對比劑微灌流影像(contrast-enhanced perfusion)是一種可以有效診斷肺部疾病的方式，對病患注射對比劑之後，在經由醫學影像設備(CT、MRI…)觀察對比劑流進及流出目標區的狀況，便可以判定患者的血流是否正常以便進行後續的診斷。
為了量化分析肺部組織血流的狀況，取得血流(pulmonary blood flow, PBF)，血量(pulmonary blood volume, PBV)，以及平均通過時間(mean transit time, MTT)與到達時間(arriving time)等各種參數，需要統計及分析左肺及右肺的組織影像，因此必須從整張胸腔影像中，圈選出左肺及右肺的ROI。然而整體肺部輪廓跟肺動脈比起來，邊界更長並更為曲折複雜，且不同人圈選之輪廓可能會有很大的差異。再者還必須去掉血管的部分，平均每圈選一次都得花上3~5分鐘，且若失敗可能得重新開始，因此導入自動化選取肺部ROI是有其必要性的。
本文利用自動肺部區域定位，針對區域內影像做強化對比與分類，多重梯度向量流動態輪廓演算法，自動分類法取得動脈血管門檻值。經由這四個步驟來取得肺部ROI，並且透過模擬真實狀況的方式驗證本方法的正確性，再由專業的放射師來評比取得的ROI，本方法所取得的ROI除了方便醫療及研究人員量化患者的肺部血流狀況外，亦可以應用在許多方面的研究上。

Contrast-enhanced perfusion imaging is an effective method for diagnosis of lung diseases. After injection of contrast agent, the contrast agent washing-in and washing-out of tissues is observed through medical equipments such as CT and MRI. Then the blood flow condition of the patient can be determined and provided for the follow-up diagnosis. The quantification analysis of blood flow condition of lung tissues is to obtain perfusion parameters such as pulmonary blood flow (PBF), pulmonary blood volume (PBV), mean transit time (MTT) and arriving time. It needs statistic calculation and analysis on the images of right and left lung parenchyma, therefore ROIs of right and left lung should be segmented from the entire image of thoracic cavity. But the whole lung boundary is longer and more complex than pulmonary artery and a big difference exists among the boundaries segmented by various persons (inter-observer variation). It also needs to exclude blood vessels. The manual process takes about 3~5 minutes for each segmentation. Therefore it is necessary to implement automatic segmentation for lung ROI.
This study adopts 4 steps to obtain the lung ROI: (1) automated lung area location. (2) automated lung area enhancing. (3) multi GVF snake model. (4) artery thresholding. To verify the feasibility and accuracy of this method, we have tested our method on both simulated data and patients. These results show that this method is feasible and good enough for the segmentation of lung ROIs. ROIs obtained by this method can be used for quantifying the blood flow condition of the patient’s lung. It can be a convenient and useful tool for quantitative analysis of pulmonary perfusion.

1. Amundsen T, Kvaerness J, Jones RA, Waage A, Bjermer L, Nilsen G, Haraldseth O. Pulmonary embolism: detection with MR perfusion imaging of lung--a feasibility study. Radiology 1997;203(1):181-185.
2. Bram van Ginneken* AFF, Joes J. Staal, Bart M. ter Haar Romeny, and Max A. Viergever. Active Shape Model Segmentation With Optimal Features. IEEE TRANSACTIONS ON MEDICAL IMAGING 2002.
3. Ray N, Acton ST, Altes T, de Lange EE, Brookeman JR. Merging parametric active contours within homogeneous image regions for MRI-based lung segmentation. IEEE Trans Med Imaging 2003;22(2):189-199.
4. Yi-Ru Lin M-TWT-YHS-YTH-WCVMMC-YCH-BP. Comparison of arterial spin labeling and first-pass dynamic contrast-enhanced MR imaging in the assessment of pulmonary perfusion in humans: The inflow spin-tracer saturation effect. Volume 52; 2004. p 1291-1301.
5. David L. Levin QCMZRREHH. Evaluation of regional pulmonary perfusion using ultrafast magnetic resonance imaging. Volume 46; 2001. p 166-171.
6. Roberts DA, Gefter WB, Hirsch JA, Rizi RR, Dougherty L, Lenkinski RE, Leigh JS, Jr., Schnall MD. Pulmonary Perfusion: Respiratory-triggered Three-dimensional MR Imaging with Arterial Spin Tagging-Preliminary Results in Healthy Volunteers. Radiology 1999;212(3):890-895.
7. Vu M. Mai SSB. MR perfusion imaging of pulmonary parenchyma using pulsed arterial spin labeling techniques: FAIRER and FAIR. Journal of Magnetic Resonance Imaging 1999;9(3):483-487.
8. Stegmann MB. Active Appearance Models: Theory, Extensions and Cases. Informatics and Mathematical Modelling, Technical University of Denmark, DTU 2000.
9. Informatics and Mathematical Modelling TUoD, DTU. http://www2.imm.dtu.dk/~aam/.
10. Stegmann MB. The AAM-API: An Open Source Active Appearance Model Implementation. Medical Image Computing and Computer-Assisted Intervention - MICCAI 2003, 6th Int 2003.
11. Michael Kass AW, Demetri TerzopoulosDOI. Snakes: Active contour models. International Journal of Computer Vision 1988.
12. Chenyang X, Jerry LP. Gradient Vector Flow: A New External Force for Snakes. Proceedings of the 1997 Conference on Computer Vision and Pattern Recognition (CVPR '97): IEEE Computer Society; 1997.
13. http://iacl.ece.jhu.edu/projects/gvf/. The Image Analysis and Communications Lab at Johns Hopkins University.
14. http://cms03p.vghks.gov.tw/Chinese/MainSite/. Kaohsiung Veterans General Hospital.
15. Mell M, Tefera G, Thornton F, Siepman D, Turnipseed W. Clinical utility of time-resolved imaging of contrast kinetics (TRICKS) magnetic resonance angiography for infrageniculate arterial occlusive disease. J Vasc Surg 2007;45(3):543-548; discussion 548.
16. Tateishi M, Koide M, Mizukami A. Tetralogy of Fallot with atypical coarctation of the aorta and carotid arterial stenosis due to fibromuscular dysplasia. Cardiol Young 2007;17(6):689-690.
17. http://www.fontanoperation.com/fontan.htm. Fontan Operation Homepage.
18. Metz CE. Basic principles of ROC analysis. Semin Nucl Med 1978;8(4):283-298.
19. Ohno Y, Hatabu H, Higashino T, Takenaka D, Watanabe H, Nishimura Y, Yoshimura M, Sugimura K. Dynamic perfusion MRI versus perfusion scintigraphy: prediction of postoperative lung function in patients with lung cancer. AJR Am J Roentgenol 2004;182(1):73-78.
20. Roman KS, Kellenberger CJ, Farooq S, MacGowan CK, Gilday DL, Yoo SJ. Comparative imaging of differential pulmonary blood flow in patients with congenital heart disease: magnetic resonance imaging versus lung perfusion scintigraphy. Pediatr Radiol 2005;35(3):295-301.