV1/FVC, FEF25-75) from PFE and 15 parameters (the proportion of suction and experiention emphysema accounted for the total lung ratio, inhalation and gas lung density, inhalation 15, 20, 50, 70, 75 percentile pixel index and gas 15, 20 , 50, 70, 75 percentibit ratio pixel index) from LDLCT were obtained, and then by the correlation coefficient analysis and consistency analysis parameters between of PET and LDLCT, the optimal interval parameters for different group were found. In second stage, the rest of 160 subjects which consist of 126 normal, 13 mild, 18 moderate, and 3 serious were used to evaluate the performance of this developed system.
The results of the first stage show the positive correlation of Pearson correlation coefficients between 15 parameters with FVC, FEV1, FEV1/FVC, and FEF25-75 are 0.21 to 0.36, 0.45 to 0.76, 0.49 to 0.82, and 0.59 to 0.87; and negative correlation are -0.33 to -0.44, -0.66 to - 0.78, -0.70 to -0.85, and -0.65 to - 0.70, respectively. These coefficients are significantly related and the analysis value of Kappa value is 0.667 to 0.933 (highly consistent). However, the 70 percent point correlation coefficient of 70 percent of gas is 0.21 to 0.48 for the medium-low correlation and the Kappa value of 70 percentire of gas is 0.26 for low fit. Thus, this parameter is excluded in testing stage. Using those 14 parameters apply to 160 samples, Kappa value is 0.618 to 0.966 which belong to highly consistent. Moreover, after applied 15 percentile section value from foreign literature to our 180 sets data for consistency analysis, the Kappa value is 0.13 (low fit). In this study, the Pearson correlation coefficient in the intake of foreign accounting pulmonary emphysema ratio (r = -0.85) is higher than other documents.
Based on the above analysis results, clinicians can make early detection and reference judgment on the degree of emphysema according to the 14 parameter results produced from low-dose pulmonary computed tomography by this developed system and make tearment in time.

並列關鍵字

Image processing ； Low dose lung computed tomography ； emphysema ； chronic obstructive pulmonary disease

參考文獻

[1] 行政院衛生署網址http://www.doh.gov.tw/cht2006/index_populace.aspx

[2] Makita, H., Nasuhara, Y., Nagai, K., Ito, Y., Hasegawa, M., Betsuyaku, T., ... & Nishimura, M. (2007). Characterisation of phenotypes based on severity of emphysema in chronic obstructive pulmonary disease. Thorax, 62(11), 932-937.

[3] Heard, Brian E. "A pathological study of emphysema of the lungs with chronic bronchitis." Thorax 13.2 (1958): 136-149.

[4] WYATT, J. P., FISCHER, V. W., & SWEET, H. C. (1962). Panlobular emphysema: anatomy and pathodynamics. Diseases of the Chest, 41(3), 239-259.

[5] Coxson, H. O., & Rogers, R. M. (2005). Quantitative computed tomography of chronic obstructive pulmonary disease1. Academic radiology, 12(11), 1457-1463.