- 學位論文
結合有限元素法之乳癌腫瘤硬度類神經模型建立
Construction of An Artificial Neural Network Model for Breast Tumor Hardness Prediction by Combining Finite Element Method
摘要
一般區分乳房內良惡性腫瘤的主要臨床依據是依靠腫瘤位置、大小、形狀、硬度等資訊來判斷，而硬度與腫瘤類型，腫瘤與周圍組織的硬度比例之關聯性高，所以乳房內部軟組織與腫瘤間的相對硬度為判斷良惡性之主要依據。我們以有限元素法建立乳房與軟組織的有限元素模型，模擬乳房組織受到下壓時的應變情形，取代驗，大幅減少所需成本以及時間。再經由下壓模擬結果，得到不同腫瘤在受到下壓時，力與位移的參數做曲線擬合，針對曲線擬合結果採以特徵萃取，用以類神經網路之輸入訓練，建立類神經網路模型以預測乳癌腫瘤硬度。目前驗證的樣本中，經由類神經網路模型預測，在合理的誤差範圍內可以有效分辨腫瘤硬度，提供醫生在乳癌腫瘤檢測時額外的參考資訊。
並列摘要
Self breast examination through palpation in comparison to visualized diagnostic tool such as mammography or ultrasonic is in advanced of its convenience and low-cost. However, the lacking of physical models of soft tissues in breasts diminishes the accuracy of self breast examinations by women themselves. This paper describes constructing breast models based on finite element method to carry out an efficient model as a non-invasive diagnostic tool in the point of bio-engineering so as to reduce the rate of false negative diagnosis due to the lacking of proficiency. Constructing breast models by using finite element method provides a instance for imitating real breast models, instead of using gelatine or other materials to approximate the parameters of breast soft tissues, so as to estimate the phenomena while applying external forces to a physical breast model with inclusions. Finite element method can help predicting the stiffness and force-displacement ratio between normal and cancerous tissue with inclusions by satisfying the constraints of initial boundary conditions, homogeneity, linearity, and other related material properties so that the total cost and time could be reduced during development of verifying the palpation strategy.