請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/88850
標題: | 基於術前電腦斷層影像以深度學習方法預測充氣後腹部表面 Surface Prediction after Insufflation Based on Preoperative Computed Tomography Scan Image by Machine Learning Method |
作者: | 林聖邦 Sheng-Bang Lin |
指導教授: | 陳永耀 Yung-Yaw Chen |
關鍵字: | 腹腔鏡手術,術前規劃,開口佈局,深度學習, Laparoscopic Surgery,Surgical Planning,Port Placement,Machine Learning, |
出版年 : | 2023 |
學位: | 碩士 |
摘要: | 在現階段的微創手術術前規劃中,仍缺少腹腔充氣後腹部表面的資訊,造成術前判斷上的困難。目前,醫師僅能透過觀察電腦斷層影像,憑藉自身經驗及判斷完成術前規劃。然而,術前電腦斷層影像與手術中腹腔充氣後表面的差異,使得醫生難以準確地進行術前規劃及開口佈局。腹腔充氣預測模型的建立能夠解決上述的困境。過去關於腹腔充氣模型的研究仍有兩個問題尚未解決。首先,過去的研究大多採用物理方法去進行腹腔形變的模擬,然而使用這些方法時需要花費大量時間去建立腹部的物理模型並計算形變。第二是過去的研究鮮少對實際應用於人體的效果進行完整的評估。本研究選擇使用深度學習的方法,透過病患的生理特徵以及術前的電腦斷層影像來預測腹腔充氣後的表面形式。
本研究一共收集20個實際接受腹腔鏡手術的病患資料。病患的生理特徵以及電腦斷層影像將作為深度學習的輸入,輸出則是病患的充氣後腹部表面。預測出的腹部膨脹表面和臨床數據比較後,平均表面誤差為6.21 mm。本研究提出了一個由深度學習方法建立腹腔膨脹預測模型的方法,能夠預測出充氣後的腹部表面。 Surgical planning in laparoscopic surgery is challenging because the patient’s abdominal insufflation model is unavailable preoperatively. Currently, surgeons have to conduct surgical planning based on their judgments on the computed tomography (CT) images, while these images do not show the abdominal surface after insufflation. The discrepancies between the abdominal surfaces in the CT images and surgery lead to difficulty in surgical planning, especially the port placement. Studies on abdominal insufflation prediction have emerged to solve the problem. Previous studies often adopted physics-based methods to simulate the deformation of the abdominal cavity after insufflation. However, these methods are time-consuming when constructing physical models of the abdomen and computing the deformation. Besides, these studies rarely discussed their method’s performance on human data. In this study, we adopt the machine learning method to predict the abdominal surface after insufflation using the patient’s CT images and individual characteristics. We collect the data from 20 patients that underwent laparoscopic surgery. The patient’s individual characteristics and CT images are the input of the machine learning process, and the abdominal surface after insufflation is the output. The comparison between the model’s output and the clinical data returned an average error of 6.21 mm. This research proposed an abdominal insufflation prediction method that can predict the patient’s abdominal surface after insufflation based on a machine learning approach. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/88850 |
DOI: | 10.6342/NTU202303617 |
全文授權: | 未授權 |
顯示於系所單位: | 電機工程學系 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-111-2.pdf 目前未授權公開取用 | 5.31 MB | Adobe PDF |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。