3M和5M显示器对不同分辨率乳腺影像图中微钙化识读的影响EI北大核心CSCD

通过使用高、低两种分辨率的乳腺影像图来比较3M与5M医用专业显示器对乳腺微钙化识读的影响。选择高、低两种分辨率的乳腺影像图各100例(各含微钙化病例40例、正常对照病例60例)。由1名高年资和1名低年资放射科医师评估两种显示器对乳腺微钙化的显示率,识别效能用ROC曲线判断,并使用Kappa分析检验两名医生的判读一致性。在低分辨率影像图组中,两名医生在3M与5M医用专业显示器上对微钙化的识别效能相同(P=0.451及0.559);在高分辨率影像图组中,高年资医生使用5M医用专业显示器对乳腺微钙化的识别率明显高于3M(P=0.022),低年资医生的识别率无显著差异(P=0.141)。两名医生在5M显示器上判读的一致性都好于3M显示器,在5M显示器上判读高分辨率影像图时,两名医生的判读有极好的一致性(K=0.862)。因此,对不同分辨率的乳腺影像图,应配套相应分辨率的显示器,高年资医师识读高分辨率影像图组时应配套5M显示器更利于微钙化的检出。     

乳腺X线图像肿块分割

乳腺肿块分割是乳腺癌计算机辅助诊断(CAD)检测和识别系统中关键的一步.由于乳腺肿块与背景相互交叠、边界不清晰、乳房密度不均匀,使得其分割比较困难.本文基于区域增长算法,研究了利用乳腺肿块自身特征得到最优分割阈值的方法,从而提出一种对乳腺X线图像肿块快速、有效的分割方法.实验结果表明该方法在保证肿块针状化特征情况下,拥有较好的分割效果.     

一种基于小波变换的乳腺X线图肿块分割方法

目的：乳腺癌的早期诊断和治疗是能够降低乳腺癌患者死亡率的有效途径。通过乳腺X线图像观察乳腺状况是目前乳腺癌普查的首选影像方法。随着图像处理技术的高速发展,计算机辅助检测技术在乳腺癌的检测方面起到越来越重要的作用。方法：本文首先利用图像处理领域的形态学处理、区域增长等相关知识,对乳腺X线图像进行预处理操作,去除图像中所包含的干扰信息。之后提出一种对图像的灰度直方图进行小波变换,并根据其小波变换的模极大值点确定图像分割阈值的方法对乳腺X线图像中的疑似肿块区域进行粗分割。在通过粗分割过程获得乳腺肿块的大致位置信息之后,再利用区域增长的方法获得肿块的边缘信息。结果：本文选取MIAS乳腺图像数据库中的65幅图像作为测试图像,保证每幅图像至少包含一个乳腺肿块。利用本文所提方法对这65幅图像进行实验,并将实验结果与该数据库中的专家标注信息作对比,实验结果为当采用db40的小波系数时的检出率为95．5％。结论：本文所述方法能够有效地分割出乳腺X线图中的肿块区域,并且有较高的检出率,具有进一步研究和应用的价值。     

基于超声体模的颈动脉血流介导血管扩张斑块高度测量方法比较

通过设计实验对超声体模的血管扩张功能(FMD)斑块高度进行了测量,对各种测量方法做出了定性与定量分析和比较,从而确定最佳的评估方法及软件,以用于颈动脉粥样硬化临床早期诊断和治疗。以颈动脉超声体模的实际尺寸作为金标准,对FMD斑块高度进行了测量,利用超声诊断仪专用软件和自行开发的分割软件,分别使用3种不同的测量方法：横向直接测量、横向间接测量和纵向直接测量,获取斑块的高度,并进行了比较和分析。结果表明,在横向间接测量方法中,超声诊断仪专用软件比自行开发的分割软件更为准确;而在横向直接测量和纵向直接测量中,自行开发的分割软件比超声诊断仪专用软件更为准确。自行开发的分割软件与超声诊断仪专用软件的测量结果在统计学上存在显著性差异(P<0.05),但两者相差的绝对值并不大（对应0.30、0.45、0.60 cm等 3种斑块高度,小于0.02 cm）。总体来说,超声诊断仪专用软件和自行开发的分割软件测量结果都接近于金标准,但自行开发的分割软件的精确性优于超声诊断仪专用软件     

乳腺MRI分割与基于光线投影的三维重建

目的乳腺MRI(magnetic resonance imaging)序列成像连续、位置固定,但二维图像无法提供便于观察的肿块的深度与外形信息,为此提出乳腺及其肿块分割与三维重建方法。方法首先利用最大类间方差法与区域生长法提取乳腺外轮廓,然后选用基于梯度和能量信息的水平集方法准确分割乳腺肿块,并对分割结果采用基于包围盒的光线投影算法进行体绘制。结果对患者的多组MRI序列进行分割实验并对几种分割算法进行客观评价,三维重建实验得到乳腺与肿块的立体图像以供医生观察。结论分区处理的水平集分割与基于光线投影的三维重建为医生提供更加形象的立体视觉效果,不仅有助于对肿块病变程度的诊断,而且能帮助外科医生在保乳手术中准确判断开刀位置。     

乳腺断层摄影合成技术的临床价值

目的:探讨数字乳腺断层合成X线成像(DBT)结合合成2D图像(SM)对乳腺微钙化的检出和诊断效能。方法:回顾性分析228例乳腺影像及病理资料。3名影像医师独立阅读DBT结合全视野数字化乳腺摄影(FFDM)、DBT结合SM、FFDM、SM 4种模式下影像资料,记录微钙化有无,根据BI-RADS 2013版对微钙化进行分类,分析不同密度乳腺类型中良、恶性微钙化的检出率及诊断效能。结果:不管在致密型乳腺或所有腺体类型乳腺中,4种阅片模式对微钙化检出敏感度的差异无统计学意义(P>0.05),特异度均为100%。DBT结合SM与DBT结合FFDM对微钙化诊断敏感度、特异度及ROC曲线下面积的差异无统计学意义(P>0.05);FFDM的敏感度高于SM,特异度低于SM,ROC曲线下面积高于SM,差异均具有统计学意义(P<0.05)。结论:DBT结合SM与DBT结合FFDM对乳腺微钙化的检出、诊断效能相似。     

电子射野影像系统（EPIDs）图像质量控制方法的探索

目的:拍摄模体边界图像,获得模体图像的对比度,信号噪声比和调制传输函数,通过获得的这些指标来寻找一个操作方便又能够量化的EPIDs图像质量控制的方法:材料和方法:Elekta iViewGT非晶硅阵列电子影像系统,ElekataPrecise加速器(光子8MV).4mm厚度平板铅体模,测量边界垂直于模体表面,同时模体边界处在照射野中心轴,模体表面垂直于射线束;在SSD=100 cm处使用8兆光子线和100 MU获得模体边界图像;使用基于matlab图像分析软件处理获得的图像,计算射线穿透模体和空气后的图像灰度信号关系,获得模体图像的对比度和信号噪声比,同时通过换算,获得探测器效率函数;根据模体的边界图像,得到边界函数,对边界函数进行微分获得线扩展函数,对线扩展函数进行傅里叶变换得到图像的调制传输函数(MTF),通过空间分辨率的方式从另一个角度描述图像的对比度;通过对比度(CNR),信噪比(SNR),调制传输函数(MTF)和探测效率等指标定量的评价EPIDs成像质量.结果:获得了模体图像的CNR、SNR、MTF(50)和探测效率的测量值,同时测量值在随时间推移而下降;结论:此种方法可行,可以长期观测EPIDs系统的成像质量,根据拍摄模体获得的数据对EPIDs系统的图像质量作监测,根据监测的结果决定对EPIDs的维护内容和频次,进而完成对EPIDs影像系统的质量控制和质量保证工作.     

基于滑动块的深度卷积神经网络乳腺X线摄影图像肿块分割算法

目的:提出一种基于滑动块的深度卷积神经网络局部分类、整图乳腺肿块分割的算法,为临床诊断提供有效的肿块形态特征。方法:首先通过区域生长算法和膨胀算法提取患者乳腺区域,并进行数据归一化操作。为了得到每一个像素位置上的诊断信息,在图像的对应位置中滑动提取肿块类及非肿块类图像块,根据卷积神经网络提取其中的纹理信息并对图像块进行分类。通过整合图像块的预测分类结果,进行由粗到细的肿块分割,获得乳腺整图中像素级别的肿块分割。结果:通过比较先进的深度卷积神经网络模型,本文算法滑动块分类结果DenseNet模型下准确率达到96.71%,乳腺X线摄影图像全图肿块分割结果F1-score最优为83.49%。结论:本算法可以分割出乳腺X线摄影图像中的肿块,为后续的乳腺病灶诊断提供可靠的基础。     

脑梗塞患者MR图像分割的研究

目的研究一种可实现脑梗塞患者MR图像脑组织分割的算法.方法根据脑梗塞患者MR图像中脑组织的区域和边缘的特性,对传统水平集算法进行改进,实现了对特定目标体分割的能力,降低了边界泄漏发生的可能性.结果通过体膜和大量脑梗塞患者MR图像实验和SPM5对比,实验证实了改进算法对MR图像分割的准确性和鲁棒性. 结论该算法为脑梗塞患者的脑图像分析和脑组织测量提供了一种有效的分割方法.     

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。     

Automatic quantitative low contrast analysis of digital chest phantom radiographs

Up to the present, the majority of low contrast object evaluations performed on phantom images have been accomplished in a subjective fashion. This is mainly due to the time and effort required to manually measure each radiograph with a densitometer to obtain quantitative results. However, with the development of digital radiographic systems, it has become feasible to automate the detection and computation processes. In this work, a method that can automatically detect and compute the subject-to-noise ratio of the low contrast disks inside a geometric chest phantom is examined. This algorithm has the ability to locate the low contrast objects to an accuracy of less than one pixel, and provides results that are consistent with the understanding of subject-to-noise ratio. This algorithm should simplify the task of quantitative evaluation of contrast detail phantoms.     

Quality assurance for the Leksell gamma unit: considering magnetic resonance image-distortion and delineation failure in the targeting of the internal auditory canal

Our aim in this study was to distinguish quantitatively between the localization accuracy of a commercially available stereotactic fixation device as claimed by the manufacturer and the target accuracy as measured by a user, applying neuroradiologic imaging in Gamma Knife planning and phantom irradiation. Missing the target is the most serious possible failure in Gamma Knife and Linac therapy. To reduce this risk, we developed a quality control algorithm and designed a phantom. To evaluate the accuracy of the targeting procedure with a Leksell Gamma unit, and to experience the possible errors in all procedural steps, irradiations of phantoms were performed, using the so-called "unknown" targeting method. Accuracy is defined by the extent of spatial deviation of the irradiated target from the calculated target. Digital imaging was used for therapy planning. GafChromic films, which had been irradiated while affixed to a specially developed phantom, were used for measuring the precision of the radiation unit. A series of MR images (in two plains: transverse and coronal) was acquired sequentially to image the three-dimensional (3-D) volume of the phantom. The results obtained for isocentric accuracy of the Leksell Gamma unit, model B, were in good agreement to the calculated position. The observed spatial deviations between calculated and irradiated targets is less than 1 mm. The newly designed phantom and quality control algorithm are useful in quality assurance measurements of stereotactic radiation therapy.     

The impact of calibration phantom errors on dual-energy digital mammography

Microcalcification is one of the earliest and main indicators of breast cancer. Because dual-energy digital mammography could suppress the contrast between the adipose and glandular tissues of the breast, it is considered a promising technique that will improve the detection of microcalcification. In dual-energy digital mammography, the imaged object is a human breast, while in calibration measurements only the phantoms of breast tissue equivalent materials are available. Consequently, the differences between phantoms and breast tissues will lead to calibration phantom errors. Based on the dual-energy imaging model, formulae of calibration phantom errors are derived in this paper. Then, this type of error is quantitatively analyzed using publicly available data and compared with other types of error. The results demonstrate that the calibration phantom error is large and dominant in dual-energy mammography, seriously decreasing calculation precision. Further investigations on the physical meaning of calibration phantom error reveal that the imaged objects with the same glandular ratio have identical calibration phantom error. Finally, an error correction method is proposed based on our findings.     

Hybrid computational phantoms of the male and female newborn patient: NURBS-based whole-body models

Anthropomorphic computational phantoms are computer models of the human body for use in the evaluation of dose distributions resulting from either internal or external radiation sources. Currently, two classes of computational phantoms have been developed and widely utilized for organ dose assessment: (1) stylized phantoms and (2) voxel phantoms which describe the human anatomy via mathematical surface equations or 3D voxel matrices, respectively. Although stylized phantoms based on mathematical equations can be very flexible in regard to making changes in organ position and geometrical shape, they are limited in their ability to fully capture the anatomic complexities of human internal anatomy. In turn, voxel phantoms have been developed through image-based segmentation and correspondingly provide much better anatomical realism in comparison to simpler stylized phantoms. However, they themselves are limited in defining organs presented in low contrast within either magnetic resonance or computed tomography images-the two major sources in voxel phantom construction. By definition, voxel phantoms are typically constructed via segmentation of transaxial images, and thus while fine anatomic features are seen in this viewing plane, slice-to-slice discontinuities become apparent in viewing the anatomy of voxel phantoms in the sagittal or coronal planes. This study introduces the concept of a hybrid computational newborn phantom that takes full advantage of the best features of both its stylized and voxel counterparts: flexibility in phantom alterations and anatomic realism. Non-uniform rational B-spline (NURBS) surfaces, a mathematical modeling tool traditionally applied to graphical animation studies, was adopted to replace the limited mathematical surface equations of stylized phantoms. A previously developed whole-body voxel phantom of the newborn female was utilized as a realistic anatomical framework for hybrid phantom construction. The construction of a hybrid phantom is performed in three steps: polygonization of the voxel phantom, organ modeling via NURBS surfaces and phantom voxelization. Two 3D graphic tools, 3D-DOCTOR and Rhinoceros, were utilized to polygonize the newborn voxel phantom and generate NURBS surfaces, while an in-house MATLAB code was used to voxelize the resulting NURBS model into a final computational phantom ready for use in Monte Carlo radiation transport calculations. A total of 126 anatomical organ and tissue models, including 38 skeletal sites and 31 cartilage sites, were described within the hybrid phantom using either NURBS or polygon surfaces. A male hybrid newborn phantom was constructed following the development of the female phantom through the replacement of female-specific organs with male-specific organs. The outer body contour and internal anatomy of the NURBS-based phantoms were adjusted to match anthropometric and reference newborn data reported by the International Commission on Radiological Protection in their Publication 89. The voxelization process was designed to accurately convert NURBS models to a voxel phantom with minimum volumetric change. A sensitivity study was additionally performed to better understand how the meshing tolerance and voxel resolution would affect volumetric changes between the hybrid-NURBS and hybrid-voxel phantoms. The male and female hybrid-NURBS phantoms were constructed in a manner so that all internal organs approached their ICRP reference masses to within 1%, with the exception of the skin (-6.5% relative error) and brain (-15.4% relative error). Both hybrid-voxel phantoms were constructed with an isotropic voxel resolution of 0.663 mm--equivalent to the ICRP 89 reference thickness of the newborn skin (dermis and epidermis). Hybrid-NURBS phantoms used to create their voxel counterpart retain the non-uniform scalability of stylized phantoms, while maintaining the anatomic realism of segmented voxel phantoms with respect to organ shape, depth and inter-organ positioning.     

Producing homogeneous cryogel phantoms for medical imaging: a finite-element approach

Tissue-mimicking phantoms with well-defined properties can help in identifying the potential weaknesses in medical imaging systems. Among the imaging systems, magnetic resonance elastography is a new noninvasive technique used to quantify the shear modulus of biological tissues, and therefore has shown promise in studying liver and brain pathologies. Polyvinyl alcohol (PVA) cryogel prepared by the freeze-thaw technique is a potential candidate for mimicking the mechanical properties of soft tissues and has been extensively used as a phantom material. However, large PVA cryogels suffer from variations in properties, partly due to the low thermal conductivity of PVA solution. The loss of homogeneity in cryogel phantoms is also attributed to inhomogeneous thawing rates during the freeze-thaw cycle. We have used a modified freeze-thaw process that imposes multiple isotherms so as to enhance the homogeneity of the produced cryogels. In addition, we have developed a finite-element modeling tool (a virtual controller) to optimize the temperature profile during the freeze-thaw cycle. Our experimental validations demonstrated the potential of the virtual controller in predicting the optimal temperature profile for the freeze-thaw process (phantom diameters: 60 and 100 mm). A robust simulation framework can fill the gap in the scientific literature with regard to phantom design for medical imaging and will help to reduce phantom development time and cost.     

A Monte Carlo based method to estimate radiation dose from multidetector CT (MDCT): cylindrical and anthropomorphic phantoms

The purpose of this work was to extend the verification of Monte Carlo based methods for estimating radiation dose in computed tomography (CT) exams beyond a single CT scanner to a multidetector CT (MDCT) scanner, and from cylindrical CTDI phantom measurements to both cylindrical and physical anthropomorphic phantoms. Both cylindrical and physical anthropomorphic phantoms were scanned on an MDCT under the specified conditions. A pencil ionization chamber was used to record exposure for the cylindrical phantom, while MOSFET (metal oxide semiconductor field effect transistor) detectors were used to record exposure at the surface of the anthropomorphic phantom. Reference measurements were made in air at isocentre using the pencil ionization chamber under the specified conditions. Detailed Monte Carlo models were developed for the MDCT scanner to describe the x-ray source (spectra, bowtie filter, etc) and geometry factors (distance from focal spot to isocentre, source movement due to axial or helical scanning, etc). Models for the cylindrical (CTDI) phantoms were available from the previous work. For the anthropomorphic phantom, CT image data were used to create a detailed voxelized model of the phantom's geometry. Anthropomorphic phantom material compositions were provided by the manufacturer. A simulation of the physical scan was performed using the mathematical models of the scanner, phantom and specified scan parameters. Tallies were recorded at specific voxel locations corresponding to the MOSFET physical measurements. Simulations of air scans were performed to obtain normalization factors to convert results to absolute dose values. For the CTDI body (32 cm) phantom, measurements and simulation results agreed to within 3.5% across all conditions. For the anthropomorphic phantom, measured surface dose values from a contiguous axial scan showed significant variation and ranged from 8 mGy/100 mAs to 16 mGy/100 mAs. Results from helical scans of overlapping pitch (0.9375) and extended pitch (1.375) were also obtained. Comparisons between the MOSFET measurements and the absolute dose value derived from the Monte Carlo simulations demonstrate agreement in terms of absolute dose values as well as the spatially varying characteristics. This work demonstrates the ability to extend models from a single detector scanner using cylindrical phantoms to an MDCT scanner using both cylindrical and anthropomorphic phantoms. Future work will be extended to voxelized patient models of different sizes and to other MDCT scanners.     

A patient-equivalent attenuation phantom for estimating patient exposures from automatic exposure controlled x-ray examinations of the abdomen and lumbo-sacral spine

The Joint Commission on Accreditation of Healthcare Organizations requires diagnostic radiology facilities to known the approximate amount of radiation received by an average patient during radiographic examinations at the facility. Automatic exposure controlled (AEC) techniques are used for many of these exams, and a standard patient-equivalent phantom is necessary when estimating patient exposure on such systems. This is of particular importance if exposures are to be compared among AEC systems with different entrance x-ray spectra. We have developed a phantom, LucA1 Abdomen, to facilitate determining the average patient exposure from AEC anteroposterior (AP) abdomen and lumbo-sacral (LS) spine radiography. The phantom is relatively lightweight, transportable, sturdy, and made of readily available inexpensive materials (Lucite and aluminum). It accurately simulates the primary and scatter transmission through the soft tissue and L-4 spinal regions of a patient-equivalent anthropomorphic phantom for x-ray spectra typically used in abdomen/LS spine radiography. A clinical evaluation to verify the patient-equivalence of three commercial anthropomorphic phantoms (Humanoid, Rando, 3-M) and two acrylic/aluminum phantoms (ANSI and LucA1 Abdomen) has been conducted. The design and development of the LucA1 Abdomen phantom and the evaluation of all phantoms is described.     

Evaluating radiographic parameters for mobile chest computed radiography: phantoms, image quality and effective dose

Conventional chest radiography is technically difficult because of wide variations in tissue attenuations in the chest and limitations of screen-film systems. Mobile chest radiography, performed bedside on hospital inpatients, presents additional difficulties due to geometric and equipment limitations inherent in mobile x-ray procedures and the severity of illness in the patients. Computed radiography (CR) offers a different approach for mobile chest radiography by utilizing a photostimulable phosphor. Photostimulable phosphors overcome some image quality limitations of mobile chest imaging, particularly because of the inherent latitude. Because they are more efficient in absorbing lower-energy x-rays than rare-earth intensifying screens, this study evaluated changes in kVp for improving mobile chest CR. Three commercially available systems were tested, with the goal of implementing the findings clinically. Exposure conditions (kVp and grid use) were assessed with two acrylic-and-aluminum chest phantoms which simulated x-ray attenuation for average-sized and large-sized adult chests. These phantoms contained regions representing the lungs, heart and subdiaphragm to allow proper CR processing. Signal-to-noise ratio (SNR) measurements using different techniques were obtained for acrylic and aluminum disks (1.9 cm diameter) superimposed in the lung and heart regions of the phantoms, where the disk thicknesses (contrast) were determined from disk visibility. Effective doses to the phantoms were also measured for these techniques. The results indicated that using an 8:1, 33 lines/cm antiscatter grid improved the SNR by 60-300 % compared with nongrid images, depending on phantom and region; however, the dose to the phantom also increased by 400-600%. Lowering x-ray tube potential from 80 to 60 kVp improved the SNR by 30-40%, with a corresponding increase in phantom dose of 40-50%. Increasing the potential from 80 to 100 kVp reduced both the SNR and the phantom dose by approximately 10%. The most promising changes in technique for trial in clinical implementation include using an antiscatter grid, especially for large patients, and potentially increasing kVp.     

Realistic CT simulation using the 4D XCAT phantom

The authors develop a unique CT simulation tool based on the 4D extended cardiac-torso (XCAT) phantom, a whole-body computer model of the human anatomy and physiology based on NURBS surfaces. Unlike current phantoms in CT based on simple mathematical primitives, the 4D XCAT provides an accurate representation of the complex human anatomy and has the advantage, due to its design, that its organ shapes can be changed to realistically model anatomical variations and patient motion. A disadvantage to the NURBS basis of the XCAT, however, is that the mathematical complexity of the surfaces makes the calculation of line integrals through the phantom difficult. They have to be calculated using iterative procedures; therefore, the calculation of CT projections is much slower than for simpler mathematical phantoms. To overcome this limitation, the authors used efficient ray tracing techniques from computer graphics, to develop a fast analytic projection algorithm to accurately calculate CT projections directly from the surface definition of the XCAT phantom given parameters defining the CT scanner and geometry. Using this tool, realistic high-resolution 3D and 4D projection images can be simulated and reconstructed from the XCAT within a reasonable amount of time. In comparison with other simulators with geometrically defined organs, the XCAT-based algorithm was found to be only three times slower in generating a projection data set of the same anatomical structures using a single 3.2 GHz processor. To overcome this decrease in speed would, therefore, only require running the projection algorithm in parallel over three processors. With the ever decreasing cost of computers and the rise of faster processors and multi-processor systems and clusters, this slowdown is basically inconsequential, especially given the vast improvement the XCAT offers in terms of realism and the ability to generate 3D and 4D data from anatomically diverse patients. As such, the authors conclude that the efficient XCAT-based CT simulator developed in this work will have applications in a broad range of CT imaging research.