自适应统计迭代重建算法在儿童头部CT扫描中的应用

目的 探讨自适应统计迭代重建(ASiR)算法在儿童头部CT扫描中的应用价值。方法 对1个水模进行CT扫描,管电压100 kV,管电流分别为200、180、160、140、120和100 mA,ASiR比例分别设置为0、10%、20%、30%、40%、50%,采用FBP和ASiR迭代重建两种重建算法进行图像重建,比较不同条件下图像CNR、SNR和图像噪声。将80例接受头部CT扫描的患儿分为对照组(n=40)和试验组(n=40)。对照组采用管电压100 kV、管电流200 mA,FBP重建算法进行图像重建;试验组采用管电压100 kV、管电流140 mA,分别采用FBP和ASiR两种重建法进行图像重建;将重建后的图像分别记为试验FBP亚组和试验ASiR亚组;对两组中图像的CNR、SNR、图像噪声、CTDIvol、DLP、ED进行比较。结果 水模研究中,采用ASiR(30%)、管电流140 mA、ASiR算法重建图像的SNR、CNR及空气噪声值与管电流 200 mA、FBP重建图像最为接近。对照组与试验ASiR亚组图像的图像噪声、灰白质CNR和灰质SNR值差异均无统计学意义(P均>0.05);对照组和试验ASiR亚组图像的噪声、灰质SNR、灰白质CNR均优于试验FBP亚组,且差异均有统计学意义(P均<0.001)。结论 采用ASiR重建算法的头部CT扫描,既可降低患儿接受的辐射剂量,又保证了图像质量,具有较高的临床应用价值。     

基于交替投影的CT图像重建算法

目的 探讨基于交替投影的CT图像重建算法的可行性。方法 将CT图像的重建转化为凸集优化问题,将重建模型分解为多个约束,并确定其对应的凸集,通过交替投影的方式在其交集中找到可行解。对TV先验项构成的凸集的求解,通过定义TV函数的上方图集,利用点到这个凸集的切向超平面的连续投影,找到起始点到TV凸集最近的点。分别采用CPTV算法、TV-POCS算法和基于交替投影的CT图像重建算法对Shepp-Logan头部图像进行重建,比较不同算法对不同角度投影图像重建后的均方根误差（RMSE）。分别采用TV-POCS算法和基于交替投影的CT图像重建算法对小鼠脊椎轴位图像进行重建,比较两种算法的归一化均方距离（d）和归一化平均绝对距离（r）。结果 CPTV算法所重建的图像平滑性较差,伪影较多,而TV-POCS算法和基于交替投影的重建算法不仅有效抑制了噪声,还保护了图像的边缘,图像质量较高。基于交替投影的重建算法的RMSE比另外两种算法下降速度更快,收敛值更小。基于交替投影的重建算法重建图像的d和r值均小于TV-POCS算法（0.064 0 vs 0.262 4,0.073 7 vs 0.298 2）。结论 采用基于交替投影的重建算法重建有限角度的CT投影图像不需参数估计,且图像质量更高,收敛速度更快。     

重建算法对SPECT空间分辨力的影响

目的 探讨重建算法对SPECT断层成像分辨力的影响。方法 采用Siemens Symbia T SPECT机,对长轴分布5条线源的椭圆柱分辨力模型进行成像。选择Hann和Butterworth 2种滤波反投影法（FBP）,每种窗函数各选择3种截止频率;Flash 3D和有序子集最大期望值（OSEM）法2种迭代法,各选择5种平滑函数;分别对图像进行断层重建。以线源半高宽（FWHM）代表重建图像的分辨力,计算5条线源6个层面上的径向、切向和轴向FWHM,并行多元线性回归分析。结果 FBP重建中,FWHM随窗函数截止频率增大而变小;同一截止频率下,Butterworth重建图像的FWHM小于Hann。迭代法中,FWHM随平滑函数核宽度增大而变大;Flash 3D和OSEM法重建图像径向及切向FWHM差异无统计学意义,而轴向Flash 3D图像的FWHM显著小于OSEM。结论 重建算法及参数设定对断层图像分辨力的影响较大,应根据临床需要和经验适当选择。     

基于正侧位锥束CT投影快速构建脊椎三维模型

目的 探讨基于C形臂的两张正侧位2D锥束CT（CBCT）投影图像进行3D模型重建的效果。方法 采用半自动化的二维投影图像特征点提取算法,选定18点的特征点集并提取其对应的正侧位二维投影图像平面坐标,然后针对C形臂CBCT系统建立坐标系,推导特征点三维空间坐标与其在投影图像中平面坐标之间的几何关系,代入转换公式获得特征点集的三维空间坐标。利用薄板样条法对三维脊椎基础模型进行空间非刚性插值获得三维脊椎目标模型。将L3石膏模型置入C形臂CBCT系统,获取375幅圆周扫描图像,利用FDK算法重建三维模型并进行表面重采样得到三维脊椎参考模型,对其进行不规则形状调制得到三维脊椎基础模型,利用本文方法对三维脊椎基础模型进行空间非刚性插值得到三维脊椎目标模型,并设置对照组对本文方法的精度进行评价。结果 相对于特征点手动提取和边缘增强提取算法,采用半自动化特征点提取算法构建获得的三维脊椎目标模型与参考模型的误差降低至1 mm以内。结论 采用本文方法可构建出近似的、精度较高的脊椎三维模型,为基于C形臂CBCT的手术导航提供3D图像支持。     

X线相衬CT成像显示人体胆道闭锁样品微脉管结构

目的 探讨X射线相衬CT（PCCT）对人体胆道闭锁（BA）样品中脉管显微结构的显示作用。方法 将4份人体BA样品进行冲洗、固定并以乙醇脱水干燥,采用上海同步辐射光源BL13W1线站进行PCCT成像,通过图像处理分别获得滤波反投影算法重建图像、样品整体及微脉管的3D重建图像,并对不同类型脉管进行区分。而后对样品进行石蜡包埋,4 μm切片,并行平滑肌细胞抗体（SMA）、细胞角蛋白19抗体（CK19）免疫组化染色及苦味酸天狼星红染色组织病理检查。对比观察滤波反投影算法及微脉管3D重建图像与相应病理图。结果 滤波反投影算法重建CT图像及微脉管3D重建图像均可用以区分人体BA样品中的肝动脉、胆管及增生胆管、门静脉,并清晰显示不同类型脉管的结构特征,所示微脉管结构与组织病理学检查相符。结论 结合滤波反投影重建算法及3D可视化技术,PCCT能够清晰显示离体BA组织的微脉管结构。     

不同CT图像重建算法下基于深度学习的肺结节检测算法效能

目的 探索CT图像重建算法对于基于深度学习（DL）的肺结节检测算法的影响。方法 选取298例接受肺部CT检查患者,依次采用肺窗重建、纵隔重建、骨窗重建3种算法重建CT图像。先由2名主治医师对入组病例进行标注,结果不一致时由1名高年资医师进行审核,以结果作为金标准。以深度神经网络为基础构建肺结节检测算法,与医师标注结果进行比对,得到算法在不同重建方法下检出肺结节的敏感度、准确率、F分数等指标以及模型检出的假阳性分布,对比分析模型在不同CT图像重建算法下的诊断效果。结果 基于DL的肺结节检测算法在肺重建、纵隔重建和骨重建3种重建方法下的敏感度分别为92.33%（313/339）、86.97%（287/330）及92.73%（319/344）,准确率分别为23.55%（313/1 329）、37.91%（287/757）及27.84%（319/1 146）,F分数分别为0.38、0.53及0.43,3种算法重建下模型检出敏感度、模型误检结节类型与医师漏标结节类型差异均无统计学意义（P均>0.05）。结论 基于DL的肺结节检测算法在肺窗、纵隔和骨窗重建下均性能优良,能帮助医生提高工作效率和诊断质量。     

非等中心锥束CT成像系统标定及三维重建

目的 研究非等中心锥束CT系统(CBCT)几何参数标定及投影图像重排, 提出新的方法以提高参数标定精度及三维(3D)重建质量, 以期突破现有算法要求等中心系统应用环境的限制。方法 针对非等中心CBCT系统建立坐标系, 推导标定模型中特定标志点空间坐标与其在投影图像中平面坐标之间的几何关系, 建立虚拟等中心CBCT系统, 提出面向非等中心CBCT系统的解析+迭代混合标定方法, 对投影图像进行重排, 在此基础上进行3D重建。将标定模型置于非等中心CBCT系统中, 获取各个旋转角度下的投影图像;提取投影图像中特定标志点坐标, 基于这些坐标数据, 采用所提方法计算虚拟等中心CBCT系统几何参数;并求取虚拟等中心锥束投影图像;最后采用FDK算法进行3D重建。结果 与文献[10]算法相比, 所提算法对转轴在成像板上的投影坐标(u^0' 和v^0' )标定精度相当, 而射线源到成像板的垂直距离(D')和射线源到转轴的距离(R')标定精度明显较优。结论 本研究混合标定算法可提高D'和R'的标定精度, 并可突破文献[10]算法等中心系统应用环境的限制。     

一种混合CBCT成像系统标定方法

目的 建立一种混合锥形束CT（CBCT）成像系统标定方法,以提高CBCT成像系统参数标定精度,突破现有算法须已知标定模型中各小球相对中平面位置的限制。方法 针对CBCT成像系统建立坐标系,推导光锥不同旋转角度下标定模型中小球中心空间坐标与其在投影图像中投影平面坐标之间的几何关系,提出解析+迭代混合标定方法。将标定模型置于CBCT成像系统中,获取各个旋转角度下的投影图像;提取投影图像中各小球投影的质心坐标,套用上述方法计算出标定参数。对小鼠标本模型进行相同扫描,采用已求标定参数用FDK算法进行3D重建。结果 解析+迭代混合法对于成像系统参数D的标定精度较高;参数R的标定不直接受D的影响。结论 解析+迭代混合法可提高部分成像参数的标定精度,且突破了须事先知晓小球相对于中平面的空间位置及受旋转半径影响较大的限制。     

基于噪声功率谱的CT图像噪声评价

目的 设计并探讨基于噪声功率谱（NPS）的CT图像噪声评价方法。方法 改变CT图像扫描参数以及重建参数获取图像,采用NPS与传统标准差（SD）两种方法对该图像进行噪声检测,并对比两种检测结果。结果 NPS曲线幅值与层厚和剂量成反比,与重建算法的增强程度成正比。重建算法、层厚和剂量对噪声指数的影响能够在NPS曲线上得到明显的体现。结论 NPS能够从幅值与频率方面反应噪声变化,与SD方法相结合可形成更为完善的噪声评价体系。     

深度学习重建算法优化能谱CT低单能量图像质量及检测肝脏低对比度小病灶能力

目的 观察深度学习重建(DLIR)算法用于优化能谱CT低单能量图像质量及提高检测肝脏低对比度小病灶能力的可行性。方法 纳入30例接受上腹部门脉期增强扫描的肝脏疾病患者,包括58个肝脏病灶,分别采用DLIR及基于混合模型的自适应统计迭代重建(ASIR-V)算法重建40~70 keV (间隔10 keV)单能量图像;根据肝脏、门静脉及肝脏病灶对比噪声比(CNR)和噪声进行主观评价,针对图像总体质量、病灶显著性和诊断信心评分进行主观评价,比较不同图像之间评价结果的差异。结果 相比ASIR-V图像,40~70 keV能级下,DLIR图像的CNR_肝脏、CNR_门静脉及CNR_肝脏病灶均显著增加而噪声均显著减少(P均<0.05);40~60 keV能级下,DLIR图像总体质量、病灶显著性及诊断信心评分均高于ASIR-V图像(P均<0.05)。结论 DLIR技术可显著减少低单能量成像噪声、改善图像质量并提高检测肝脏低对比度小病灶的能力。     

Electrical impedance tomography of human brain function using reconstruction algorithms based on the finite element method

Electrical impedance tomography (EIT) is a recently developed technique which enables the internal conductivity of an object to be imaged using rings of external electrodes. In a recent study, EIT during cortical evoked responses showed encouraging changes in the raw impedance measurements, but reconstructed images were noisy. A simplified reconstruction algorithm was used which modelled the head as a homogeneous sphere. In the current study, the development and validation of an improved reconstruction algorithm are described in which realistic geometry and conductivity distributions have been incorporated using the finite element method. Data from computer simulations and spherical or head-shaped saline-filled tank phantoms, in which the skull was represented by a concentric shell of plaster of Paris or a real human skull, have been reconstructed into images. There were significant improvements in image quality as a result of the incorporation of accurate geometry and extracerebral layers in the reconstruction algorithm. Image quality, assessed by blinded subjective expert observers, also improved significantly when data from the previous evoked response study were reanalysed with the new algorithm. In preliminary images collected during epileptic seizures, the new algorithm generated EIT conductivity changes which were consistent with the electrographic ictal activity. Incorporation of realistic geometry and conductivity into the reconstruction algorithm significantly improves the quality of EIT images and lends encouragement to the belief that EIT may provide a low-cost, portable functional neuroimaging system in the foreseeable future.     

Level-set-based reconstruction algorithm for EIT lung images: first clinical results

We show the first clinical results using the level-set-based reconstruction algorithm for electrical impedance tomography (EIT) data. The level-set-based reconstruction method (LSRM) allows the reconstruction of non-smooth interfaces between image regions, which are typically smoothed by traditional voxel-based reconstruction methods (VBRMs). We develop a time difference formulation of the LSRM for 2D images. The proposed reconstruction method is applied to reconstruct clinical EIT data of a slow flow inflation pressure-volume manoeuvre in lung-healthy and adult lung-injury patients. Images from the LSRM and the VBRM are compared. The results show comparable reconstructed images, but with an improved ability to reconstruct sharp conductivity changes in the distribution of lung ventilation using the LSRM.     

Imaging of conductivity changes and electrode movement in EIT

Electrical impedance tomography (EIT) attempts to reconstruct the internal impedance distribution in a medium from electrical measurements at electrodes on the medium surface. One key difficulty with EIT measurements is due to the position uncertainty of the electrodes, especially for medical applications, in which the body surface moves during breathing and posture change. In this paper, we develop a new approach which directly reconstructs both electrode movements and internal conductivity changes for difference EIT. The reconstruction problem is formulated in terms of a regularized inverse, using an augmented Jacobian, sensitive to impedance change and electrode movement. A reconstruction prior term is computed to impose a smoothness constraint on both the spatial distribution of impedance change and electrode movement. A one-step regularized imaging algorithm is then implemented based on the augmented Jacobian and smoothness constraint. Images were reconstructed using the algorithm of this paper with data from simulated 2D and 3D conductivity changes and electrode movements, and from saline phantom measurements. Results showed good reconstruction of the actual electrode movements, as well as a dramatic reduction in image artefacts compared to images from the standard algorithm, which did not account for electrode movement.     

Direct reconstruction of tissue parameters from differential multifrequency EIT in vivo

The basic purpose of electrical impedance tomography (EIT) is the reconstruction of conductivity distributions. While multifrequency measurements are of common use, the majority of reconstructed images are still conductivity distributions from one single frequency. More interesting than conductivities at each frequency are electrical tissue parameters, which describe the frequency-dependent conductivity changes of tissue. These parameters give information about physiological or electrical properties of tissues. By using this spectral information, a classification of different tissue types is possible. To get a distribution of tissue parameters, usually a posterior fitting of a tissue model to the conductivity spectra obtained with classical reconstruction algorithms at various frequencies is used. In this work, a single-step reconstruction algorithm for differential imaging was developed for the direct estimation of Cole parameters. This method is termed differential parametric reconstruction. The Cole model was used as the underlying tissue model, where only the relative changes of the two conductivity parameters sigma(0) and sigma(infinity) were reconstructed and the other two parameters of the model which are less identifiable were set to constant values. The reconstruction algorithm was tested with simulated noisy datasets and real measurement data from EIT measurements on the human thorax. These measurements were taken from healthy subjects and from patients with a serious lung injury. The new method yields a good image quality and higher robustness against noise compared to conventional reconstruction methods.     

Real-time imaging and detection of intracranial haemorrhage by electrical impedance tomography in a piglet model

The aim of this study was to use electrical impedance tomography (EIT) to detect and image acute intracranial haemorrhage (ICH) in an animal model. Blood was infused into the frontal lobe of the brains of anaesthetized piglets and impedance was measured using 16 electrodes placed in a circle on the scalp. The EIT images were constructed using a filtered back-projection algorithm. The mean of all the pixel intensities within a region of interest--the mean resistivity value (MRV)--was used to evaluate the relative impedance changes in the target region. A symmetrical index (SI), reflecting the relative impedance on both sides of the brain, was also calculated. Changes in MRV and SI were associated with the injection of blood, demonstrating that EIT can successfully detect ICH in this animal model. The unique features of EIT may be beneficial for diagnosing ICH early in patients after cranial surgery, thereby reducing the risk of complications and mortality.     

Multi-frequency electrical impedance tomography of the breast: new clinical results

Electrical impedance tomography (EIT) has been used in the recent past for a number of clinical applications. In this work we present recent tomographic and spectroscopic findings for breast imaging from clinical exams completed at Dartmouth. The results presented here are based on 18 normal and 24 abnormal subjects. The participants were classified as normal or abnormal using the American College of Radiology (ACR) indexing system for mammograms. The EIT data were collected for ten discrete frequencies in the range 10 kHz-1 MHz using a single array of 16 electrodes. The finite element method was used to reconstruct the images. The images were examined visually and were compared with mammograms. The results were also analyzed based on zonal averages of property values and breast tissue radiodensities. Statistical analysis showed a significance difference between the mean conductivity and permittivity values of normal and abnormal subjects for various zones defined on the reconstructed images. Tissues with high radiodensity also had increased conductivity and permittivity.     

Reconstruction of conductivity changes and electrode movements based on EIT temporal sequences

Electrical impedance tomography (EIT) reconstructs a conductivity change image within a body from electrical measurements on the body surface; while it has relatively low spatial resolution, it has a high temporal resolution. One key difficulty with EIT measurements is due to the movement and position uncertainty of the electrodes, especially due to breathing and posture change. In this paper, we develop an approach to reconstruct both the conductivity change image and the electrode movements from the temporal sequence of EIT measurements. Since both the conductivity change and electrode movement are slow with respect to the data frame rate, there are significant temporal correlations which we formulate as priors for the regularized image reconstruction model. Image reconstruction is posed in terms of a regularization matrix and a Jacobian matrix which are augmented for the conductivity change and electrode movement, and then further augmented to concatenate the d previous and future frames. Results are shown for simulation, phantom and human data, and show that the proposed algorithm yields improved resolution and noise performance in comparison to a conventional one-step reconstruction method.     

Frequency-difference electrical impedance tomography (fdEIT): algorithm development and feasibility study

Frequency-difference electrical impedance tomography (fdEIT) has been proposed to deal with technical difficulties of a conventional static EIT imaging method caused by unknown boundary geometry, uncertainty in electrode positions and other systematic measurement artifacts. In fdEIT, we try to produce images showing changes of a complex conductivity distribution with respect to frequency. Simultaneously injecting currents with at least two frequencies, we find differences of measured boundary voltages between those frequencies. In most previous studies, real parts of frequency-difference voltage data were used to reconstruct conductivity changes and imaginary parts to reconstruct permittivity changes. This conventional approach neglects the interplay of conductivity and permittivity upon measured boundary voltage data. In this paper, we propose an improved fdEIT image reconstruction algorithm that properly handles the interaction. It uses weighted frequency differences of complex voltage data and a complex sensitivity matrix to reconstruct frequency-difference images of complex conductivity distributions. We found that there are two major sources of image contrast in fdEIT. The first is a contrast in complex conductivity values between an anomaly and background. The second is a frequency dependence of a complex conductivity distribution to be imaged. We note that even for the case where conductivity and permittivity do not change with frequency, the fdEIT algorithm may show a contrast in frequency-difference images of complex conductivity distributions. On the other hand, even if conductivity and permittivity values significantly change with frequency, there is an example where we cannot find any contrast. The performance of the proposed method is demonstrated by using computer simulations to validate its feasibility in future experimental studies.     

Imaging cryosurgery with EIT: tracking the ice front and post-thaw tissue viability

Cryosurgery employs freezing for targeted destruction of undesirable tissues such as cancer. Ice front imaging has made controlled treatment of deep body tumors possible. One promising method, recently explored for this task, is EIT, which recovers images of electrical impedance from measurements made at boundary electrodes. However, since frozen tissue near the ice front survives, ice front imaging is insufficient. Monitoring treatment effect would enable iterative cryosurgery, where extents of ablation and need for further treatment are assessed upon thawing. Since lipid bilayers are strong barriers to low frequency electrical current and cell destruction implies impaired membranes, EIT should be able to detect the desired effect of cryosurgery: cell death. Previous work has tested EIT for ice front imaging with tank studies while others have simulated EIT in detecting cryoablation, but in vivo tests have not been reported in either case. To address this, we report 3D images of differential conductivity throughout the freeze-thaw cycle in a rat liver model in vivo with histological validation, first testing our system for ice front imaging in a gel and for viability imaging post-thaw in a raw potato slice.     

Validation of a 3D reconstruction algorithm for EIT of human brain function in a realistic head-shaped tank

Previous work has demonstrated that electrical impedance tomography can be used to image human brain activity during evoked responses, but two-thirds of the reconstructed images fail to localize an impedance change to the expected stimulated cortical area. The localization failure may be caused by modelling the head as a homogenous sphere in the reconstruction algorithm. This assumption may lead to errors when used to reconstruct data obtained from the human head. In this study a 3D reconstruction algorithm, based on a model of the head as a homogenous sphere, was characterized by simulating the algorithm model, the head shape and the presence of the skull in saline-filled tanks. EIT images of a sponge, 14 cm3 volume with a resistivity contrast of 12%, were acquired in three different positions in tanks filled with 0.2% saline. In a hemispherical tank, 19 cm in diameter, the sponge was localized to within 3.4-10.7% of the tank diameter. In a head-shaped tank, the errors were between 3.1 and 13.3% without a skull and between 10.3 and 18.7% when a real human skull was present. A significant increase in localization error therefore occurs if an algorithm based on a homogeneous sphere is used on data acquired from a head-shaped tank. The increased error is due to the presence of the skull, as no significant increase in error occurred if a head-shaped tank was used without the skull present, compared to the localization error within the hemispherical tank. The error due to the skull significantly shifted the impedance change within the skull towards the centre of the image. Although the increased localization error due to the skull is not sufficient to explain the localization errors of up to 50% of the image diameter present in the images of some human subjects, the future use of a realistic head model in the reconstruction algorithm is likely to reduce the localization error in the human images due to the presence of the skull.