胸部CT中肺实质的提取与辅助诊断

根据临床应用需求,研究胸部高分辨率CT图像中肺实质的提取与量化诊断问题。首先讨论肺区分割与肺实质提取问题,自动分割法采用全局自适应阈值将躯干和背景分离,然后应用轮廓跟踪方法获取到肺部轮廓;人工分割则是在勾勒若干轮廓点后,应用Bresenham扫描线法得到连续的肺部轮廓。利用基于四邻域的背景标记扫描线方法,提取肺部实质区域。最后,根据肺气肿量化诊断标准,进行量化分析与诊断。实验证明,该方法能快速、准确地分割肺实质,实现肺气肿的量化分析与辅助诊断。     

基于小波变换和感兴趣区域编码的ECG压缩方法

提出了一种基于小波变换和感兴趣区域编码的ECG压缩方法：首先使用正交小波变换对去均值处理后的信号进行多层分解。然后根据对原始信号特征提取的结果，找到感兴趣区域，进而找到与感兴趣区域对应的系数，视这些系数为重要系数而予以保留。对非感兴趣区域系数从小到大排序，根据目标PRDBE（Percentage Rootmean-square Difference with Baseline Eliminated）指标，计算该区域系数阈值并阈值化。通过扫描所有小波系数得到重要系数图。最后对重要系数进行标量量化。对重要系数图进行RLE（Run Length Encoding）编码，并使用Huffman编码进一步提高压缩比。使用MIT／BIH心律失常数据库测试表明。本方法在最大程度保存诊断信息，获得好的信号质量的同时，也获得了基本满足实际应用需要的压缩比。     

基于分水岭算法的交互式三维分割方法

背景：在临床中准确对人体组织进行三维分割能提高临床诊断的准确性,但传统的分水岭算法存在过度分割问题,难以实现人体组织的三维分割。 目的：为准确三维分割人体组织,减少图像中伪极小值点对图像分割的影响,提出了一种基于控制标记符分水岭的交互式三维分割方法。 方法：提取CT序列图像的内部和外部标记符,以此修正梯度图像并进行分割;在此基础上,根据序列图像上下层的相似性,利用人机交互进行组织结构的三维分割。首先在第一张序列图像上手工选取感兴趣区域上的一个点,借助同一组织在连续CT序列图像上面积的重叠关系即可从三维序列图上提取出感兴趣区域。 结果与结论：基于控制标记符的分水岭算法解决了直接应用梯度图像进行分割的过度分割问题,便于进一步分割图像。利用基于分水岭算法的交互式三维分割方法得到的三维分割结果经过三维可视化后可清晰、准确地反映组织的三维特征。     

独立分量分析在乳腺钼靶X片感兴趣区域提取中的应用

为乳腺癌早期诊断和乳腺X线影像微钙化点计算机辅助检测的前期预处理,本研究提出基于独立分量分析（ICA）的自动提取新算法并且将其应用于乳腺图像感兴趣区域的自动提取.其具体思路是：（1）将乳腺区域图像提取成等大的子图像作为待测乳腺图像感兴趣区域;（2）将ICA应用于乳腺图像感兴趣区域得到基图像;（3）将待识别乳腺图像感兴趣区域在基图像所构成的子空间进行投影求得待测乳腺图像感兴趣区域的特征矢量;（4）用人工神经网络分类方法进行乳腺图像感兴趣区域的模式判别.对临床实际病例的试验结果表明,该方法的检出率为91%,与同类研究检出率相当.本研究方法简单有效,并具有较高的智能性,为ROI的自动提取提供了新的研究思路.     

基于纹理分析的带标记线心脏MR图像分割

背景：左心室边界的准确分割是对左心室运动及形变进行分析的前提。由于受带标记线心脏核磁共振图像中标记线强梯度的影响,对左心室内膜的提取变得非常困难。 目的：为了抑制标记线对图像分割的影响,提出了一种基于最小值-方差能量图的纹理分析方法。 方法：首先对局部最小值和方差进行加权求和,得到能量图;然后利用中值滤波滤除能量图中的伪影并保持边界;最后,应用GVF-snake模型提取左心室内膜。 结果与结论：针对标记线在心脏MR图像中的分布特征,提出了一种基于最小值-方差的纹理分析方法,该方法有效地去除了标记线。结果提示,对使用该纹理分析方法生成的能量图应用GVF-snake模型可以较好地提取左心室内膜。     

基于病理“金标准”标记脊髓损伤慢性期MRI图像坏死区域的方法

脊髓损伤慢性期胶质瘢痕及其所包绕的囊腔所构成的脊髓坏死区域在磁共振成像（MRI）中形态各异，边界模糊。针对单纯根据图像无法准确识别脊髓坏死区域这一问题，提出一种基于病理切片“金标准”来标记MRI图像实际坏死区域的方法。双边滤波去噪并刚体配准脊髓损伤慢性期的多模态MRI图像，使用Snake模型进行脊髓的分离，并利用双三次插值法对脊髓进行拉直。采用双树复小波变换融合不同模态的图像后，基于病理切片的胶质瘢痕及囊腔区域与MRI图像的脊髓坏死区域具有空间相对位置一致性的特点，建立从脊髓病理切片到MRI图像的空间映射关系，最终实现MRI图像中实际坏死区域的标记。以圆度和偏心率作为形状评价指标，对本方法标记的MRI坏死区域与病理切片“金标准”结果进行对比，相似度分别达到了0.93±0.03和0.97±0.02。结果表明，应用本方法构建的映射关系能较为准确、客观地标记MRI图像的实际坏死区域，为后期计算机自动识别与分割MRI坏死区域奠定基础。     

基于纹理分析和双边滤波方法对加标记心脏核磁共振图像分割

背景:加标记心脏核磁共振成像方式提供了左心室内外心膜的边缘信息,该边缘信息可由分割图像得到。但是,所引入的标记线加大了这类图像边界分割的困难。目的:针对目前在加标记心脏核磁共振图像中对左心室分割困难的问题,提出了一种新的自动分割的方法。方法:首先,使用全局直方图规定化方法增强标记和非标记区域的对比度;然后,利用一种简单的纹理分析方法区分血流充盈的心腔(非纹理)区域和加标记心肌(纹理)区域;再应用双边滤波在保持边界的同时滤掉图像的伪影;最后,用GVF-snake模型自动提取左心室图像的边界。结果与结论:提出了一种简单的纹理分析方法来移除标记线:用局部窗口中的最大灰度值与最小灰度值之差来代替原象素点灰度值,再运用双边滤波滤除图像伪影并保持边界,最后应用GVF-snake模型实现了左心室边界的有效提取。实验结果显示,该方法能够较好地提取部分加标记心脏核磁共振图像中血流充盈区的边界。     

交互式GACV模型的肿瘤图像分割

目的:提取医学图像中肿瘤区域,用以测量肿瘤体积问题。方法:提出一种基于GACV(Geodesic-Aided C-Vmethod)的交互式模型。该模型首先人工选取感兴趣区域,并在区域内设定初始水平集与肿瘤内部种子点,然后在感兴趣区域上应用将图像梯度边缘信息与图像区域灰度特性统一到同一分割中的GACV模型,得到肿瘤的粗分割结果。最后为去除目标内外孔洞,提出一种无损边缘的膨胀搜索算法,作为细分割。结果:将该模型应用于不同形状的肿瘤图像中,能成功检测肿瘤轮廓。通过实验与其它活动轮廓分割方法结果对比,结果显示该模型在准确分割肿瘤边界与分割算法耗时方面均具有良好表现。结论:本文提出的分割方法能高效率、准确识别肿瘤区域。     

CT图像肺结节自动检测

肺癌是对人类生命健康危害最大的恶性肿瘤之一。计算机辅助诊断系统对肺部CT图像进行自动分析后,可提示医生可疑肺结节,从而克服医生在诊断中的一些主观因素,为此本文提出了一种基于胸部CT图像的可疑肺结节自动检测算法。首先,根据胸部组织的特殊结构,利用一种新的分割算法提取出肺实质部分;在此基础上提取出灰度与结节相近的感兴趣区域,包括结节、肺血管、支气管;然后,以已标记的结节数据作为样本集,计算结节的面积、灰度均值、灰度方差、圆形度、形状矩、体积、球形度等特征值,利用最近邻法建立分类器判别函数;最后,计算测试集感兴趣区域的上述特征,对其进行判别、分类,并标记出结节。试验结果表明,该算法综合考虑了肺结节特征,具有较高的准确度。     

基于奇异值分解和频率选择的磁共振波谱频域量化方法

时域的磁共振波谱(magnetic resonance spectroscopy,MRS)信号经过傅立叶变换得到频域波谱,通常我们仅对某段频率范围内的波谱感兴趣。本文采用基于奇异值分解(singular value decomposition,SVD)和频率选择的方法,仅对感兴趣的频率范围内的波谱进行量化,降低了其它频率的波谱对感兴趣区域的影响,并大大减少了运算量。通过与时域标准的SVD方法比较实验,表明该算法准确、省时。     

Automatic segmentation algorithm for the extraction of lumen region and boundary from endoscopic images

A new segmentation algorithm for lumen region detection and boundary extraction from gastro-intestinal (GI) images is presented. The proposed algorithm consists of two steps. First, a preliminary region of interest (ROI) representing the GI lumen is segmented by an adaptive progressive thresholding (APT) technique. Then, an adaptive filter, the Iris filter, is applied to the ROI to determine the actual region. It has been observed that the combined APT-Iris filter technique can enhance and detect the unclear boundaries in the lumen region of GI images and thus produces a more accurate lumen region, compared with the existing techniques. Experiments are carried out to determine the maximum error on the extracted boundary with respect to an expert-annotated boundary technique. Investigations show that, based on the experimental results obtained from 50 endoscopic images, the maximum error is reduced by up to 72 pixels for a 256 × 256 image representation compared with other existing techniques. In addition, a new boundary extraction algorithm, based on a heuristic search on the neighbourhood pixels, is employed to obtain a connected single pixel width outer boundary using two preferential sequence windows. Experimental results are also presented to justify the effectiveness of the proposed algorithm.     

Fully automated gradient based breast boundary detection for digitized X-ray mammograms

Accurate segmentation of the breast from digital mammograms is an important pre-processing step for computerized breast cancer detection. In this study, we propose a fully automated segmentation method. Noise on the acquired mammogram is reduced by median filtering; multidirectional scanning is then applied to the resultant image using a moving window 15×1 in size. The border pixels are detected using the intensity value and maximum gradient value of the window. The breast boundary is identified from the detected pixels filtered using an averaging filter. The segmentation accuracy on a dataset of 84 mammograms from the MIAS database is 99%.     

Broadband ultrasound attenuation imaging: algorithm development and clinical assessment of a region growing technique

This paper presents a computerized method for the selection of an irregular region of interest (ROI) in broadband ultrasound attenuation (BUA) images. A region growing algorithm searches an initial region in the posterior part of the calcaneus until the pixel with the lowest attenuation value is found; this is the starting seed. Then, the algorithm evaluates the values of the eight pixels neighbouring the starting seed. Pixels that have the closest value to the starting seed are accepted. This procedure is the first processing level. The procedure is repeated for the group of pixels neighbouring those accepted from the previous processing level. The algorithm ceases when the number of accepted pixels reaches a user-specified number. The clinical part of this study compares measurements of BUA at an automatic ROI implemented on a quantitative ultrasound imaging device, defined as the circular region of lowest attenuation in the posterior part of the calcaneus, and at irregular ROIs of various sizes generated by the algorithm developed in this study. The algorithm was applied to BUA images obtained from 24 post-menopausal women with hip fractures and 26 age-matched healthy female subjects. The use of the irregular ROI with a size of 2400 pixels is proposed because that region yielded better clinical results compared to irregular ROIs with different size and the circular automatic ROI.     

基于深度学习的颈动脉超声图像内中膜厚度测量

目的 颈动脉血管内中膜厚度(IMT)是衡量动脉粥样硬化程度的重要标准.一般采用人工标定进行测量,该过程耗时且繁琐,由此提出一种总体性能较好的全自动分割(AS)算法.方法 该算法首先利用卷积神经网络(CNN)识别出颈动脉血管远端,进而提取包含颈动脉内膜、中膜部分的感兴趣区域(ROI).采用基于堆栈式自编码器(SAE)构造的模式分类器将ROI中的像素进行分类.最后利用分类区域的面积信息和位置信息对分类结果进行甄别,运用曲线拟合提取边界完成测量任务.结果 针对本研究所用图像库中的84幅颈动脉超声图像进行实验,金标准(GT)由两名专家4次测量的平均值产生,其与AS之间的绝对误差和标准差为(13.3±20.5) μm,协方差系数为0.990 7.结论 实验结果表明,此算法总体性能较好,能够实现超声颈动脉血管内中膜全自动、快速、准确分割,从而满足临床需要.     

Computer aided detection of masses in mammography using subregion Hotelling observers

We propose to investigate the use of the subregion Hotelling observer for the basis of a computer aided detection scheme for masses in mammography. A database of 1320 regions of interest (ROIs) was selected from the DDSM database collected by the University of South Florida using the Lumisys scanner cases. The breakdown of the cases was as follows: 656 normal ROIs, 307 benign ROIs, and 357 cancer ROIs. Each ROI was extracted at a size of 1024 x 1024 pixels and sub-sampled to 128 x 128 pixels. For the detection task, cancer and benign cases were considered positive and normal was considered negative. All positive cases had the lesion centered in the ROI. We chose to investigate the subregion Hotelling observer as a classifier to detect masses. The Hotelling observer incorporates information about the signal, the background, and the noise correlation for prediction of positive and negative and is the optimal detector when these are known. For our study, 225 subregion Hotelling observers were set up in a 15 x 15 grid across the center of the ROIs. Each separate observer was designed to "observe," or discriminate, an 8 x 8 pixel area of the image. A leave one out training and testing methodology was used to generate 225 "features," where each feature is the output of the individual observers. The 225 features derived from separate Hotelling observers were then narrowed down by using forward searching linear discriminants (LDs). The reduced set of features was then analyzed using an additional LD with receiver operating characteristic (ROC) analysis. The 225 Hotelling observer features were searched by the forward searching LD, which selected a subset of 37 features. This subset of 37 features was then analyzed using an additional LD, which gave a ROC area under the curve of 0.9412 +/- 0.006 and a partial area of 0.6728. Additionally, at 98% sensitivity the overall classifier had a specificity of 55.9% and a positive predictive value of 69.3%. Preliminary results suggest that using subregion Hotelling observers in combination with LDs can provide a strong backbone for a CAD scheme to help radiologists with detection. Such a system could be used in conjunction with CAD systems for false positive reduction.     

A novel background field removal method for MRI using projection onto dipole fields (PDF)

For optimal image quality in susceptibility-weighted imaging and accurate quantification of susceptibility, it is necessary to isolate the local field generated by local magnetic sources (such as iron) from the background field that arises from imperfect shimming and variations in magnetic susceptibility of surrounding tissues (including air). Previous background removal techniques have limited effectiveness depending on the accuracy of model assumptions or information input. In this article, we report an observation that the magnetic field for a dipole outside a given region of interest (ROI) is approximately orthogonal to the magnetic field of a dipole inside the ROI. Accordingly, we propose a nonparametric background field removal technique based on projection onto dipole fields (PDF). In this PDF technique, the background field inside an ROI is decomposed into a field originating from dipoles outside the ROI using the projection theorem in Hilbert space. This novel PDF background removal technique was validated on a numerical simulation and a phantom experiment and was applied in human brain imaging, demonstrating substantial improvement in background field removal compared with the commonly used high-pass filtering method.     

A novel PET tumor delineation method based on adaptive region-growing and dual-front active contours

To more accurately and precisely delineate a tumor in a 3D PET image, we proposed a novel, semi-automatic, two-stage method by utilizing an adaptive region-growing algorithm and a dual-front active contour model. First, a rough region of interest (ROI) is manually drawn by a radiation oncologist that encloses a tumor. The voxel having the highest intensity in the ROI is chosen as a seed point. An adaptive region growing algorithm successively appends to the seed point all neighboring voxels whose intensities > = T of the mean of the current region. When T varies from 100% to 0%, a sharp volume increase, indicating the transition from the tumor to the background, always occurs at a certain T value. A preliminary tumor boundary is determined just before the sharp volume increase, which is found to be slightly outside of the known tumor in all tested phantoms. A novel dual-front active contour model utilizing region-based information is then applied to refine the preliminary boundary automatically. We tested the two-stage method on six spheres (0.5-20 ml) in a cylindrical container under different source to background ratios. Comparisons between the two-stage method and an iterative threshold method demonstrate its higher detection accuracy for small tumors (less than 6 ml). One patient study was tested and evaluated by two experienced radiation oncologists. The study illustrated that this two-stage method has several advantages. First, it does not require any threshold-volume curves, which are different and must be calibrated for each scanner and image reconstruction method. Second, it does not use any iso-threshold lines as contours. Third, the final result is reproducible and is independent of the manual rough ROIs. Fourth, this method is an adaptive algorithm that can process different images automatically.     

A simple derivation and analysis of a helical cone beam tomographic algorithm for long object imaging via a novel definition of region of interest

We derive and analyse a simple algorithm first proposed by Kudo et al (2001 Proc. 2001 Meeting on Fully 3D Image Reconstruction in Radiology and Nuclear Medicine (Pacific Grove, CA) pp 7-10) for long object imaging from truncated helical cone beam data via a novel definition of region of interest (ROI). Our approach is based on the theory of short object imaging by Kudo et al (1998 Phys. Med. Biol. 43 2885-909). One of the key findings in their work is that filtering of the truncated projection can be divided into two parts: one, finite in the axial direction, results from ramp filtering the data within the Tam window. The other, infinite in the z direction, results from unbounded filtering of ray sums over PI lines only. We show that for an ROI defined by PI lines emanating from the initial and final source positions on a helical segment, the boundary data which would otherwise contaminate the reconstruction of the ROI can be completely excluded. This novel definition of the ROI leads to a simple algorithm for long object imaging. The overscan of the algorithm is analytically calculated and it is the same as that of the zero boundary method. The reconstructed ROI can be divided into two regions: one is minimally contaminated by the portion outside the ROI, while the other is reconstructed free of contamination. We validate the algorithm with a 3D Shepp-Logan phantom and a disc phantom.     

Prostate cancer multi-feature analysis using trans-rectal ultrasound images

This note focuses on extracting and analysing prostate texture features from trans-rectal ultrasound (TRUS) images for tissue characterization. One of the principal contributions of this investigation is the use of the information of the images' frequency domain features and spatial domain features to attain a more accurate diagnosis. Each image is divided into regions of interest (ROIs) by the Gabor multi-resolution analysis, a crucial stage, in which segmentation is achieved according to the frequency response of the image pixels. The pixels with a similar response to the same filter are grouped to form one ROI. Next, from each ROI two different statistical feature sets are constructed; the first set includes four grey level dependence matrix (GLDM) features and the second set consists of five grey level difference vector (GLDV) features. These constructed feature sets are then ranked by the mutual information feature selection (MIFS) algorithm. Here, the features that provide the maximum mutual information of each feature and class (cancerous and non-cancerous) and the minimum mutual information of the selected features are chosen, yielding a reduced feature subset. The two constructed feature sets, GLDM and GLDV, as well as the reduced feature subset, are examined in terms of three different classifiers: the condensed k-nearest neighbour (CNN), the decision tree (DT) and the support vector machine (SVM). The accuracy classification results range from 87.5% to 93.75%, where the performance of the SVM and that of the DT are significantly better than the performance of the CNN.     

Computerized detection of clustered microcalcifications: a modular approach with non-linear filters

The ideal computerized mammogram processing system still needs to be developed. In order to achieve maximum flexibility we suggest a modular scheme, dividing the processing sequence into functionally autonomous modules. This paper provides a general scheme for detection and/or automated recognition of microcalcifications. Some modules that perform ROI selection are introduced, using special non-linear filters designed for microcalcification detection. A first type of filter selects pixels with specific statistical local features, as compared to the local mean. Among these, only pixels satisfying particular constraints on the local standard deviation are kept. Another type of filter then checks the local mean values of gradient components, so that sharp variations, unrelated to small close objects, can be eliminated. The scheme thus applies different non-linear filters in combination, making precise identification of clustered microcalcifications possible. This modular approach seems greatly to simplify system maintenance and consistency, as well as affording a comparison of different processing techniques and parameters.