Skip-cycleGAN:一种果园苹果异源图像配准模型

针对有监督的配准模型的性能受限于给定的标签以及循环一致性生成对抗网络训练不稳定,收敛速度较慢,易过拟合,对复杂场景的图像处理效果不佳的问题,基于循环一致性生成对抗网络从3个方面(生成器、鉴别器和损失函数)进行改进,提出一种无监督的异源图像配准模型。生成网络的下采样与上采样之间引入带有特征转换残差层的跳跃连接,可以确保梯度的有效传递,减少前向与反向传播过程中信息损失,实现低级特征和高级特征的结合,从而缓解梯度消失和梯度爆炸,促进神经网络的收敛,有助于网络学习更多的上下文信息。在一个自建果园苹果数据集和两个公共数据集上对模型进行评估,实验得出在改进后的生成器基础上,对于形变比较大的数据集选取70×70 PatchGAN鉴别器更合适,对于形变比较小的数据集选取PixelGAN鉴别器更合适。与8个经典算法进行对比,用6个性能指标进行评估,实验结果表明该模型在异源果园苹果数据集上的综合表现优于对比算法。未来将提升模型对异源图像亮度和对比度的鲁棒性,并进行轻量化模型的工作。

基于SKIP宏程序编程CNC无线测头在线智能检测系统设计

针对数控铣削加工精密零件尺寸精度测量时,因离线测量易产生测量误差、效率低、成本高等问题,通过赋值相关检测路径及补偿的参数,编写零件在线检测及误差补偿宏程序,开发了基于三菱M70数控系统跳过指令SKIP在线检测与误差补偿宏程序控制系统。应用结果表明:该系统在线检测重复定位精度能达到0.001~0.003 mm,能适用于精密零件数控铣削加工,具有一定的推广使用价值。

Current status of nerve infiltration by skip metastases in colorectal cancer

Colorectal cancer(CRC)is a prevalent malignant tumor,with the global new cases reaching 1.9316 million and deaths reaching 935,200 in 2022.In China,therewere 555,500 new cases of CRC,with an age-standardized incidence rate of 24.07 per 10million,and 286,200 deaths.China accounts for approximately 30%of new cases and deaths from CRC worldwide,with East Asia accounting for over 75%.Initially,CRC presents as local tumor growth,but it has the potential to spread to other body parts over time.Perineural infiltration(PNI)is a relatively less discussed route of diffusion,yet it plays a crucial role in the progression and prognosis of CRC.PNI often occurs alongside local lymph nodes and distant metastases,posing challenges for treatment and management.Clinical symptoms,radiographic findings,and histopathological examination can be used to diagnose PNI with skipmetastasis.Symptoms commonly include local pain,paresthesia,andmotor impairment.Imaging helps identify the mass’s location and relationship to nerves,whereas histopathological examination confirms the diagnosis.Treatment of PNI skipmetastases is similar to other CRC metastases,including surgical resection,chemotherapy,radiotherapy,and targeted therapy.Surgical resection is the primary therapeutic approach,but the wider range of metastasis in PNI skip transfer may limit its feasibility.In cases where surgical resection is not possible,chemotherapy,radiotherapy,and targeted therapy are used to control tumor metastasis.In conclusion,PNI skip metastases increase the risk of poor prognosis for CRC,requiring a comprehensive approach with multiple treatments to prevent disease progression.Early detection and treatment are vital to improving prognosis.

基于LSTNet-Skip的综合能源系统多元负荷超短期预测

随着用户侧用能需求多元化的发展,多元负荷的超短期预测对于动态的大型综合能源系统的规划和优化至关重要。为此,该文提出一种基于长期和短期时间序列网络的多元负荷超短期预测模型。首先采用卷积神经网络来提取多元负荷之间的局部依赖关系,然后使用长短期记忆网络捕获负荷序列的长期依赖关系,使用具有循环跳过结构的长短期记忆网络充分学习负荷序列的超长期重复模式,最后采用自回归层和全连接层进行组合预测。使用平均绝对百分比误差和均方根误差作为评价指标,利用美国亚利桑那州立大学坦佩校区综合能源系统数据集进行验证,并与3种负荷预测方法比较。实验结果表明,提出的预测模型均优于其他方法且有较高的预测精度。

冉冉升起的肺癌靶向治疗新星:MET exon14-skipping

2014年7月,美国癌症基因研究组（The CancerGenome Atlas,TCGA）在《Nature》上发布了230例可切除肺腺癌的包含DNA、mRNA、micro RNA水平的综合基因谱分析结果。通过对mRNA和DNA高通量测序结果及序列比对分析,发现约4%（10/230）肺腺癌因MET基因DNA水平第14外显子剪接区域（splice-sites）的突变导致MET第14外显子在mRNA水平出现部分或完全跳跃缺失（exon14．skipping）。

Pulse Skipping Modulated Buck Converter - Modeling and Simulation

Modeling and simulation results of a pulse skipping modulated buck converter for applications involving a source with widely varying voltage conditions with loads requiring constant voltage from full load down to no load is presented. The pulses applied to the switch are blocked or released on output voltage crossing a predetermined value. The regulator worked satisfactorily over a wide input voltage range with good transient response but with higher ripple content. Input current spectrum indicates a good EMI performance with crowding of components at audio frequency range for the se-lected switching frequency.

Quantization Skipping Method for H.264/AVC Video Coding

This paper presents a quantization skipping method for H. 264/AVC video coding standard. In order to reduce the conputational-cost of quantization process coming from integer discrete cosine transform of H. 264/AVC, a quantization skipping condition is derived by the analysis of integer transform and quantization procedures. The expeerimental results show that the proposed algorithm has the capability to reduce the computional cost about 10% -2.5%.

Skipping Metastasis to Mediastinal Lymph Nodes in Non-small Cell Lung Cancer: A Clinical Study on the Reasonable Extent of Dissection

Objective To elucidate the characteristics and metastastic pattern of skipping mediastinal lymph node metastasis (skipping N2) in non-small cell lung cancer (NSCLC), and investigate reasonable extent of lymph node dissection. Methods From 1990 to 1998, lobectomy combined with systematic mediastinal lymph node dissection was performed in 109 patients with NSCLC. A retrospective study was carried out to elucidate the characteristics of skipping N2 disease and to compare the difference between skipping N2 and non-skipping N2 diseases. Results Twenty-one patients (19%) had skipping N2 diseases. Of the skipping N2 group, 18 cases (86%) were adenocarcinoma. Skipping N2 disease was more common in T1 and T2 group than that in T3 and T4 group (P<0.01). All skipping N2 diseases only involved one nodal station, and most of them were regional mediastinal nodal metastasis. Skipping N2 from upper lobe tumors mainly involved superior tracheobronchial or subaortic lymph nodes, and skipping N2 from lower lobe tumors involved subcarinal lymph nodes. Conclusion Skipping N2 disease presents certain clinical characteristics and metastastic pattern, and mediastinal nodal dissection might be modified according to the pattern.

Analysis of Current Situation of Fancy Rope Skipping Curriculum in Regular Institutions of Higher Learning in Sichuan Province

This paper adopts the research methods such as literature material method, questionnaire investigation method, expert interview method and mathematical statistics to investigate and study the conditions of fancy rope skipping curriculum in 22 universities in Sichuan. The research shows that the university teachers highly recognize the fancy rope skipping curriculum; construction of teaching resources should be strengthened; cultivation of the professional teaching staffs should be strengthened; in terms of teaching materials, more premium teaching materials with high pertinence and professionalism are needed; the promotion and implementation of the curriculum system of fancy rope skipping should be boosted in different steps and stages; cultivation of the professional teachers should be enlarged; related scientific research of the curriculum system of fancy rope skipping should be strengthened.

Aurora A Kinase Plays a Key Role in Mitosis Skip during Senescence Induced by Ionizing Radiation

Objective To investigate the fate and underlying mechanisms of G2 phase arrest in cancer cells elicited by ionizing radiation(IR).Methods Human melanoma A375 and 92-1 cells were treated with X-rays radiation or Aurora A inhibitor MLN8237(MLN)and/or p21 depletion by small interfering RNA(si RNA).Cell cycle distribution was determined using flow cytometry and a fluorescent ubiquitin-based cell cycle indicator(FUCCI)system combined with histone H3 phosphorylation at Ser10(p S10 H3)detection.Senescence was assessed using senescence-associated-β-galactosidase(SA-β-Gal),Ki67,andγH2AX staining.Protein expression levels were determined using western blotting.Results Tumor cells suffered severe DNA damage and underwent G2 arrest after IR treatment.The damaged cells did not successfully enter M phase nor were they stably blocked at G2 phase but underwent mitotic skipping and entered G1 phase as tetraploid cells,ultimately leading to senescence in G1.During this process,the p53/p21 pathway is hyperactivated.Accompanying p21 accumulation,Aurora A kinase levels declined sharply.MLN treatment confirmed that Aurora A kinase activity is essential for mitosis skipping and senescence induction.Conclusion Persistent p21 activation during IR-induced G2 phase blockade drives Aurora A kinase degradation,leading to senescence via mitotic skipping.

基于UNet3+生成对抗网络的视频异常检测

为解决传统视频异常检测方法在不同场景下多尺度特征提取不完全的问题,提出两种方法:一种是用于简单场景的基于UNet3+的生成对抗网络方法(简称U3P^(2)),另一种是用于复杂场景的基于UNet++的生成对抗网络方法(简称UP^(3))。两种方法分别对连续输入的视频帧生成预测,引入多种损失函数和光流模型学习其外观与运动信息,通过计算AUC进行性能评估。U3P^(2)方法以6.3 M参数量在Ped2数据集的AUC提升约0.6%,而UP^(3)方法在Avenue数据集的AUC提升约0.8%,验证其能够有效应对不同场景下的异常检测任务。

基于多目标优化的跨介质航行体水面滑跳初始运动参数设计

基于耦合欧拉-拉格朗日方法,开展了初始运动参数对带滑板跨介质航行体近水面滑跳过程的影响规律研究。采用参数回归方法分别建立最大浸深、最大过载和俯仰角变化3个变量随航行体初始速度、入水角、滑板倾角变化的关系式,得到了跨介质航行体稳定滑跳的参数包络,在入水角2°~5°、滑板倾角10°~25°、速度60~80 m/s范围之间,且在范围内各参数之间会相互制约。在此基础上,基于多目标优化算法获得了最适宜航行体稳定滑跳的初始参数为入水速度60.1 m/s、滑板倾角18.8°、入水角2°,在此参数范围内,跨介质航行体的水面滑跳过程可以在不超过航行体过载要求的条件下获得相对最小的入水浸深和俯仰角变化。

加入跳跃连接的深度嵌入K-means聚类

现有的深度聚类算法大多采用对称的自编码器来提取高维数据的低维特征,但随着自编码器训练次数的不断增加,数据的低维特征空间在一定程度上发生了扭曲,这样得到的数据低维特征空间无法反映原始数据空间中潜在的聚类结构信息.为了解决上述问题,本文提出了一种新的深度嵌入K-means算法(SDEKC).首先,在低维特征提取阶段,在对称的卷积自编码器中相对应的编码器与解码器之间以一定的权重加入两个跳跃连接,以减弱解码器对编码器的编码要求同时突出卷积自编码器的编码能力,这样可以更好地保留原始数据空间中蕴含的聚类结构信息;其次,在聚类阶段,通过一个标准正交变换矩阵将低维数据空间转换为一个新的揭示聚类结构信息的空间;最后,本文以端到端的方式采用贪婪算法迭代优化数据的低维表示及其聚类,在6个真实数据集上验证了本文提出新算法的有效性.

一期后路手术治疗跳跃性胸腰椎布病脊柱炎

目的探讨一期后路手术治疗跳跃性胸腰椎布病脊柱炎的临床疗效。方法回顾性分析2016年9月~2019年10月收治的12例病灶呈跳跃性分布的胸腰椎布病脊柱炎患者。比较手术前后疼痛视觉模拟评分(visual analogue scale,VAS)、腰痛日本骨科协会评分(Japanese orthopedic association score,JOA)、Oswestry功能障碍指数(Oswestry disability index,ODI)、美国脊髓损伤协会(American spinal injury association,ASIA)分级及红细胞沉降率(erythrocyte sedimentation rate,ESR)、C-反应蛋白(C-reactive protein,CRP)。结果所有患者术后2周及术后1年,VAS、JOA、ODI与术前比较均有统计学差异(均为P<0.001);术后1年的VAS、JOA、ODI与术后2周比较,也均有统计学差异(均为P<0.001)。术后2周,所有患者ESR、CRP较术前均有所下降,但与术前比较均无统计学差异(P=0.430,P=0.395)。但术后1年的ESR、CRP与术前及术后2周比较,均有统计学差异(均为P<0.001)。9例脊髓或神经压迫症状患者的ASIA分级由术前B级1例、C级2例、D级6例,恢复至术后1年时D级1例、E级8例,具有统计学差异(P=0.011)。结论一期后路病灶清除、减压、内固定、椎间植骨融合手术治疗跳跃性胸腰椎布病脊柱炎,手术安全、可行。

结构参数对跨介质飞行器近水面滑跳运动特性影响分析

跨介质飞行器可借助其特殊的滑板结构实现在不同介质中反复运动,为研究滑板参数对飞行器滑跳性能的影响,文中采用标准k-ε湍流模型和重叠网格技术,对具有不同滑板固定角以及滑板构型的飞行器近水面滑跳运动过程进行了数值计算研究.研究发现,当滑板固定角在10°左右时,该飞行器的近水面滑跳水动力性能最佳,稳定滑跳次数最多.而对于不同构型的滑板,矩形平滑板的砰击载荷最大,有利于飞行器的滑跳,但易对滑板造成损坏,而V型曲面滑板不仅具有优良的水动性能,同时兼具载荷适中的优点,最符合实际工程中飞行器滑板的设计要求.

融合编码器多尺度特征的RGB-D图像语义分割

针对语义分割任务中,室内场景中目标物体尺寸变化较大的问题,在ACFNet的基础上,提出融合编码器多尺度特征的RGB-D语义分割网络。首先,为有效利用网络提取的多尺度特征,提出结合池化操作的多尺度特征融合模块(PMFM),选择编码器不同阶段RGB和深度特征的融合特征作为该模块的输入;其次,设计改进的跳跃连接模块(ISCM),使用下一层级包含更多语义信息的特征图辅助修正当前层级的特征图,再经跳跃连接以拼接的方式传输到解码器对应阶段。将提出的网络模型应用到NYUD V2和SUN RGB-D数据集上,平均交并比分别达到了52.6%和48.8%。通过这两项改进,实验结果表明,上述方法达到了较高的分割准确率,优于对比的语义分割方法。

赛沃替尼治疗MET基因突变非小细胞肺癌的药物浓度与疗效和不良反应的关系

目的:探讨药物浓度与疗效和不良反应之间的关系,了解赛沃替尼治疗MET14外显子跳跃突变晚期非小细胞肺癌患者的疗效和安全性。方法:回顾性分析2018年4月至2019年9月北京大学肿瘤医院收治的携带MET14外显子跳跃突变,并且接受赛沃替尼治疗的16例局部晚期或转移性非小细胞肺癌患者的临床资料及其血浆药物浓度,评估赛沃替尼治疗的疗效和安全性,分析药物浓度与疗效及不良反应的相关性。随访患者的无进展生存期(progression-free survival,PFS)和总生存期(overall survival,OS),采用Kaplan-Meier法计算中位PFS和OS。结果:16例患者中男性9例(56.3%),年龄范围51~84岁。肺癌病理类型包括:腺癌11例,肉瘤样癌5例,鳞癌1例。赛沃替尼的客观缓解率为50.0%(8/16),疾病控制率达到87.5%(14/16)。评效达到部分缓解患者和未达到者相比,赛沃替尼的药物谷浓度分别为77.2 ng/mL和146.7 ng/mL(P=0.06)。治疗中出现过2级以上不良反应的患者和仅出现过1级不良反应者,其药物谷浓度分别为116.5 ng/mL和93.2 ng/mL(P=0.59)。所有患者的中位PFS为8.5个月,中位OS为14.1个月。结论:赛沃替尼治疗MET基因突变非小细胞肺癌的效果较好,不良反应可耐受。常规用药剂量下,赛沃替尼的血浆药物谷浓度与疗效可能存在一定负相关,但药物浓度与不良反应的程度无明显相关性。

一种基于伪谱法的助推-补能跳跃滑翔轨迹优化方法

针对大气层边缘补能-跳跃滑翔运动,建立了一种助推-补能跳跃滑翔全程空间运动模型以及补能滑翔段无量纲运动方程。提出了一种基于Gauss伪谱法的动态分段串行优化策略,可对助推段俯仰程序角、补能-跳跃滑翔段飞行攻角、倾侧角以及补能时刻等轨迹控制变量进行求解,实现含连续/离散控制变量的轨迹优化一体化设计。分别对助推-补能跳跃滑翔和助推-拟平衡滑翔轨迹优化开展仿真,验证了针对补能-跳跃滑翔轨迹的分段串行优化策略的有效性,同时分析了相同助推规模和飞行器质量条件下,补能跳跃轨迹相较拟平衡滑翔轨迹在增大飞行时间、航程和机动过载能力上更具优势。

金字塔渐进融合低照度图像增强网络

针对现有低照度图像增强网络对不同尺度特征信息存在感知与表达能力不足的问题,提出金字塔渐进融合低照度图像增强网络模型。网络对图像进行多次下采样操作以组成特征金字塔,通过在特征金字塔的三个不同分支上加入跳跃连接,将不同尺度的特征图进行相互融合。通过精细恢复模块进一步提取精炼信息,将特征图恢复到正常的光照图像。结果表明,该网络模型不但能有效地提升低照度图像的整体亮度,而且能很好地保持图像中的细节信息和清晰的物体边缘轮廓,同时能够有效地抑制图像中的暗部噪声,使增强后的图像整体画面真实自然。

基于条件生成对抗网络的大视角单图像人脸纹理重建

针对当前的人脸重建方法、尤其是纹理重建部分,应用于自遮挡或物体遮挡区域表现不佳,使得人脸纹理重建后的结果不真实的问题,本文提出了基于条件生成对抗网络的大视角单图像人脸纹理重建的方法,实现人脸图像的纹理补全。补全网络基于改进的条件生成对抗网络,包括编码器和解码器的粗层之间的跳跃连接来保存高频细节;每个卷积层的输出上叠加了高斯噪声映射;将U-V纹理映射与其翻转版本共同连接输入的方法来提高纹理重建的质量以及真实性。使用Multi-PIE数据集与CFP数据集进行评估,整体网络能够实现更高的纹理重建精度,尤其在±90°图像重建上,能获得更为完整的纹理图像。