以幾何雜湊技術進行醫學影像之對位量測

近年來醫學影像之應用方面，在臨床診斷時，利用不同的造影技術所獲得之醫學影像，以顯示身體部份的組織結構或生理變化，然而，單一結構性或功能性之醫學影像，無法同時全面地提供組織的空間相關位置，及生化變化參數以供醫務人員參考，因此其臨床實用價值有限。如能發展一簡易、精確及自動化之定量方式，將兩大類的醫學影像相結合，提供精確對位之功能結構性影像，達到醫學影像資訊彙整之各項醫療診斷和應用。將兩組醫學影像對位之前，系統必須具有識別影像物件之能力，探尋出其空間位置的關連性，以決定各幾何轉換參數間(旋轉量與位移量)之最佳解。因此，本論文是利用資料結構上的雜湊索引技術，在無碰撞及溢位的情況下，只要一次就可擷取到所要之特徵值，替代傳統對全體樣版特徵點，進行逐一搜尋比對，不僅可增加識別速度，且可保留更多的特徵點資訊；提出幾何雜湊技術之對位基本原理，配合簡捷法(Downhill Simplex Method)，快速求得最佳幾何參數解。經由實驗驗證，本論文所提之幾何雜湊技術，可在不理想的影像影像品質中正確及有效的辨別物件。

關鍵字

醫學影像對位；幾何雜湊技術；簡捷法

並列摘要

Since many of imaging techniques, such as computed tomography, magnetic resonance imaging and positron emission tomography can be used to investigate the structural and functional information of human tissues. A major problem in the field of medical research is image registration and object recognition. By registering, the structural and functional information can be combined to obtain quantitative evaluation of the physiological activity of human organs. The purpose of the study is to recognize and register the images obtained from different imaging procedures by the geometric hashing based registration technique developed in this paper. The registration method has two phases: preprocessing, and recognition. In the first phase, each of the pattern models in the database is processed. For each model, the geometric information is encoded in a hash table. And then, to give an object in a scene, its features are extracted. These features are used to map the object to multiple entries in the hash table. The models in these hash table entries receive a vote. By the voting procedure, the models that receive the largest number of votes can be selected to be the object we need. In this study, the 2-D and 3-D registration techniques were developed, and applied on medical image studies. The pixel error is less than 3% in average, and the running times are of the order of a couple of minutes for both the preprocessing and the recognition procedure. The results show the ability to register the images efficiently. In brief, this developed method is not only used for object recognition, but image registration precisely and automatically.

並列關鍵字

geometric hashing technique ； medical image registration ； Simplex Method

參考文獻

1 P.J. Besl and R.C. Jain, “Three-Dimensional Object Recognition”, ACM

2 R.T. chin and C.R. Dyer, ”Model-Based Recognition in Robot Vision“, ACM

4 J. Schwartz and M. Sharir, “Identification of partially Obscured Objects in

5 A. Kalvin et al., “Two-Dimensional Model-Based Noundary Matching Using

7 Z.Zhang, “Iterative Point Matching for Registration of Free-Form Curves and

國際替代計量

以幾何雜湊技術進行醫學影像之對位量測

未授權

主題瀏覽