腎臟的體積在臨床診斷上是個很重要的參數,無論是對成人、新生兒、或是三個月後的胎兒都是如此,以成人為例,醫師會藉由腎臟的長度跟體積來評估或是追蹤病人是否患有尿道感染、腎血管狹窄等疾病。至於新生兒或是三個月後的胎兒,醫師會藉由異常大的腎臟體積來判斷是否患有新生兒腎盂積水等疾病,所以腎臟邊界的偵測對於醫師判讀病理來說,是很重要的一環。而至今的文獻,對於腎臟超音波影像邊界偵測的問題,仍然沒有提出一套完整而且有效的解決方法出來。本研究則是以星狀演算法為主軸,並搭配馬可夫隨機場的理論基礎,來偵測腎臟超音波影像的邊界,目的是希望能結合星狀演算法與馬可夫隨機場的優點,發展出一套近乎全自動化、且運作快速的演算法,以達到電腦輔助診斷的目的。
Renal volume is an important parameter in clinical settings both for the adult, newborns and fetuses. About the former, evaluation and follow-up of patients with urinary tract infections, renal vessels stenosis and others are done in terms of both the length and the volume within the organ. About newborns and fetuses, the neonatal hydroneohrosis is detected by means of abnormal large volumes enclosed by the organ. Therefore, the boundary detection problem of kidney is important for the doctors to diagnose pathology. But Solutions proposed so far in the literature are very application-driven so they do not constitute a complete and valid method when applied to the boundary detection problem of ultrasonic kidney images. This research is based on star algorithm, and matches the rationale of markov random fields to detect the boundary of ultrasonic kidney images. We hope to combine the advantages of star algorithm and markov random fields, and develop an algorithm with near automation and high speed for the purpose of the computer aided diagnosis.