基于深度学习单投影的CT断层成像三维重建

【摘要】目的:提出一种基于深度学习的计算机断层扫描（CT）单视图断层成像三维（3D）重建方法,在减少数据采集量和降低成像剂量的情况下对不同患者进行CT图像的3D重建。方法:对不同患者的CT图像进行数据增强和模拟生成对应的数字重建放射影像（DRR）,并进行数据归一化操作。利用预处理后的数据通过卷积神经网络训练出一个普适于不同患者的神经网络模型。将训练好的神经网络模型部署在测试数据集上,使用平均绝对误差（MAE）、均方根误差（RMSE）、结构相似性（SSIM）和峰值信噪比（PSNR）对重建结果进行评估。结果:定性和定量分析的结果表明,该方法可以使用不同患者的单张2D图像分别重建出质量较高的3D CT图像,MAE、RMSE、SSIM和PSNR分别为0.006、0.079、0.982、38.424 dB。此外,相比特定于单个患者的情况,该方法可以大幅度提高重建速度并节省70%的模型训练时间。结论:构建的神经网络模型可通过不同患者的2D单视图重建出相应患者的3D CT图像。因此,本研究对简化临床成像设备和放射治疗当中的图像引导具有重要作用。     

The HUDSEN Atlas: a three-dimensional (3D) spatial framework for studying gene expression in the developing human brain

We are developing a three-dimensional (3D) atlas of the human embryonic brain using anatomical landmarks and gene expression data to define major subdivisions through 12 stages of development [Carnegie Stages (CS) 12-23; approximately 26-56 days post conception (dpc)]. Virtual 3D anatomical models are generated from intact specimens using optical projection tomography (OPT). Using MAPAINT software, selected gene expression data, gathered using standard methods of in situ hybridization and immunohistochemistry, are mapped to a representative 3D model for each chosen Carnegie stage. In these models, anatomical domains, defined on the basis of morphological landmarks and comparative knowledge of expression patterns in vertebrates, are linked to a developmental neuroanatomic ontology. Human gene expression patterns for genes with characteristic expression in different vertebrates (e.g. PAX6, GAD65 and OLIG2) are being used to confirm and/or refine the human anatomical domain boundaries. We have also developed interpolation software that digitally generates a full domain from partial data. Currently, the 3D models and a preliminary set of anatomical domains and ontology are available on the atlas pages along with gene expression data from approximately 100 genes in the HUDSEN Human Spatial Gene Expression Database (http://www.hudsen.org). The aim is that full 3D data will be generated from expression data used to define a more detailed set of anatomical domains linked to a more advanced anatomy ontology and all of these will be available online, contributing to the long-term goal of the atlas, which is to help maximize the effective use and dissemination of data wherever it is generated.     

A novel segmentation method using multiresolution analysis with 3D visualization for X-ray coronary angiogram images

Coronary angiography is a widely used tool in the diagnosis and treatment of cardiac diseases. The main cause of coronary artery disease is atherosclerosis, which leads to the narrowing of artery lumen, resulting in decreased blood supply to heart muscles. Determination of narrowing of the lumens mainly depends upon the quality of the segmented image; with improved segmentation technique there is better accuracy in identification of blocks. The main purpose of the paper is to develop an automatic, accurate segmentation technique with 3D visualization for the segmented images. 3D visualization provides clearer information regarding the shape and severity of the lesion. The thresholding technique is one of the oldest and simplest techniques used for segmentation. This paper proposes a multithresholding approach using the entropy measure and multiresolution analysis to ensure automatic and accurate segmentation by overcoming some of the problems encountered in other techniques. Also, segmentation performance analysis was conducted for various segmentation methods. This method is tested with different real coronary angiographic images and was found to perform better than the other techniques.     

The intrinsic innervation of the lung is derived from neural crest cells as shown by optical projection tomography in Wnt1-Cre;YFP reporter mice

Within the embryonic lung, intrinsic nerve ganglia, which innervate airway smooth muscle, are required for normal lung development and function. We studied the development of neural crest-derived intrinsic neurons within the embryonic mouse lung by crossing Wnt1-Cre mice with R26R-EYFP reporter mice to generate double transgenic mice that express yellow fluorescent protein (YFP) in all neural crest cells (NCCs) and their derivatives. In addition to utilizing conventional immunohistochemistry on frozen lung sections, the complex organization of lung innervation was visualized in three dimensions by combining the genetic labelling of NCCs with optical projection tomography, a novel imaging technique that is particularly useful for the 3D examination of developing organs within embryos. YFP-positive NCCs migrated into the mouse lung from the oesophagus region at embryonic day 10.5. These cells subsequently accumulated around the bronchi and epithelial tubules of the lung and, as shown by 3D lung reconstructions with optical projection tomography imaging, formed an extensive, branching network in association with the developing airways. YFP-positive cells also colonized lung maintained in organotypic culture, and responded in a chemoattractive manner to the proto-oncogene, rearranged during transfection (RET) ligand, glial-cell-line-derived neurotrophic factor (GDNF), suggesting that the RET signalling pathway is involved in neuronal development within the lung. However, when the lungs of Ret^−/− and Gfrα1^−/− embryos, deficient in the RET receptor and GDNF family receptor α 1 (GFRα1) co-receptor respectively, were examined, no major differences in the extent of lung innervation were observed. Our findings demonstrate that intrinsic neurons of the mouse lung are derived from NCCs and that, although implicated in the development of these cells, the role of the RET signalling pathway requires further investigation.     

3D reconstruction method from biplanar radiography using non-stereocorresponding points and elastic deformable meshes

Standard 3D reconstruction of bones using stereoradiography is limited by the number of anatomical landmarks visible in more than one projection. The proposed technique enables the 3D reconstruction of additional landmarks that can be identified in only one of the radiographs. The principle of this method is the deformation of an elastic object that respects stereocorresponding and non-stereo-corresponding observations available in different projections. This technique is based on the principle that any non-stereocorresponding point belongs to a line joining the X-ray source and the projection of the point in one view. The aim is to determine the 3D position of these points on their line of projection when submitted to geometrical and topological constraints. This technique is used to obtain the 3D geometry of 18 cadaveric upper cervical vertebrae. The reconstructed geometry obtained is compared with direct measurements using a magnetic digitiser. The order of precision determined with the point-to-surface distance between the reconstruction obtained with that technique and reference measurements is about 1 mm, depending on the vertebrae studied. Comparison results indicate that the obtained reconstruction is close to the actual vertebral geometry. This method can therefore be proposed to obtain the 3D geometry of vertebrae.     

A semi-automated method using interpolation and optimisation for the 3D reconstruction of the spine from bi-planar radiography: a precision and accuracy study

The 3D reconstruction of the spine in upright posture can be obtained by bi-planar radiographic methods, developed since the 1970s. The principle is to identify 4–25 anatomical landmarks per vertebrae and per images. This identification time is hardly manageable in clinical practice. A semi-automated method is used: 3D standard vertebral models are positioned along with a 3D curve (identified all the way through the vertebral bodies). The silhouettes of the models of C7 and L5 vertebrae are first adjusted and the positions of the other vertebrae are interpolated and optimised. The inter- and intra-operator variabilities and the errors between the semi-automated method and the manual identification of six anatomical landmarks per vertebra are evaluated on 20 pairs of X-ray images of subjects with different spinal deformities. The identification time for the semi-automated method is 5 min. For scolitic subjects, the precision is under 2.2° and the accuracy is under 3.2° for all lateral, sagittal and axial rotations.     

Comparison of mandibular landmarks from computed tomography and 3D digitizer data

We recorded 3D coordinates for 28 mandibular landmarks from three-dimensional reconstructions of CT axial slices using the image analysis program eTDIPS. The images were acquired from a pediatric series of human mandibles (neonate to 13 years of age) from the Bosma collection (Shapiro and Richtsmeier, 1997, Am. J. Phys. Anthropol. 103:415-416). To test the accuracy of these coordinate data, we recorded the same 28 landmarks directly on the Bosma mandibles using a Polhemus 3Space digitizer. The directly digitized landmarks serve as a gold standard upon which to evaluate the eTDIPS data. Standard deviations of landmark placement using eTDIPS show a greater degree of variation compared to the data gathered using the digitizer, although this error is more heavily concentrated in certain types of landmarks. All possible linear distances between unique pairs of landmarks were calculated, and like linear distances were compared between the two data collection methods. The absolute difference for all like linear distances ranged from 0.001-3.9 mm (mean = 0.377 mm; SD = 1.136), with the eTDIPS data being consistently larger than the digitizer coordinates. This study demonstrates that landmark coordinate data can be reliably collected from digital CT images of the human mandible. We define a set of mandibular landmarks useful in evaluating the effects of craniofacial disorders, growth and other biological processes.     

IVUSAngio Tool: A publicly available software for fast and accurate 3D reconstruction of coronary arteries

There is an ongoing research and clinical interest in the development of reliable and easily accessible software for the 3D reconstruction of coronary arteries. In this work, we present the architecture and validation of IVUSAngio Tool, an application which performs fast and accurate 3D reconstruction of the coronary arteries by using intravascular ultrasound (IVUS) and biplane angiography data. The 3D reconstruction is based on the fusion of the detected arterial boundaries in IVUS images with the 3D IVUS catheter path derived from the biplane angiography. The IVUSAngio Tool suite integrates all the intermediate processing and computational steps and provides a user-friendly interface. It also offers additional functionality, such as automatic selection of the end-diastolic IVUS images, semi-automatic and automatic IVUS segmentation, vascular morphometric measurements, graphical visualization of the 3D model and export in a format compatible with other computer-aided design applications. Our software was applied and validated in 31 human coronary arteries yielding quite promising results. Collectively, the use of IVUSAngio Tool significantly reduces the total processing time for 3D coronary reconstruction. IVUSAngio Tool is distributed as free software, publicly available to download and use.     

一种基于u/v控制线实现股骨三维重建的曲面造型方法

本文阐述了一种基于u/v控制线实现股骨三维重建的曲面造型方法，该方法通过对u方向和v方向控制线的处理，来实现股骨三维重建的曲面造型，实验证明利用该方法，可以得到较满意的曲面造型结果，实现了股骨的三维重建。     

Localization of motoneurons that extend axons through the ventral rami of cervical nerves to innervate the trapezius muscle: a study using fluorescent dyes and 3D reconstruction method

It has been suggested that in addition to motor axons, which extend directly into the spinal accessory nerve (SAN), ventral rami-associated motor fibers of cervical nerves also innervate the trapezius muscle. Using fluorescent dye labeling and 3D reconstruction in adult rats, this study clarifies the localization of motoneurons, which extend axons either directly through the SAN or through the ventral rami of cervical nerves to innervate the trapezius. DiI or DiI and DiO were used to label the ventral rami of cervical nerves entering the SAN, as well as branches of the SAN. We show that motoneurons whose axons pass through the ventral rami of cervical nerves and then enter the SAN, and those extending axons directly through the SAN are distributed within the same area. The neurons that extend axons through the SAN had a greater diameter than those axons that pass through the cervical nerves en route to the trapezius muscle. In addition, the axons that ultimately extend through the SAN exit the spinal cord dorsolaterally, while those that pass through the cervical nerves extend out the spinal cord through the ventral roots. We presume that the neurons that extend axons through the SAN are mainly alpha-motoneurons and that those projecting axons through the cervical nerves to the trapezius are mainly gamma-motoneurons. Taken together, these results could explain why patients in whom the SAN was used to treat brachial plexus injury retain some control of the trapezius muscle.     

Microvascularization of the human central and peripheral nervous system: A new microcomputed tomography method

1. Download : Download high-res image (148KB)
2. Download : Download full-size image

     

Consistency of unitary shapes in dual lead recordings from myelinated fibres in human peripheral nerves: evidence for extracellular single-unit recordings in microneurography

Percutaneous microneurography is a powerful technique allowing studies of activity in single nerve fibres of conscious humans. However, the mechanisms by which single-unit recordings are achieved with this technique are not fully understood. To further elucidate these mechanisms, dual-lead recordings, using a modified concentric needle electrode with two separate recording surfaces at the tip, were performed in normal subjects. Sixty-two single units supplied by large myelinated afferents were studied. The majority (90%) of the units were recorded simultaneously on both surfaces but with different action potential amplitudes. Four types of unitary waveforms were encountered. The potentials recorded on the two channels were of the same type, although occasionally some details differed. Parallel waveform changes of the same units occurred simultaneously on the two surfaces. A displacement of a single fibre from one recording surface to the other with or without concomitant waveform-type transitions was observed when the electrode was slightly repositioned intraneurally. The results provided direct evidence to confirm that concentric needle electrodes record single-unit activity extracellularly from myelinated nerve fibres, probably at or close to a node of Ranvier. All the types of action potentials encountered with conventional tungsten electrodes were also identified in dual-channel recordings with concentric electrodes, which casts doubt on the previous explanation that single-unit activity recorded with tungsten electrodes is derived from intracellular sources. Some biological and technical aspects of the findings are discussed, especially concerning the applicability of in vivo measurements of the time course of the action potentials in humans and ways to improve microneurography towards multichannel recordings. Received: 8 September 1997 / Accepted: 15 December 1997     

A versatile 3D reconstruction system of the spine and pelvis for clinical assessment of spinal deformities

This paper presents a three-dimensional (3D) reconstruction system of the human spine for the routine evaluation of musculoskeletal pathologies like idiopathic scoliosis. The main objective of this 3D reconstruction system is to offer a versatile and robust tool for the 3D analysis of spines in any healthcare centre with standard clinical setup using standard uncalibrated radiographic images. The novel system uses a self-calibration algorithm and a weak-perspective method to reconstruct the 3D coordinates of anatomical landmarks from bi-planar radiographic images of a patient’s trunk. Additionally, a small planar object of known dimensions is proposed to warrant an accurately scaled model of the spine. In order to assess the validity of the 3D reconstructions yielded by the proposed system, a clinical study using 60 pairs of digitized X-rays of adolescents was conducted. The subject cohort in the study group was composed of 51 scoliotic and 9 non-scoliotic patients, with an average Cobb angle on the frontal plane of 25°. For each case, a 3D reconstruction of the spine and pelvis was obtained with the previous system used at our hospital (which requires a positioning apparatus and a calibration jacket), and with the proposed method. Results show that 3D reconstructions obtained with the new system using uncalibrated X-ray images yield geometrically accurate models with insignificant differences for 2D and 3D clinical indexes commonly used in the evaluation of spinal deformities. This demonstrates the system to be a viable and accurate tool for clinical studies and biomechanical analysis purposes, with the added advantage of versatility to any clinical setup for routine follow-ups and surgical planning. This paper was supported in part by the National Sciences and Engineering Research Council of Canada (NSERC) and the Fonds Quebecois de la Recherche sur la Nature et les Technologies (FQRNT).     

A 3D kinematic estimation of knee prosthesis using X-ray projection images: clinical assessment of the improved algorithm for fluoroscopy images

In this paper, we propose three ideas to improve a kinematic estimation algorithm for total knee arthroplasty. The first is a two-step estimation algorithm that improves estimation accuracy by excluding certain assumptions needed for the pattern matching algorithm reported by Banks and Hodge. The second is incorporating a 3D geometric articulation model into the algorithm to improve estimation accuracy substantially for the depth translation, and to introduce contact points’ trajectories between the articular surfaces. The third is an algorithm to process estimation even when the silhouettes of two components overlap. To assess our algorithm’s potential for clinical application, we carried out two experiments. First, we used a robot to position the prosthesis. Estimation accuracy was checked by comparing input data to the robot with the estimates from X-ray photographs. Incorporating our articulation model remarkably reduced the error in the depth translation. Next, we performed a clinical assessment by applying the algorithm and articulation model to fluoroscopy images of a patient who had recently had TKA.     

Multislice CT angiography in abdominal aortic aneurysm treated with endovascular stent grafts: evaluation of 2D and 3D visualisations

Endovascular repair of abdominal aortic aneurysms has been introduced into the clinical practice for more than a decade and has been confirmed to be an effective alternative to conventional open surgery, especially in patients with co-morbid medical conditions. Helical CT angiography is the preferred imaging method in the follow-up of endovascular repair. Recent introduction of multislice CT scanners has augmented its diagnostic role in this area. Diagnostic value of multlislice CT has been complemented by a series of 3D post-processings, which assist vascular surgeons in accurately assessing the effect of endovascular repair by providing additional information when compared to conventional 2D axial images. These reconstructions include multiplanar reformation, curved multiplanar reconstruction, shaded surface display, maximum intensity projection, volume rendering and virtual endoscopy. This article aims to demonstrate the generation of these 2D/3D reconstructions based on multislice CT data acquired from a group of patients with abdominal aortic aneurysm following endovascular repair. A brief introduction of generating each reconstruction was provided; potential clinical applications of each reconstruction were briefly discussed. Images were presented in a dynamic format with the aim of allowing the reader to easily understand the post-processing of these reconstructions.     

A computed tomography and magnetic resonance imaging study of the variations of the sciatic nerve branches of the pediatric knee: Implications for peripheral nerve blockade

In pediatric patients, peripheral anesthetic blockade of the sciatic nerve is useful for postoperative pain relief after orthopedic procedures of the lower limb. The identification of the sciatic and its branches relative to the surrounding anatomical structures of the knee significantly contribute to the successful nerve blockade. However, reports have demonstrated complications arising from the inadequate and inconsistent understanding of the anatomy of the nerves in the knee region. In addition, the paucity of information regarding the location of the nerves of the knee throughout the pediatric development needs to be addressed in order for peripheral nerve blockades to become more readily used among the pediatric population. This clinical imaging study therefore aimed to document and analyze the relative morphometric relationship of the nerves in the knee region throughout the different stages of pediatric development. The locations of the sciatic, tibial, and common fibular nerves were measured in relation to relevant anatomical structures in 142 pediatric magnetic resonance imaging and computed tomography. A strong correlation was found between age and the distance from the common fibular nerve to the semitendinosus muscle, and thus can be used to predict stature. Significant differences were found between males and females, specifically after puberty, suggesting that sexual dimorphism emerges more distinctively at puberty. In order to predict the location of the nerves at the popliteal fossa, a regression formula using the patient's age and bicondylar width was created. These finding may have significant implications for popliteal approach of the sciatic nerve blockade in pediatric patients. Clin. Anat. 32:836–850, 2019. © 2019 Wiley Periodicals, Inc.     

虚拟中国人女性一号松质骨图像数据的配准与三维重建

目的：研究从虚拟人体数据集中松质骨连续切片图像的分割、配准、及三维重建的技术方法。方法：利用现有的虚拟中国人女性一号数据集中腰椎和股骨部分解剖连续切片数据集，用基于外置标记点和分割—计数法两种方法进行参数计算，依参数对图像进行刚体变换完成配准，将配准后的切片图像输入二维图像处理软件进行分割，提取感兴趣区域后输入三维重建软件进行三维重建。结果：重建后的松质骨三维立体图像呈均匀、致密的立体网状结构，骨小梁连接清晰可见。结论：利用现有软件及技术可重建虚拟人体的精细结构。