利用胎兒3D超音波影像計算臟器體積之研究

Computing Fetal Organ Volume Using 3D Ultrasound Image

許哲誠(Che-Cheng Hsu)

指導教授 ： 張瑞峰

國立臺灣大學/電機資訊學院/資訊工程學研究所/碩士(2020年)

https://doi.org/10.6342/NTU202003806

摘要

隨著醫學影像發展，超音波診斷科技已成為孕婦檢查先天性胎兒畸形不可缺少的工具。然而，過去婦科醫生在產前檢查中使用二維超音波影像，無法提供更詳細的胎兒資訊。近年來隨著三維超音波篩查科技的發展，胎兒三維超音波可以提供全面的器官資訊來評估胎兒的狀態，胃體積的估計是確定胎兒狀態的一個重要特徵，準確的胃體積計算可以幫助婦科醫生做出决定，雖然三維超音波影像能提供更詳細的資訊，但超音波雜訊會影響婦科醫生診斷的準確性。此外，胃體積的估計是一個依賴婦科醫生的過程，不同的婦科醫生可能會有所不同。因此，在本研究中，我們提出一個全自動的電腦輔助體積估測系統，以協助婦科醫師獲得胎兒胃體積。該系統由圖像處理、胃區域定位和胃體積估計組成，在圖像處理中，採用了anisotropic和sigmoid濾波器來消除超音波雜訊，提高了影像的對比度，然後採用候選區域生成和胃選擇相結合的胃區域定位方法，對胃進行定位，分割出粗糙區域。最後，利用Level-set和體積計算相結合的胃體積估計方法，對胃區域進行細化，計算出寶貴的胃體積。由實驗結果，該系統對23例患者進行了處理，胃體積和胃直徑的總體準確率分別達到73.9%和82.6%。綜上所述，基於實驗結果，本系統可以為胎兒三維超音波數據提供可靠與精準的胎兒胃體積量測。它克服了人工操作對超音波結果解釋的局限性。

關鍵字

胎兒 ； 超音波 ； 器官大小 ； 電腦輔助診斷

並列摘要

With the high development of medical imaging, the ultrasound (US) diagnostic technology has become an indispensable tool for pregnant women to check for congenital fetal abnormalities. However, in the past, gynecologists use two-dimensional US images in the prenatal examination, which cannot provide more detailed information with fetal. With the development of three- dimensional (3-D) US screening in recent, the fetus 3-D US could provide comprehensive organ information to evaluate the status of the fetus. The stomach volume estimation is a crucial feature for determining the fetus's status. The accuracy of stomach volume calculation can help the gynecologist make a decision. Although the 3-D US image can provide more detailed information, the US noise would affect the accuracy of gynecologist diagnosis. Moreover, the stomach volume estimation is the gynecologist-dependent process. The estimation result may be different by the different gynecologists. Therefore, in this study, a fully automatic computer-aided volume estimation system is proposed for helping the gynecologist to obtain the fetal stomach volume. The proposed system is composed of the image processing, the stomach region locating, and stomach volume estimation. In image processing, the anisotropic and sigmoid filters are used to eliminate the US noise and enhance the image contrast. Then, the stomach region locating composed of candidate region generation and stomach selection is employed to locate the stomach and segment out the rough region. Finally, the stomach volume estimation made up of the level-set method, and volume calculation is utilized to refine the stomach region and compute the precious stomach volume. In the experimental results, the proposed system is processed with 23 cases, and the overall accuracy of stomach volume and diameter achieve 73.9% and 82.6%. In conclusion, based on the experiment result, our system can offer reliable and precise fetal stomach volume measurements for fetal 3-D ultrasound data. It can overcome the limitation of manual operation to interpret ultrasound results.

並列關鍵字

fetal ； ultrasound ； volume diagnosis

參考文獻

[1] U. M. Reddy, R. A. Filly, and J. A. Copel, "Prenatal imaging: ultrasonography and magnetic resonance imaging," (in eng), Obstet Gynecol, vol. 112, no. 1, pp. 145-57, Jul 2008, doi: 10.1097/01.AOG.0000318871.95090.d9.

[2] W. Lee et al., "Fetal growth parameters and birth weight: their relationship to neonatal body composition," (in eng), Ultrasound Obstet Gynecol, vol. 33, no. 4, pp. 441-6, Apr 2009, doi: 10.1002/uog.6317.

[3] I. Goldstein, E. A. Reece, S. Yarkoni, M. Wan, J. L. J. Green, and J. C. Hobbins, "Growth of the fetal stomach in normal pregnancies," Obstetrics and Gynecology, Article vol. 70, no. 4, pp. 641-644, 1987. [Online]. Available: https://www.scopus.com/inward/record.uri?eid=2-s2.0-0023198358 partnerID=40 md5=952bb02293c74a24acd12a891ea59409.

[4] T. Hata, H. Tanaka, J. Noguchi, E. Inubashiri, T. Yanagihara, and S. Kondoh, "Three-Dimensional Sonographic Volume Measurement of the Fetal Stomach," Ultrasound in Medicine Biology, vol. 36, no. 11, pp. 1808-1812, 2010/11/01/ 2010, doi: https://doi.org/10.1016/j.ultrasmedbio.2010.06.013.

[5] B. Rahmatullah, A. T. Papageorghiou, and J. A. Noble, "Image Analysis Using Machine Learning: Anatomical Landmarks Detection in Fetal Ultrasound Images," in 2012 IEEE 36th Annual Computer Software and Applications Conference, 16-20 July 2012 2012, pp. 354-355, doi: 10.1109/COMPSAC.2012.52.