临床试验数据采集与报告文件设计制作的技术规程

临床试验数据采集与报告文件的设计和制作对保证试验数据的质量与完整性至关重要。参照临床试验数据管理规范 (GCP)并结合具体实践经验制订本规程。设计规程包括与方案设计同步、文件格式、问题和提示、冗余数据、数据记录指南、标准化、版本管理与培训 ;印制规程包括审批、印制说明书以及厂商选择。     

临床试验机构研究者标准操作规程

依据SFDA《药品临床试验管理规范》．ICH GCP，WHO GCP，以及我院临床试验的实践，制订临床试验机构研究者的标准操作规程，包括研究者资格与条件，试验前的准备，受试者的招募和筛选，受试者的知情同意，方案的依从性，受试者的医疗．随机化程序和破盲，安全性报告，源文件和源数据．病例报告表，试验用药的管理，试验的终止或暂停，进展报告和总结报告，档案等。     

cyclin D1基因沉默对K562细胞化疗增敏作用的体外研究

目的 cyclin D1基因在细胞周期调控中起关键性作用.研究表明其过表达与肿瘤的发生、发展及预后密切相关;并且与肿瘤细胞对化疗药物的抗拒性有关.抑制cyclin D1蛋白表达可达到化疗增敏作用.研究目的在于利用RNA干扰(RNAi)技术体外观测其对K562细胞cyclin D1基因的沉默效应及其增强多柔比星(ADM)对K562细胞毒性的作用.方法 体外构建靶向cyclin D1基因的小发夹RNA(shRNA)表达质粒,通过壳聚糖介导转染K562细胞,Western Blotting分析检测转染前后cyclin D1蛋白表达变化,流式细胞术检测细胞周期分布及对凋亡率的影响,MTF法检测K562细胞对ADM敏感性的变化.结果 构建了靶向cyclin D1基因的shRNA表达质粒(pshRNA-419和pshRNA-575)经壳聚糖转染后,能显著抑制cyclin D1基因表达;影响K562细胞周期分布,诱导细胞凋亡.并降低ADM半数抑制浓度,提高了化疗敏感性.而设计一碱基突变的序列所构建的质粒并无上述生物学效应.结论 沉默K562细胞cyclin D1基因表达可达到有效的化疗增敏目的.     

中草药生物转化工程与中医药创新

在中医药理论和现代人体生命科学理论的基础上,从中药成分代谢研究中探析,结合系统生物学等现代学科理念,运用先进的分析检测技术和生物转化技术,尤其是体外生物转化技术,可改善中药制剂的疗效与个体差异,提出新的中医药产业工程——中草药生物转化工程,以及相应的创新中药——生物中药。重点探讨了微生物转化与酶法转化技术在创新中药研发体系中的应用前景。     

三维斑点追踪超声心动图在心尖肥厚型心肌病诊断中的应用

目的应用三维斑点追踪超声心动图（three-dimensional speckle tracking echocardiography, 3D-STE）定量测定其心尖部心肌应变及旋转角度,探讨3D-STE在心尖肥厚型心肌病（apical hypertrophic cardiomyopathy, AHCM）诊断中的应用价值。方法分别对42例AHCM患者和20例正常被检者行心脏磁共振检查及超声心动图检查。将42例AHCM患者根据其MRI测量的左心室心尖部心肌厚度进行分型,其中早期心尖肥厚型心肌病患者（P-AHCM,舒张末期心尖壁厚度≥12 mm且<15 mm）20例、典型心尖肥厚型心肌病（T-AHCM,舒张末期心尖壁厚度≥15 mm）22例。所有受检者应用彩色超声诊断仪获取二维及三维全容积动态图像,并对左心室整体三维应变峰值（left ventricle global 3D strain, LVG3DS）、左心室心尖各节段三维应变峰值（left ventricle 3D strain, LV3DS）等三维应变参数及左心室心尖部各节段旋转角度（rotation angle, Rot）、左心室整体旋转角度（global rotation angle, G-Rot）进行分析,比较三组间上述指标的差异。结果与对照组比较,P-AHCM组和T-AHCM组的LVG3DS、LV3DS均减低（P<0.01）,且T-AHCM组LV3DS减低更为显著（P<0.05）,而P-AHCM组和T-AHCM组间的LVG3DS差异无统计学意义（P>0.05）。左心室心尖部旋转角度在T-AHCM组最高,P-AHCM组次之,对照组最低（P<0.01）。结论3D-STE能够定量测定AHCM患者左室心尖部心肌应变及旋转运动的改变,常规超声心动图结合3D-STE检查能够提高AHCM的诊断准确率。     

INAUGURAL ARTICLE by a Recently Elected Academy Member:A topological solution to object segmentation and tracking

The world is composed of objects, the ground, and the sky. Visual perception of objects requires solving two fundamental challenges: 1) segmenting visual input into discrete units and 2) tracking identities of these units despite appearance changes due to object deformation, changing perspective, and dynamic occlusion. Current computer vision approaches to segmentation and tracking that approach human performance all require learning, raising the question, Can objects be segmented and tracked without learning? Here, we show that the mathematical structure of light rays reflected from environment surfaces yields a natural representation of persistent surfaces, and this surface representation provides a solution to both the segmentation and tracking problems. We describe how to generate this surface representation from continuous visual input and demonstrate that our approach can segment and invariantly track objects in cluttered synthetic video despite severe appearance changes, without requiring learning.

Through a process of perceptual organization that is still not well understood, the primate visual system transforms visual input consisting of a stream of retinal images into a percept of stable, discrete objects. This process has traditionally been broken down into two separate problems: the “segmentation problem,” which addresses how visual pixels can be grouped into distinct objects within a single image (1), and the “tracking problem,” which addresses how objects can be identified across images despite changing appearance (2).Both problems are highly challenging. Segmentation is difficult because distant pixels of different color/texture can belong to the same object, while neighboring pixels of the same color/texture can belong to different objects (Fig. 1A). Tracking is difficult because the appearance of the same object can change drastically due to object deformation, changing perspective, or dynamic occlusion (Fig. 1B). The segmentation problem has classically been tackled through intensity-, color-, and texture-based region-growing approaches relying upon properties extracted from single images (3), and more recently through deep learning approaches. The tracking problem has been approached through probabilistic dynamical modeling (4) or “tracking by detection” (5–8), with recent methods incorporating deep learning (9–15). While earlier learning approaches to segmentation and tracking were supervised (10, 16), requiring large labeled training sets, more recently unsupervised approaches have emerged (15, 17). In this paper, we explore the computational origin of the ability to segment and invariantly track objects and show that this problem can in principle be solved without learning, supervised or unsupervised.Open in a separate windowFig. 1.The challenge of object segmentation and tracking and Gibson’s proposed framework for solution. (A) The challenge of segmentation: Points 1 and 2 are nearby and have the same color but belong to different objects, while points 2 and 3 are distant and different in color, but belong to the same object. (B) The challenge of invariant tracking: The three views of the horse are very different in shape and pixel composition yet represent the same object. (C) Gibson’s ecological approach to visual perception. An array of light rays from objects in the environment is sensed at each point in the observation space (two are illustrated). Gibson asserted that transformations between these arrays contain all the information necessary to solve the segmentation and invariant tracking problems (reproduced from ref. 20).Complementing image-based approaches to segmentation and tracking, a geometry-based approach considers vision as an inverse graphics problem (18). In this framework, the visual system infers three-dimensional (3D) surfaces from images by inverting a 3D graphics model. However, because the third dimension is lost during perspective projection onto the retina, this inverse inference process is considered to be not fully constrained (19), implying that extensive learning from experience is necessary. In this paper, we show that the problem of inferring 3D surfaces from images is in fact fully constrained, if the input is in the form of a sequence of images of a scene in which either the observer or objects are moving. We demonstrate through both mathematical analysis and computational experiments that with only two natural assumptions, namely, 1) the world is composed of objects, that is, a discrete set of smooth textured surfaces with locally constant lighting, and 2) animals view the world from a moving observation point, it is possible to solve the problem of segmenting and invariantly tracking each discrete surface in the environment without requiring learning. Our computational experiments are limited to synthetic video, and we assume access to high-quality images, but as we argue below, our approach should be readily extendable to natural conditions.Our paper is essentially a mathematical translation of the “ecological approach to visual perception” developed by the psychologist J. J. Gibson (20). Gibson pointed out that the key to understanding human vision is to insert between the 3D environment and the eye a new item, the field of ambient optic arrays. The ambient optic array at one point in space consists of the 2D distribution of light rays passing the point from illuminated surfaces in the environment (Fig. 1C). Gibson pointed out that the field of ambient optic arrays is governed by a set of laws which he dubbed “ecological optics,” and these laws can explain much of visual perception: “Instead of making the nervous system carry the whole burden of explaining perception, I wish to assign part of this burden to light itself. Ecological optics is my way of doing so” (20). In the decades since Gibson proposed his ecological optics approach to vision, this important concept has attracted growing attention in the computer vision community (21, 22).We explain how Gibson’s theory can be formulated in precise mathematical terms and be implemented computationally. Mathematical analysis shows that object surface information is redundantly represented by the field of ambient optic arrays through two of its topological structures: the pseudogroup of stereo diffeomorphisms and the set of infinitesimal accretion borders. Formulated in terms of ecological optics, vision is a fully constrained, well-posed problem. Complete information for perception of objects as discrete, persistent units is contained in the visual environment itself within the field of ambient optic arrays.The main paper has three parts. In the first part, we give a broad overview of our approach. In the second part, we present the mathematical theory of ecological optics (this part heavily references SI Appendix and may be skipped without loss of comprehension of the remainder of the paper). In the third part, we show how to exploit ecological optics computationally to solve the segmentation and invariant tracking problems. In addition, SI Appendix provides a self-contained and expanded exposition of the ideas.     

A Design for the High Yield Photoneutron Source Target Station

Low energy accelerator driven neutron sources are promising candidates to obtain a neutron yield as high as 10¹⁴ n/s, which is required for a variety of applications, such as boron neutron capture therapy, neutron imaging, and neutron scattering. The methods to generate neutrons can be divided into two categories: hadron-based and photon-based methods. In order to better understand which kind of source would be the better choice for delivering a brilliant neutron beam robustly, in this paper, the underlying principles of neutron production, as well as the simulation results of neutron yield, target heat dissipation, thermal stress, and reaction byproducts concentration of these two types of neutron sources, will be elaborated on. A preliminary photoneutron target station design based on a 50 MeV/50 kW electron linear accelerator, including the optimized neutron yield, thermal hydraulic analysis, and shielding calculation, is presented as well to demonstrate the method to deliver brilliant thermal neutron beam of 1.03 × 10¹⁰ cm⁻² s⁻¹ sr⁻¹.     

Characteristics and Prognosis of Stroke in Living Donor Renal Transplant Recipients

Aims: We aimed to determine the characteristics and vascular outcomes of stroke in renal transplant (RT) recipients and compare them with those in patients on hemodialysis (HD) and those with no renal replacement therapy (RRT). Methods: In this prospective observational study, 717 patients (mean age, 70.8 years; male, 60.5%) with acute ischemic stroke within one week of onset were consecutively enrolled and followed for one year. The patients were classified into three groups: (1) living donor RT recipients (n=27); (2) patients on maintenance HD before the index stroke (n=39); and (3) those with no history of RRT (n=651). The primary outcome was a composite of major adverse cardiovascular events (MACE). Results: Diabetic nephropathy was the most common reason for RRT in both RT and HD patients. RT patients were more likely to have embolic stroke of undetermined source (33.3%) than others, whereas HD patients more often had cardioembolism (51.3%). No difference was observed in the MACE risk between the patients in RT and non-RRT groups (annual rate, 11.3% vs. 13.1%; log-rankP=0.82; hazard ratio [95% confidence interval], 0.92 [0.29-2.98]). In contrast, HD patients had a greater risk of MACE than those with no RRT (annual rate, 28.2% vs. 13.1%; log-rankP=0.019; hazard ratio [95% confidence interval], 2.24 [1.16-4.3]). Conclusions: The underlying etiologies of stroke differed in RT and HD patients. The one-year risk of MACE for stroke patients who had received an RT was lower than that for patients undergoing HD and comparable with that of patients with no RRT.     

组织非均匀性对宫颈癌近距离治疗剂量评估影响研究

目的:比较近距离治疗两种临床常用的剂量计算方法和基于CT影像的蒙特卡罗程序计算的剂量差异,探讨组织非均匀性对宫颈癌近距离治疗剂量评估的影响。方法:回顾性选取2018年1月至2020年6月在安徽省肿瘤医院接受三维近距离治疗的宫颈癌患者11例,分别采用美国医学物理师协会(AAPM)TG43号报告的纯水剂量计算方法(TG43...     

Association of left ventricular global area strain derived from resting 3D speckle-tracking echocardiography and exercise capacity in individuals undergoing treadmill exercise test

Background: Left ventricular (LV) global area strain (GAS) is a novel index derived from resting 3D speckle-tracking echocardiography (STE), and its clinical significance has rarely been studied. We examined the association of LV GAS and exercise capacity in a health check-up population.Methods: We recruited 94 symptom-free participants (52.2 ± 11.7 years, 62.8% male) without substantial structural heart disease or coronary heart diseases who were undergoing a routine health examination. All participants underwent resting echocardiography and symptom-limited treadmill exercise test according to the Bruce protocol. Four strain parameters were obtained from the analysis, namely 3D GAS (GAS3d), global longitudinal strain, global circumferential strain, and global radial strain.Results: After multivariate analysis for factors of exercise time, we observed a significant association in LV GAS3d (P < 0.001). We divided participants into preserved and impaired exercise capacity groups according to the cutoff value of 8 metabolic equivalent of tasks. LV GAS3d (OR 1.24, 95% CI 1.10-1.39, P < 0.001) was an independent predictor of impaired exercise capacity and the optimal cut-off value was -19.96% at a sensitivity of 77.8% and at a specificity of 92.1%. LV GAS3d could improve the discriminatory power of exercise capacity in individuals with early mitral filling velocity to average mitral annulus velocity ratio (E/e'') ≥ 8.Conclusions: LV GAS3d was significantly associated with exercise time and exhibited incremental predictive value on E/e'' for exercise capacity in participants undergoing treadmill exercise test.