以乳房超音波的B-mode特徵為基礎的電腦輔助診斷

Abstract

乳癌已經成為女性中最具性命威脅的疾病；做為乳房攝影的輔助，超音波已經被使用來評估乳房組織中的可疑變異，為了要建立更有效率的診斷程序，許多的電腦輔助診斷系統已經被開發出來，對於乳房超音波來說，B-mode影像是最普遍的成像技術並且廣泛的用於臨床檢查之中，因此，開發以B-mode特徵為基礎的電腦輔助診斷系統將可以提供方便且立即的協助，以放射師用來描述腫瘤的B-mode特徵來說，先前的文獻已將其量化為電腦輔助診斷系統之用，在這篇論文研究中，我們呈現了更進步的電腦輔助診斷系統，包含針對臨床病理特徵的診斷程序，用來提高偵測惡性腫瘤的敏感度，還有利用超音波斑點的特性，建立出準確性更高、更有效率的診斷，並在先進的全乳房超音波成像上成功應用，最後，將電腦輔助診斷系統應用於醫師判讀的BI-RADS 3 (近似良性)病例上，得出可將其中的潛在惡性腫瘤再篩選出的成果。

Breast cancer has become the most fatal disease among women. As an adjunct to mammography, ultrasound has been used for evaluating suspicious abnormalities in breast tissues. To establish an efficient diagnostic procedure, various computer-aided diagnosis (CAD) systems have been developed. For breast ultrasound, B-mode images are the most common imaging techniques and are widely used on clinical examination. Consequently, the development of CAD system based on B-mode features can provide a convenient and immediate assistance. B-mode features used by radiologists for tumor assessment have been quantified in previous studies. In this study, we present several improvements for CAD system. With a quantified finding based diagnosis procedure, the sensitivity was increased to reveal the malignancy of tumors. The features of speckle patterns in breast ultrasound were used to establish an efficient diagnostic procedure and were also successfully applied to automated breast ultrasound. At the end, numerous malignant tumors in BI-RADS 3 (probably benign) cases were upgraded via the proposed CAD system.