基于灰度阈值分割和动态规划的超声图像乳腺肿瘤边缘提取

目的 寻求提高超声图像乳腺肿瘤边缘提取效果的最优方法。方法 先根据超声图像的灰度分布采用灰度闽值分割定位法提取乳腺肿瘤的初始边缘，然后根据图像灰度的梯度信息采用动态规划法进行边缘的修正，从而准确提取超声图像乳腺肿瘤的边缘。结果 对18例超声图像进行乳腺肿瘤的边缘提取，结果显示本文方法相比单一的灰度阈值分割法或单一的动态规划法能更为准确地提取超声图像乳腺肿瘤的边缘。结论 本文方法可以有效地用于超声图像乳腺肿瘤的边缘提取。     

一种新的用于宫颈癌粘连细胞图像分割的分水岭算法

目的：探讨一种能有效分割宫颈癌粘连细胞图像的算法,完成各粘连细胞边缘的准确分割。方法利用水平集算法从背景区域中提取目标细胞图像,应用极小值的距离变换算法将图像归一化后与感兴趣区域的梯度图像点乘来抑制无用的梯度信息,然后运用标记分水岭算法对粘连细胞图像进行分割。结果与结论实验结果表明,该算法实现了染色不均的粘连宫颈癌细胞的有效分割,在粘连细胞的边缘建立了较传统的分水岭分割方法更准确的分割线,具有显著的临床应用价值。     

基于动态增强磁共振成像的影像组学模型对肉芽肿性乳腺炎与乳腺癌的鉴别诊断价值

目的 探讨基于动态增强磁共振成像的影像组学模型对肉芽肿性乳腺炎与乳腺癌的鉴别诊断价值。资料与方法 回顾性收集2019年2月—2022年1月于中国中医科学院西苑医院经病理证实为肉芽肿性乳腺炎和乳腺癌的患者MRI资料共82例,基于动态增强磁共振成像增强扫描第一期图像,分别用半自动分割法和手工逐层勾画分割感兴趣区,随机分配99个感兴趣区到训练组69个,测试组30个,比较两种方法所提取数据组间一致性差异。将半自动分割法所提取原始数据通过相关性分析和多因素逻辑回归法进行特征筛选。采用6种分类器（逻辑回归、支持向量机、朴素贝叶斯、决策树、随机森林、K最邻近）构建预测模型,评估各模型的诊断效能、准确度、敏感度和特异度的差异。结果 82例患者共分割出99个病灶（肉芽肿性乳腺炎37个,乳腺癌62个）。采用两种感兴趣区分割法所提取影像组学数据组间一致性欠佳[组内相关系数为0.68(0.51,0.78)]。半自动分割法所提取数据构建的6个预测模型中,逻辑回归模型和支持向量机模型诊断效能显著优于其他模型,逻辑回归模型诊断效能和稳定性最佳（训练组：曲线下面积0.928,准确度0.855,敏感度0.837,特异度...     

基于心肌信号强度阈值的半自动心脏MR影像分割法的临床可行性

正摘要目的与常规基于轮廓的图像分割法和主动脉血流测量比较,评价基于心肌信号强度阈值的半自动心脏MRI分割法的准确性和效率。方法分别用常规的和基于阈值的2种分割方法 ,在短轴电影影像上测量148例病人[(55±18)岁,男81例]的左心室容积和心肌质量。采用相位对比影像     

从CT图像中自动分割出肺部区域的算法研究

目的为用于肺部疾病的计算机辅助诊断，研究设计从CT图像中提取肺部区域的自动分割算法。方法在最优闽值分割的基础上，用自动区域生长去除气管／支气管区域，对边界跟踪法进行改进以快速去除背景干扰和获得肺部边界，最后进行肺部边界修补得到完整的肺部图像。算法采用迭代法寻找最优阈值解决了阈值选取的敏感性问题，提出了基于前层图像中气管／支气管位置的气管／支气管提取方法，避免了种子点的人工选取，基于前次搜索方向改进了八邻域搜索方法来提高边界跟踪的速度：结果用该算法对不同病人的4组胸部CT序列进行处理，能自动、快速地分割出肺部区域且精度较高：结论提出的算法能有效地从CT图像中自动提取肺部区域。     

肿瘤分割策略对机器学习预测肾透明细胞癌病理分级影响的研究

目的 探究肿瘤边缘分割策略对基于CT的机器学习预测肾透明细胞癌(ccRCC)病理分级的影响。方法 回顾性分析2009年1月至2019年12月经病理证实的ccRCC患者三期CT图像数据及临床病理学资料共546例,分为建模及内部验证集(n=311)、独立外部验证集(n=235)。按以下肿瘤图像分割策略进行分组：聚焦肿瘤边缘分割(MF组);边缘范围分别扩大1 mm(E1组)、3 mm(E3组);边缘范围分别缩小1 mm(S1组)、3 mm(S3组)。基于三期CT图像并利用CatBoost建立ccRCC病理分级(高/低级别)的机器学习预测模型。对比分析各组所提取的纹理特征及权重因子、预测模型的敏感度(SEN)、特异度(SPC)、阳性预测值(PPV)、阴性预测值(NPV)、准确率(ACC)、受试者工作特征曲线(ROC)及曲线下面积(AUC)。结果 分割边缘范围与所提取的纹理特征权重因子大小成正比,权重因子随着肿瘤图像分割边缘收缩而减小。采用如上述肿瘤图像分割策略的ccRCC病理分级预测模型的AUC分别为：MF 0.8037、E1 0.8161、E3 0.8165、S1 0.8010、S3 0.7...     

基于Snake模型的海马结构MR图像分割方法的应用研究

目的：介绍一种海马结构的磁共振图像分割方法Snake模型,并利用该模型对海马结构的磁共振图像进行分割研究。材料和方法：首先在MR图像工作站上采集图像,并进行图像预处理;然后编程,利用Snake模型进行海马结构的边缘检测,从磁共振图像中分割出海马结构。结果：除通过人机交互,由操作者给出初始轮廓外,Snake模型能从磁共振图像中自动分割出海马结构。Snake分割与人工分割总的重叠率是84％,标准差SD＝8．30。结论：Snake模型是一种快速、有效的海马结构分割方法,具有很强的鲁棒性和稳定性。     

高分辨率增强MR血管成像中的零点充填技术

目的在解剖学仿真动脉模型上对2种零点充填(zero-filling interpolation,ZIP)技术进行比较研究。方法 8个解剖学仿真动脉模型,腔内径为2～10 mm,在1.5 T MR 扫描仪上用头线圈进行钆喷替酸葡甲胺增强 MR 血管成像。快速扰相位梯度回波(FSPGR)序列的参数设置如下:反转角45°,TR6 ms,TE 1.4 ms,带宽31.2 kHz,层厚1.4 mm。研究过程中上述参数保持不变,而 ZIP 则选择1024×1024或512×512。重组最大信号强度投影(MIP)图像后,原始单层图像和 MIP 图像均用于图像质量的评估。在 ZIP 1024×1024和 ZIP 512×512图像上均测量了信噪比(SNR)。利用宽度中点(FWHM)法确定血管边缘后计算腔内直径。结果 8个解剖学仿真动脉模型在2种 ZIP 技术中均得到了很好的显示。在所有仿真模型图像中,ZIP 1024×1024技术获得的图像血管边缘均较 ZIP512×512技术的更清晰;而在 FWHM 结果中,8个仿真模型腔内直径的计算结果准确性均相同。虽然ZIP 1024×1024技术的平均 SNR(26.7±3.8)比 ZIP 512×512的平均 SNR(31.6±4.1)低(t=4.018,P<0.01),但 ZIP 1024×1024图像的总体质量均比 ZIP 512×512的更好。结论 ZIP 1024×1024技术的图像总体质量优于 ZIP 512×512技术。研究方向可着重于扫描序列的修改及参数的调整,以期同时获得较高的分辨率和 SNR。     

基于深度学习联合分割模型对扩张型心肌病心肌纤维化的定量分析

目的探讨基于深度学习的心肌-纤维化区域联合分割模型对扩张型心肌病(DCM)患者心肌纤维化定量分析的效果。方法回顾性分析徐州市中心医院2015年1月至2022年4月确诊为DCM, 并接受心脏MR-钆延迟强化检查显示左心室心肌纤维化的200例患者资料, 分为训练集120例、验证集30例、测试集50例。由影像科医师勾勒左心室心肌轮廓和选取正常心肌区域, 应用标准差法(SD)计算阈值提取纤维化心肌, 作为左心室分割和纤维化量化的参考标准。通过凸形先验的U-Net网络分割左心室心肌, 然后应用VGG图像分类网络识别正常心肌图像块, 计算SD阈值提取纤维化心肌。采用精确度、召回率、交并比和Dice系数评价心肌分割效果。采用组内相关系数(ICC)评价深度学习联合分割模型与手动提取测得的左心室心肌纤维化比率的一致性。根据纤维化比率中位数, 将测试集样本分为轻度组和重度组, 通过Mann-WhitneyU检验比较纤维化量化效果差异。结果在测试集中, 心肌分割精确度为0.827(0.799, 0.854), 召回率为0.849(0.822, 0.876), 交并比为0.788(0.760, 0.816),...     

基于改进大津法和模糊聚类的CT图像肺实质分割方法的比较研究(英文)

目的比较改进后的大津阈值法和模糊C均值聚类(FCM)两种分割算法在CT图像肺实质分割中的应用。方法选取40例肺部图像,分别采用改进的大津阈值法和模糊C均值聚类法,分割出肺实质区域,同时剔除肺部纵膈、气管。最后,采用主观评价和客观分析(图象一致性和信息熵指标)评价分割效果。结果主观分析显示,改进的大津法得到的图像中虽仍存在肺实质内部间隙,但图中孤立像素点明显减少,且颗粒的边缘也更加光滑。FCM算法分割出的肺实质空洞较少,获得的图像较完整。但有些许粘连,且提取中主气管容易错误地分割到肺实质区域,造成肺实质分割不彻底。客观分析表明,在一致性准则上,两种方法分割结果相差不大,分割出的区域都具有较高的内部区域一致性;从信息熵的角度,FCM法分割效果较好。结论肺部CT图像肺实质的分割中,对于目标与背景灰度有强对比的图像,Otsu法优于FCM法,但在阈值自适应和运算时间方面需要进一步提升;而对于存在不确定性和模糊性的图像,FCM法优于Otsu法,但在抗噪性和分割精度方面需要进一步改进。     

Segmentation of avascular necrosis of the femoral head using 3-D MR images

Avascular necrosis of the femoral head (ANFH) is a common clinical disorder in the orthopedic field. Traditional approaches to study the extent of ANFH rely primarily on manual segmentation of clinical magnetic resonance images (MRI). However, manual segmentation is insufficient for quantitative evaluation and staging of ANFH. This paper presents a new computerized approach for segmentation of necrotic lesions of the femoral head. The segmentation method consists of several steps including histogram based thresholding, 3-D morphological operations, oblique data reconstruction, and 2-D ellipse fitting. The proposed technique is rapid and efficient. In addition, it is available as a Microsoft Windows free software package on the Internet. Feasibility of the method is demonstrated on the data sets of 30 patients (1500 MR images).     

Segmentation of tumors in magnetic resonance brain images using an interactive multiscale watershed algorithm

RATIONALE AND OBJECTIVE: This article presents the evaluation of an interactive multiscale watershed segmentation algorithm for segmenting tumors in magnetic resonance brain images of patients scheduled for neuronavigational procedures. MATERIALS AND METHODS: The watershed method is compared with manual delineation with respect to accuracy, repeatability, and efficiency. RESULTS: In the 20 patients included in this study, the measured volume of the tumors ranged from 2.7 to 81.9 cm(3). A comparison of the tumor volumes measured with watershed segmentation to the volumes measured with manual delineation shows that the two methods are interchangeable according to the Bland and Altman criterion, and thus equally accurate. The repeatability of the watershed method and the manual method are compared by looking at the similarity of the segmented volumes. The similarity for intraobserver and interobserver variability for watershed segmentation is 96.4% and 95.3%, respectively, compared with 93.5% and 90.0% for manual outlining, from which it may be concluded that the watershed method is more repeatable. Moreover, the watershed algorithm is on average three times faster than manual outlining. CONCLUSION: The watershed method has an accuracy comparable to that of manual delineation and outperforms manual outlining on the criteria of repeatability and efficiency.     

MRI brain image segmentation by multi-resolution edge detection and region selection

Combining both spatial and intensity information in image, we present an MRI brain image segmentation approach based on multi-resolution edge detection, region selection, and intensity threshold methods. The detection of white matter structure in brain is emphasized in this paper. First, a multi-resolution brain image representation and segmentation procedure based on a multi-scale image filtering method is presented. Given the nature of the structural connectivity and intensity homogeneity of brain tissues, region-based methods such as region growing and subtraction to segment the brain tissue structure from the multi-resolution images are utilized. From the segmented structure, the region-of-interest (ROI) image in the structure region is derived, and then a modified segmentation of the ROI based on an automatic threshold method using our threshold selection criterion is presented. Examples on both T1 and T2 weighted MRI brain image segmentation is presented, showing finer brain tissue structures.     

Local segmentation of biomedical images

In this paper, a new algorithm for local segmentation of biomedical images is presented. First, a relatively small region is selected for segmentation on the basis of dispersion measurement of local gray values. This small region is then segmented using a segmentation algorithm based on quantization approach. While quantizing a signal, the range of input signal is divided into a number of segments. All signal values within a segment are assigned a unique reconstruction value. In segmentation of gray level images, the problem is to classify or code gray values of the pixels into two or more groups. An N-level threshold selection method for segmentation thus becomes the design of an N-level optimal quantizer. This new approach is suitable for a number of biomedical applications where the objects of interest appear as small and localized in the images. Some experimental results are also provided which illustrate the success of the new scheme.     

Automatic segmentation of white matter hyperintensities by an extended FitzHugh & nagumo reaction diffusion model

Purpose:

To evaluate the efficiency and reproducibility of the extended FitzHugh & Nagumo (FHN) reaction‐diffusion model proposed in this study for white matter hyperintensities (WMH) segmentation.

Materials and Methods:

Five types of magnetic resonance T2‐weighted fluid‐attenuated inversion‐recovery (T2FLAIR) images of 127 patients with different scanning parameters from five clinical scanner systems were selected for this study. After skull and scalp removal and denoise, the T2FLAIR images were processed by the proposed extended FHN model to obtain WMH. This new technique replaced the global threshold constant with a local threshold matrix.

Results:

There was no significant difference between the segmentation results of the training set and the manual contouring against those between the test set and the manual contouring based on similarity index (SI) values (P = 0.5217). The SI values of the five types of T2FLAIR images were 86.0% ± 15.4%, 85.8% ± 10.5%, 84.1% ± 14.8%, 87.2% ± 14.6%, 86.3% ± 12.7%, respectively, comparing the segmentation results using the proposed method to the manual delineations. The overall SI value of the images was 86.5% ± 14.5%. This approach also demonstrated a better WMH segmentation performance over its classic form (P < 0.001).

Conclusion:

The proposed approach is efficient and could provide a more effective and convenient tool for clinical quantitative WMH analysis. J. Magn. Reson. Imaging 2013;37:343–350. © 2012 Wiley Periodicals, Inc.     

Three-dimensional human airway segmentation methods for clinical virtual bronchoscopy

RATIONALE AND OBJECTIVES: The segmentation of airways from CT images is a critical first step for numerous virtual bronchoscopic (VB) applications. Automatic or semiautomatic methods are necessary, since manual segmentation is prohibitively time consuming. The methods must be robust and operate within a reasonable time frame to be useful for clinical VB use. The authors developed an integrated airway segmentation system and demonstrated its effectiveness on a series of human images. MATERIALS AND METHODS: The authors' airway segmentation system draws on two segmentation algorithms: (a) an adaptive region-growing algorithm and (b) a new hybrid algorithm that uses both region growing and mathematical morphology. Images from an ongoing VB study were segmented by means of both the adaptive region-growing and the new hybrid methods. The segmentation volume, branch number estimate, and segmentation quality were determined for each case. RESULTS: The results demonstrate the need for an integrated segmentation system, since no single method is superior for all clinically relevant cases. The region-growing algorithm is the fastest and provides acceptable segmentations for most VB applications, but the hybrid method provides superior airway edge localization, making it better suited for quantitative applications. In addition, the authors show that prefiltering the image data before airway segmentation increases the robustness of both region-growing and hybrid methods. CONCLUSION: The combination of these two algorithms with the prefiltering options allowed the successful segmentation of all test images. The times required for all segmentations were acceptable, and the results were suitable for the authors' VB application needs.     

Automatic vessel segmentation using active contours in cine phase contrast flow measurements.

The segmentation of images obtained by cine magnetic resonance (MR) phase contrast velocity mapping using manual or semi-automated methods is a time consuming and observer-dependent process that still hampers the use of flow quantification in a clinical setting. A fully automatic segmentation method based on active contour model algorithms for defining vessel boundaries has been developed. For segmentation, the phase image, in addition to the magnitude image, is used to address image distortions frequently seen in the magnitude image of disturbed flow fields. A modified definition for the active contour model is introduced to reduce the influence of missing or spurious edge information of the vessel wall. The method was evaluated on flow phantom data and on in vivo images acquired in the ascending aorta of humans. Phantom experiments resulted in an error of 0.8% in assessing the luminal area of a flow phantom equipped with an artificial heart valve. Blinded evaluation of the volume flow rates from automatic vs. manual segmentation of gradient echo (FFE) phase contrast images obtained in vivo resulted in a mean difference of -0.9 +/- 3%. The mean difference from automatic vs. manual segmentation of images acquired with a hybrid phase contrast sequence (TFEPI) within a single breath-hold was -0.9 +/- 6%.     

A method for 3D reconstruction of coronary arteries using biplane angiography and intravascular ultrasound images

The aim of this study is to describe a new method for the three-dimensional reconstruction of coronary arteries and its quantitative validation. Our approach is based on the fusion of the data provided by intravascular ultrasound images (IVUS) and biplane angiographies. A specific segmentation algorithm is used for the detection of the regions of interest in intravascular ultrasound images. A new methodology is also introduced for the accurate extraction of the catheter path. In detail, a cubic B-spline is used for approximating the catheter path in each biplane projection. Each B-spline curve is swept along the normal direction of its X-ray angiographic plane forming a surface. The intersection of the two surfaces is a 3D curve, which represents the reconstructed path. The detected regions of interest in the IVUS images are placed perpendicularly onto the path and their relative axial twist is computed using the sequential triangulation algorithm. Then, an efficient algorithm is applied to estimate the absolute orientation of the first IVUS frame. In order to obtain 3D visualization the commercial package Geomagic Studio 4.0 is used. The performance of the proposed method is assessed using a validation methodology which addresses the separate validation of each step followed for obtaining the coronary reconstruction. The performance of the segmentation algorithm was examined in 80 IVUS images. The reliability of the path extraction method was studied in vitro using a metal wire model and in vivo in a dataset of 11 patients. The performance of the sequential triangulation algorithm was tested in two gutter models and in the coronary arteries (marked with metal clips) of six cadaveric sheep hearts. Finally, the accuracy in the estimation of the first IVUS frame absolute orientation was examined in the same set of cadaveric sheep hearts. The obtained results demonstrate that the proposed reconstruction method is reliable and capable of depicting the morphology of coronary arteries.     

Development of a method for reconstructing three-dimensional data from axial, sagittal, and coronal MR images

Magnetic resonance (MR) imaging is useful for the diagnosis of brain atrophy and intracranial abnormalities. We have developed a method of automated volumetry to evaluate the degree of brain atrophy for the diagnosis of dementia. Whole-brain MR images with thin slices without gaps are required for segmentation and volumetry. However, obtaining such images requires that the patient remain at rest for a prolonged period, thereby reducing the throughput of MR imaging examinations. Therefore, a method is needed for the reconstruction of isotropic three-dimensional (3D) data using routine axial, sagittal, and coronal MR images with 30% gaps and measurement of brain volume. The method of reconstructing 3D data consists of four processes: 1) segmentation of the brain region on axial, sagittal, and coronal MR images using the region-growing technique; 2) setting data to a 3D domain; 3) registration by manual operation; and 4) interpolation between the data based on linear interpolation. In clinical MR images, the differences between this method and the conventional technique were less than 10%. These results demonstrate that this technique is able to construct 3D data from axial, sagittal, and coronal MR images.     

Image processing on Macintosh II: a practical boundary finding algorithm for biomedical measurement

With increasing research interest in displaying and analyzing biomedical images, a practical personal computer based off-line image processing software would be useful. This paper describes the implementation of an image processing work station on a Macintosh II which features a novel edge detection capability useful for biomedical measurement. The boundary finding algorithm is coded in Turbo Pascal, and operates at a speed comfortable for interactive operation. Depending on the complexity of the problem, it usually takes less than a minute for the measurement of an image. The edge detection algorithm has an built-in edge detector with decision-making capability, and can be efficiently controlled by a mouse. In this way, the local accuracy of an automatic edge detection operator, and the global accuracy of the human eye (through manual control of a mouse) are combined.