本論文旨於結合超音波影像與同時擷取所得之觸覺特徵應用於乳房腫瘤之診斷,將可疑之非正常組織特性透過超音波探頭擷取所得之影像及力量感測器所回傳之非正常組織上方之觸覺特徵,透過電腦輔助診斷(Computer Aided Diagnosis, CAD)系統進行組織特性之鑑別,用以區分非正常組織所代表的腫瘤型態為良性或為惡性。本論文所研究之電腦輔助診斷系統利用我們所研發之混合式觸覺結合超音波影像系統進行訊號擷取,透過機器手臂所量測之探頭位置,結合力量感測器獲得的資料與超音波探頭所擷取的影像,透過特徵萃取,我們可得到觸覺的特徵及超音波影像之腫瘤大小及深度等特徵,再將以上特徵送入類神經網路進行腫瘤特性分類。本文對本系統所研製的混合式觸覺結合超音波影像系統進行可行性之先期研究,並使用實驗室及臨床測試進行驗證。
This paper aims to diagnose the characteristics of tumors in breast by using the extracted features from ultrasound image and tactile information. The ultrasound probe in conjunction with force sensor is palpated on the suspicious abnormal area above for data acquisition, the features of the acquired data is extracted and send into our computer aided diagnosis system for tumor classification. The tumor in breast can be distinguished out as benign or malignant by our computer aided diagnosis system. This study develops a robot arm to mount the ultrasound and force sensor to acquire the tactile information and ultrasound image simultaneously. The extracted tactile feature, tumor stiffness ratio, and extracted ultrasound image feature including texture, morphology, mobility can be obtained. These features can help the tumor classification by using support vector machine. In this study, we validate the efficacy of this hybrid system containing features that combine both tactile information and ultrasound image. Laboratory and clinical trials are both proceeded and its efficacy are proved.