Abstract
摘要:神經造影(Neuroimaging)為一強而有力的工具去探索正常和有病變大腦的組織和功能。由於磁振造影(Magnetic Resonance Imaging, MRI)具有高對比度、高解析度和多重頻譜特性,它已成為神經造影中最常使用的設備之一。在磁振影像分析上,雜訊(Noise)一直是影像品質惡化的主因之一。影像變質不僅影響視覺上的臨床檢查,並會更進一步影響組織分類(Tissue classification)、影像分割(Segmentation)和結構對位(Registration)等電腦化處理結果。因此,對隨後的眾多影像處理分析及應用,去除在腦部磁振影像上的雜訊是相當重要且必要的。
在文獻上,已有許多去雜訊(Denoising)演算法被不同的研究學者提出。然而,大部份的方法需要繁瑣費力的參數調整,且常對腦部影像的特徵(Feature)和紋理(Texture)相當敏感。因此,藉由人工智慧(Artificial intelligence)技術來自動化參數調整將是莫大的助益。不過,這又引發另一個來尋求最適當且有意義特徵值的問題。這計畫的第一年將嚐試在為數眾多的不同影像特徵種類中,有系統地探討對此自動化程序裡最顯著的紋理特徵值。在第二年裡,我們將研發一個以結合倒傳遞類神經網路(Back propagation neural network)與紋理特徵分析(Texture feature analysis)為基礎的自動化雙邊濾波(Bilateral filter)系統。在此,因考量雙邊濾波器之有效性和簡潔性,故採用為此參數自動化系統的第一個實例。
本研究計畫的終極目的乃是研發一個自動且強健的雜訊去除(Noise removal)系統,希冀促進臨床上和學術上神經影像處理之應用。基於此,一個以結合紋理特徵和類神經網路為基礎的雙邊濾波器被獨創地提出。大量不同類型的影像資料將被用來評估這一新系統在去除腦部磁振影像雜訊的成效。而此一系統的開發成功,將能大幅減輕人力、物力和財力在神經科學研究和磁振影像前處理的投資。
Abstract: Neuroimaging is one powerful tool used to investigate the structure and function of the brain in both health and disease. Magnetic resonance imaging (MRI) has been one of the most frequently used neuroimaging modalities due to its high contrast among different soft tissues, high spatial resolution across the entire field of view, and multi-spectral characteristics. In MR image analysis, noise is one of the main sources of quality deterioration not only for visual inspection but also in computerized processing such as tissue classification, segmentation and registration. Consequently, noise removal in brain MR images is important and essential for a wide variety of subsequent processing applications.
In the literature, abundant denoising algorithms have been proposed, most of which require laborious tuning of parameters that are often sensitive to specific image features and textures. Automation of these parameters through artificial intelligence techniques will be highly beneficial. However, this will induce another problem of seeking appropriate meaningful attributes among a huge number of image characteristics for the automation process. The first year of this study is in an attempt to systematically investigate significant attributes from image characteristics to facilitate subsequent automation process. More specifically, we will investigate a broad range of potential texture features, from which several best texture features will be incorporated into an artificial neural network. In the second year, a back propagation network framework associated with texture feature analysis will be developed to optimize the parameters of the bilateral filter process. Herein, the bilateral filter is adopted as the first example of the automation system for its effectiveness and compactness.
The ultimate goal of this study is to develop an automatic and robust noise removal system for neuroimaging processing applications both clinically and academically. Toward that end, an artificial neural network based bilateral filter associated with texture feature analysis is originally proposed for obtaining reliable denoising in brain MR images. A wide variety of image data from worldwide public domains will be used to evaluate the proposed system. The success of
this study will dramatically reduce the labor and effort of noise removal to facilitate subsequent
studies in neuroscience and MR image processing applications.
Keyword(s)
雜訊去除
雙邊濾波器
類神經網路
紋理特徵
磁振造影。
Noise removal
bilateral filter
neural network
texture feature
MRI.