Structure Analysis for 3D Ultrasonic Data

瀏覽統計 Email 通知 RSS Feed

簡歷

基本資料

Project title

Code/計畫編號

NSC99-2221-E019-041

Translated Name/計畫中文名

三維超音波影像結構分析

Project Coordinator/計畫主持人

Shyh-Kuang Ueng

Funding Organization/主管機關

National Science and Technology Council

Department/Unit

Department of Computer Science and Engineering

Website

https://www.grb.gov.tw/search/planDetail?id=2116391

Year

2010

Start date/計畫起

01-08-2010

Expected Completion/計畫迄

31-07-2011

Bugetid/研究經費

492千元

ResearchField/研究領域

資訊科學--軟體

肝腫瘤為國人常罹患的致命疾病，早期偵測並給予適當治療，可以大幅提高治瘉率，降低社會成本。電腦斷層掃描、核磁共振、與超音波影像是偵測肝腫瘤的常用方法。電腦斷層掃描可提供極精細且品質良好的影像，對手術的評估和治療計畫極有幫助；但電腦斷層掃描會產生大量的幅射量，對病人日後的健康恐有影響，除非必要，否則電腦斷層掃描不應時常實施。核磁共振的副作用較少，其影像品貿也夠清晰，但它的成本高，且手續繁雜，無法在診療室實施。因此，也只能偶爾為之。相較之下，超音波影像成本低廉，操作簡便，且幾乎無副作用，在診療室即可實施；因此，它是最方便的醫學影像檢驗方法。本研究計畫將研製一套三維超音波影像的視覺化系統，以方便醫護人員進行肝癌的偵測、治療、與追蹤的工作。超音波影像含有大量的雜訊，這些雜訊降低了影像品質，阻擋了人體內部組織的顯現，它們是超音波影像處理上的最大難題。在本研究計畫中，我們提出一套以結構分析為基礎的三維超音波資料顯影技術；首先我們先將三維超音波資料重新取樣並進行初步的濾雜訊工作，因此可獲得一個多階層的資料金字塔結構。然後從最高層(解析度最低) 的資料中算出其海森矩陣與此矩陣的特徵值和特徵向量；並使用這些資訊判定各部位的組織形狀和大小。然後利用較高解析度的資料逐漸修正判讀結果，以求得整個掃描範圍的幾何屬性。幾何屬性計算完成後，我們會利用特製化的濾波器對資料做進一步的除雜訊工作。其目的在於排除不清晰且無用處的資訊，並加強邊界和管狀組織的效果。接著以資料的紋理、幾何屬性、梯度、和灰階值，將資料分類並給予不同的顏色、透明度、和圖案，以建構一個新的三維資料集合。最後利用硬體加速的實體成像繪圖方法顯示人體中的各組織和異常的區域，以方便醫護人員進行診治、追蹤和評估的工作。 Liver cancers are fatal to our health. Detecting liver cancers is crucial for curing the diseases. Blood-testing, CT scan images, MRI images, and Ultrasound images are useful methodologies for detecting liver cancers. However, blood-testing can only be used to survey the functionalities of our livers. It can not be utilized to locate cancer cells or estimate the size of cancer tissues. CT and MRI images can offer high-resolution portrays of our livers. They are very useful for surgical assessment and planning. Nonetheless, the high-dose X-ray radiation used in CT scan may cause cancers in the future. This method should not be employed unless it is un-avoidable. MRI-imaging is very expensive. Taking MRI images requires a lot of efforts too. Doctors can not perform MRI scanning too often. They can not take MRI images in their clinical examination procedures unless they have a team of engineers to help them. Compared with other medical imaging methods, ultrasound-imaging is easy to conduct and less expensive. Unlike CT scan, it is un-intrusive and causes very little side-effects. Doctors can take ultrasonic images if the medical procedures require the revealing of some special organs and tissues. Thus ultrasound imaging is widely used for detecting liver cancers and tracing the results of treatments or surgeries. One major problem reduce the usefulness of ultrasound imaging, the speckle noises in ultrasound images. Speckles are created by the constructive and destructive interferences of sound beams. They come with ultrasound images. They block insight data such that it is difficult to obtain clear displays of our internal organs and tissues. In this research, we propose a structure-based approach to post-process ultrasound images. At first, we create a set of hierarchical data set. Then the diffuse tensor field of the data is computed at each level of data. The eigen-vectors and eigen-values of the tensor field are used to detect the underlying structure in each region. Then based on the shapes and sizes of the structures, adaptive filters are used to enhance features and suppress noises. The structure information obtained in all levels are combined to form the full description and profile of the entire 3D ultrasound data volume. We do not intend to reconstruct the geometries of livers. Instead, advanced volume rendering methods will be employed to display the fine details of livers. Beside cancer cells, we will pay more attention to the linear structures, for example, blood veins since they are vital for surgical planning. The major goals of the research include: 1. Design and test a 3D structure analysis procedure for 3D ultrasonic data sets. 2. Improve current filtering methods such that more linear structures and cancer cells are preserved while speckles are reduced. 3. Create a post-processing system to help doctors

Keyword(s)

超音波影像處理
結構分析
特徵強化
雜訊抑制
Ultrasound image processing
structure analysis
feature enhancement
speckle reduction

DSpace CRIS

Structure Analysis for 3D Ultrasonic Data

基本資料

Description