عنوان مقاله
مجازی سازی حجم کمی آماری
فهرست مطالب
مقدمه
دسته بندی آماری کلی
مجازی سازی داده های دسته بندی شده
پارامتری کردن
اجرا
نتیجه گیری و کار آتی
بخشی از مقاله
برش و کاوش
مرحله اول در مجازی سازی داده ها با استفاده از سیستم پیشنهادی ، بازبینی داده های خام به صورت برش به برش می باشد. رابط ابزار برش بعد از استفاده از رابط کاربر برای داده های پزشکی مدلسازی گردید.
دسته بندی و قطعه بندی
با وجود توسعه الگوریتم های دسته بندی تخصصی مختلف ، عمدتاً بر پروژه مشترک با هدف توسعه الگوریتم های منبع باز برای ثبت تصویر، دسته بندی و قطعه بندی موسوم به Insight Toolkit (ITK) تکیه می کنیم.
کلمات کلیدی:
Statistically Quantitative Volume Visualization Joe M. Kniss∗ University of Utah Robert Van Uitert† National Institutes of Health Abraham Stephens‡ University of Utah Guo-Shi Li§ University of Utah Tolga Tasdizen University of Utah Charles Hansen¶ University of Utah Abstract Visualization users are increasingly in need of techniques for assessing quantitative uncertainty and error in the images produced. Statistical segmentation algorithms compute these quantitative results, yet volume rendering tools typically produce only qualitative imagery via transfer functionbased classification. This paper presents a visualization technique that allows users to interactively explore the uncertainty, risk, and probabilistic decision of surface boundaries. Our approach makes it possible to directly visualize the combined ”fuzzy” classification results from multiple segmentations by combining these data into a unified probabilistic data space. We represent this unified space, the combination of scalar volumes from numerous segmentations, using a novel graph-based dimensionality reduction scheme. The scheme both dramatically reduces the dataset size and is suitable for efficient, high quality, quantitative visualization. Lastly, we show that the statistical risk arising from overlapping segmentations is a robust measure for visualizing features and assigning optical properties. Keywords: volume visualization, uncertainty, classification, risk analysis 1 Introduction Volume visualization endeavors to provide meaningful images of features ”embedded” in data. There has been a significant amount of research over the past 17 years on providing visualization of volume data [4, 6, 17, 19]. Interactive volume visualization strives to allow the user to highlight features of interest in the volume data, such as material boundaries or different tissue types. Such features are dependent on a number of factors: the kind of data, domain specific knowledge, and the user’s semantics. Simultaneously, there has been progress towards classifying features from volumetric data [5, 7, 21]. While segmentation is not considered to be a solved problem, there exist many different methods for segmenting volume data [10]. The demand for more quantitative measures in visualization has grown both within the visualization community and with the users of visualization tools. In volume rendering applications, transfer functions have typically been used for both classification and assignment of optical properties. However, using transfer functions for classification limits the user’s ability to change the type of classification that occurs and does not provide any quantifiable measure ∗e-mail: jmk@cs.utah.edu †e-mail:uitertr@cc.nih.gov ‡e-mail:abe@sci.utah.edu §e-mail:lig@sci.utah.edu ¶e-mail:hansen@cs.utah.edu A) Transfer Function-based Classification B) Unsupervised Probabilistic Classification Figure 1: A comparison of transfer function-based classification versus data-specific probabilistic classification. Both images are based on T1 MRI scans of a human head and show fuzzy classified whitematter, gray-matter, and cerebro-spinal fluid. Subfigure A shows the results of classification using a carefully designed 2D transfer function based on data value and gradient magnitude. Subfigure B shows a visualization of the data classified using a fully automatic, atlasbased method that infers class statistics using minimum entropy, non-parametric density estimation [21]. of uncertainty. Transfer functions also tend to be unintuitive to use and do not provide the user with a clear concept of how classification is being performed on the data. Statistical classification and segmentation methods incorporate a probabilistic model of the data and feature behaviors, a sophisticated notion of spatial locality, as well as the ability for the user to input their expertise. Interaction with this kind of probabilistic data and decision rules can provide each user the ability to define what information is important to his/her particular task as part of the visualization. In this paper, we propose a system that provides the user access to the quantitative information computed during fuzzy segmentation. The decision making step of classification is
