利用機器學習技術於PET/CT上之肝癌電腦輔助診斷系統

PET/CT是核子醫學近年來發展的造影技術，其結合功能性及解剖性的資訊可供醫師用來診斷癌症、感染及發炎疾病、與癲癇。肝癌(HCC)為世界上第四常見的癌症，尤其在台灣，死亡人數與死亡率排名第二，發生率排名第三。本論文的目的為建立一個電腦輔助診斷系統，使用決策樹與支持向量機這兩種機器學習技術，藉由分析肝臟PET/CT影像進而判斷病人是否罹患肝癌。我們總共分析37位病人的PET/CT資料，包含20位肝癌與17位未罹患肝癌。使用決策樹的分類結果，與實際的診斷相比較，得到平均靈敏度(sensitivity)為84.667%、平均特異度(specificity)為96.667%、平均陽性預測值(positive predictive value)為90%。使用支持向量機的預測結果，與實際的診斷相比較，得到平均靈敏度(sensitivity)為85.34%、平均特異度(specificity)為83.34%、平均陽性預測值(positive predictive value)為83.32%。若未來有更多的筆資料加入分析，可以建立更高準確度的預測模型。本論文建立一個分析肝臟PET/CT影像的電腦輔助診斷系統，若配合自動化肝臟切割，則可輔助臨床醫師的診斷。

關鍵字

電腦輔助診斷系統； PET/CT ；肝癌；決策樹；支持向量機；影像處理

並列摘要

PET/CT is a recent development in nuclear medicine imaging techniques and it well combines functional and anatomical information, and has been used in the diagnosis of cancers, infections, inflammatory diseases and epilepsy. Hepatocellular cancer (HCC) is the fourth most common cancer in the world. In Taiwan, the death toll and mortality of HCC is the second, and the incidence is the third. The goal of this thesis is to build a computer aided diagnosis (CAD) system based on the decision tree classification model and the support vector machine (SVM) prediction model to identify HCC on PET/CT images. There are 37 patients including 20 patients diagnosed as HCC and the remaining 17 patients diagnosed as normal. Compare the prediction from decision tree classification model with practical diagnosis, the average sensitivity is 84.667%, average specificity is 96.667% and the average positive predictive value (PPV) is 90%. Compare the prediction from SVM with practical diagnosis, the average sensitivity is 85.34%, average specificity is 83.34% and the average PPV is 83.32%. More data would probably help to build a more precise prediction model. In the combination of automatic liver segmentation, we can implement a fully automatic computer aided diagnosis system to make a diagnosis providing help for clinicians.

並列關鍵字

Computer Aided Diagnosis ； PET/CT ； Hepatocellular cancer ； Decision tree ； Support Vector Machine ； image processing

參考文獻

[11] Chih-Wei Hsu, Chih-Chung Chang, and Chih-Jen Lin,“A Practical Guide to Support Vector Classication”,

[9] John M. Ollinger, Jeffrey A. Fessler, “Positron-Emission Tomography”, Signal Processing Magazine, IEEE, 14, 1, 43-55, 1997

[1] Kapoor V, McCook BM, Torok FS,“An introduction to PETCT imaging”, Radiographics, 24, 2, 523–43, 2004

[4] Sang-wook Leea, Soon Yuhl Namb, Ki Chun Imc, Jae Seung Kimc, Eun Kyung Choia, Seung Do Ahna, Sung Ho Parka, Sang Yoon Kimb, Bong-Jae Leeb, Jong Hoon Kima “Prediction of prognosis using standardized uptake value of 2-[(18)F] fluoro-2-deoxy-d-glucose positron emission tomography for nasopharyngeal carcinomas”, Radiotherapy and Oncology, 87, 2, 211–216, 2008

[5] Gerd Muehllehner, Joel S Karp, “Positron emission tomography”, Physics in Medicine and Biology, 51, 13, 2006

國際替代計量

利用機器學習技術於PET/CT上之肝癌電腦輔助診斷系統

全文下載

主題瀏覽