發展乳房仿真模型應用於腫瘤硬度探測之研究

Development of Artificial Breast Model for Study of Tumor Hardness Exploration

張凱揚(Kai-Yang Chang)

指導教授 ： 顏炳郎

國立臺北科技大學/機電學院/自動化科技研究所/碩士(2007年)

摘要

乳癌是女性最常發生的惡性腫瘤之一，近年乳癌更名列女性十大癌症之中，然而乳房腫瘤良惡性的篩檢在乳癌防治上佔了很重要的角色。經李實驗結果，確認乳癌探測法與類神經網路具有分辨乳癌特性的可行性，本文進行設計並製作軟組織性質更近似真實乳房之仿真3D矽膠乳房模型，針對模型製作上做出了合理的假設與模型簡化，使模型較易於分析，再以仿真乳房模型利用下壓與橫推方向之兩軸力量感測器進行乳房探測實驗並評估其效能，並找出具有描述腫瘤特性且更有效的下壓方向與橫推方向力量特徵參數群與較佳之探測策略，進而利用類神經網路推算出腫瘤與周圍組織的硬度比。本研究使用11個樣本做驗證，結果顯示利用下壓方向與橫推方向力量特徵參數群做為網路輸入參數，對於不同硬度之腫瘤及對於大、小、深、淺之腫瘤特性都有很好的預測能力，能夠有效區分不同硬度之腫瘤。本研究結果可作為有限元素分析驗證之比較和臨床乳癌診斷之理論基礎。

關鍵字

乳房仿真模型 ； 腫瘤 ； 特徵萃取 ； 倒傳遞神經網路

並列摘要

The breast cancer is one of the malignant tumor that women happen most frequently. The breast cancer ranked in women's ten major cancers in recent years. And examine of breast tumor has taken very important role in the breast cancer prevented and cured. Confirm through experimental result of Li, breast palpation method and artificial neural network can distinguish characteristic of breast cancer. In this study, we designed and made the model of artificial 3D silicon of soft tissue that is more approximate then true breast. Use two axis force sense for exploration experiment and assess its efficiency. Find out describe the tumor characteristic more effective feature and best explore strategy to use ANN for forecast stiffness ratio of tumor with soft tissue. In this study, we used two axis force feature for input of ANN and 11 data sets for validation. Results depict that prediction accuracy can be achieved very well for inclusion properties, such as big, small, deep and shallow tumor. This result can compared with finite element analysis and makes to human basis of breast clinical diagnosis.

並列關鍵字

Artificial Breast Model ； Tumor ； Feature Extraction ； Back Propagation Neural Network

參考文獻

[2]. 李俊毅，類神經網路應用於觸診乳房硬塊之初步研究，碩士論文，國立台北科技大學自動化科技研究所，台北，2004。

[3]. Y. C. Fung, Biomechanics: mechanical properties of living tissues. New York, NY: Springer-Verlag, 1993.

[4]. Jianchao Zang, Yue Wang, ”Color Feature Based Finger Tracking for Breast Palpation Quantification”, Proceedings of IEEE International Conference on Robotics and Automation, Volume:3, pp.20-25, April 1997.

[7]. Du-Yih TSAI “Classification of Breast Tumors in Mammograms using a Neural Network:Utilization of Selected Features”Proceedings of 1993 International Joint Conference on Neural Networks, pp.967-970.

[8]. Elizabeth, G., Violante, A. and Francisco, C.,“Using neural networks for differential diagnosis for Alzheimer disease and Vascular dementia” Expert Systems with Applications, 14,pp.219-225, 1998.