人工智能与专家Greulich-Pyle图谱法骨龄评估的一致性比较

背景：人工智能骨龄评估已成为研究热点，国内外研究表明人工智能技术在医学影像领域发展迅速，为更准确快速的对骨龄进行评估提供了可能。目的：探讨人工智能Greulich-Pyle (智能GP)图谱骨龄评估系统与人工专家Greulich-Pyle(专家GP)图谱评估结果的一致性。方法：骨龄片采样对象为6-15岁儿童青少年，其中男672人，女650人。同一张骨龄片分别用智能GP与专家GP读出骨龄值。智能GP读片结果的准确性采用差值绝对值反映；骨龄结果一致性采用Pearson相关、Bland-Altamn分布图；评估一致性采用Kappa检验。结果与结论：(1)智能GP骨龄与专家GP骨龄差值95%置信区间男女总体分别为0.39岁(0.37-0.41岁)，0.32岁(0.29-0.34岁)，智能GP骨龄与专家GP骨龄差值Bland-Altamn偏差男女分别为(-0.096±0.482)岁，(0.014±0.415)岁；(2)智能GP骨龄与专家GP骨龄男女相关结果分别为r=0.991,r=0.992,P<0.0001；男女所有年龄段差值中位数均在0.5岁之内；Kappa检验值除男子9岁年龄段，其他...     

腰椎智能医学图像建模系统的构建及精度检验

目的　构建腰椎智能医学图像建模系统，替代分析人员完成有限元建模过程中对腰椎活动度的繁复计算工作。 方法　利用python语言构建一个附和系统，包括录制程序与回放程序。通过Geomagic的偏差分析功能评估智能建模与人工建模二者的偏差分布，并分别构建腰椎的非线性有限元模型，验证模型对有限元分析结果的影响。 结果　本课题建立智能腰椎复位系统，成功构建腰椎体的三维模型。与人工建模进行偏差分析，超过98%的区域0偏差。有限元活动度存在些许差异，经配对t检验，差异无统计学意义（P=0.2）。 结论　利用智能建模系统构建的三维腰椎模型，精确度较高且对有限元分析结果未产生影响。     

基于深度主动学习的白带白细胞智能检测方法研究

白带显微图像中白细胞的数量可以提示阴道炎症的严重程度。目前对白带中白细胞的检测主要依靠医学专家们的人工镜检,这种人工检查耗时、昂贵且容易出错。近年来,有研究提出基于深度学习技术对白带白细胞实现智能检测,但是这类方法通常需要人工标注大量的样本作为训练集,标注代价高。因此,本研究提出运用深度主动学习算法来实现对白带显微图像中白细胞的智能检测。在主动学习框架下,首先以少量的标注样本作为基础训练集,采用更快的卷积神经网络(Faster R-CNN)训练检测模型,再自动挑选最有价值的样本进行人工标注,从而迭代更新训练集和相应的检测模型,使模型的性能不断提高。实验结果表明,深度主动学习技术能在较少的人工标注样本下获得较高的检测精度,对白细胞检测的平均精度达到了90.6%,可以满足临床常规检查要求。     

基于达芬奇技术的菌类发酵控制系统的研究

目的在于设计基于达芬奇技术的嵌入式系统,以实现菌类图像分割以及菌体自动计数和面积测算功能,根据这些菌体特征参数输出控制信号,优化菌类的生长条件,提高目标产物率.在硬件设计方面,充分利用基于TMS320DM6446芯片的达芬奇开发板强大的视频图像处理能力和丰富的外设接口,构造体积小、处理速度快、控制灵活的菌类发酵控制系统硬件平台.在软件实现方面,重点研究达芬奇软件框架Codec Engine和水平集分割算法,并基于Codec Engine的编程框架,实现了菌类图像水平集分割以及菌体计数和面积测算功能.实验数据统计分析结果表明,该系统的菌体自动计数结果与人工计数结果基本一致,并且实现速度快;通过面积测算,可建立菌体面积与干重的线性关系,从而根据菌体数目和面积快速判断菌类生长状况.实践证明,所设计的达芬奇菌类发酵控制系统具有良好的准确性、高效性和稳定性.     

基于物联网和可穿戴技术的智能监护系统研发及其应用模式探索研究

可穿戴技术是一种低生理、心理负荷的监测技术,具有长时间连续监测的优点,代表了未来监护技术的一个发展方向。本文以穿戴式生理参数监测技术为基础,结合物联网和人工智能等技术,研发了基于物联网可穿戴技术的智能监护系统,包括可穿戴硬件、病区物联网平台、连续生理数据分析算法与软件三大部分。基于该系统,经过大量临床实践探索了连续生理数据的临床应用价值,给出了实时监护、病情评估、预测预警和康复训练四大价值方向;依托真实临床应用环境,探索了可穿戴技术在普通病房监护、心肺康复、院内-院外一体化监测等领域的应用模式。研究结果表明,本监护系统能够有效用于院内患者监护、心肺功能评估与训练以及院外患者管理。     

脑卒中患者上肢功能智能评估系统研究进展EI北大核心CSCD

目前临床上常采用量表方法评估脑卒中患者的上肢功能,但这种方法存在耗时长、评估结果一致性差、需康复医师参与度高等问题。为克服量表方法的短板,结合传感器和机器学习算法的上肢功能智能评估系统成为了近年来的研究热点之一。本文首先对常用的临床上肢功能评估方法做了分析总结,随后对近年来智能评估系统的研究进行了综述,重点对智能评估系统中数据采集和数据处理部分使用的技术及其优缺点进行了分析总结,最后对目前智能评估系统面临的挑战和未来的发展方向展开讨论,以期为相关领域的研究学者提供有价值的参考信息。     

脑卒中患者上肢功能智能评估系统研究进展

目前临床上常采用量表方法评估脑卒中患者的上肢功能,但这种方法存在耗时长、评估结果一致性差、需康复医师参与度高等问题。为克服量表方法的短板,结合传感器和机器学习算法的上肢功能智能评估系统成为了近年来的研究热点之一。本文首先对常用的临床上肢功能评估方法做了分析总结,随后对近年来智能评估系统的研究进行了综述,重点对智能评估系统中数据采集和数据处理部分使用的技术及其优缺点进行了分析总结,最后对目前智能评估系统面临的挑战和未来的发展方向展开讨论,以期为相关领域的研究学者提供有价值的参考信息。     

膝骨性关节炎运动质量评估系统设计

为膝骨性关节炎患者提供一种运动康复训练的监护系统,患者可以通过监护系统了解自身运动规范程度并作适当的调整.设计了一种基于ZigBee无线通信技术的人体下肢运动质量评估系统,以评估膝骨性关节炎运动理疗法的动作规范性.该系统将装有微型加速度传感器的ZigBee模块穿戴在人体的下肢,获取运动时的三维加速度信号,将加速度信号经过Haar小波变换后,采用粒子群算法提取小波特征值,将提取的特征向量应用神经网络分类器对动作质量进行识别评估.通过对20名年龄在24～30周岁的健康男性直腿抬高训练的动作质量评估测试,系统对规范抬腿、抬腿过高、保持时间太短和非平行抬腿这4类训练取真率的均值和标准差分别为(89.1±2.0)％、(93.4±1.7)％、(89.5±2.3)％、(90.1±1.8)％.实验结果表明,本系统能有效地识别训练过程中的不规范动作,较好地实现了对直腿抬高训练的运动质量监测与评估,满足健康监护系统的应用需求.     

消化道内窥镜图像异常的人工智能诊断方法研究进展

随着深度学习的出现,图像处理不再局限于人工提取特征,转而对图像进行端到端的预测,实现了人工智能在图像处理领域的又一历史性飞越。作为人工智能医疗领域的热点应用,内镜图像异常检测能够准确快速地筛选整个消化道的异常,为医生提供诊断帮助。该文围绕消化道图像最为常见的息肉、出血、溃疡等异常,对其智能诊断方法展开研究,并探讨机器学习在消化内镜异常检测的应用现状,最后展望了未来消化道内窥镜病灶智能诊断的研究方向。     

基于视频目标检测的中药饮片调剂预警系统

目的 构建基于视频目标检测的中药饮片调剂预警系统,降低中药饮片调剂混淆事件发生率,保障用药安全。方法 设计基于人工智能的中药饮片调剂预警系统开发方案,分为视频采集、目标检测、文字识别和预警四个模块。预警系统可以驱动监控视频实时检测药剂师中药饮片调剂内容,与处方实时比对,对漏配、错配和多配等错误调剂信息进行语音和图文信息预警。本文以小柴胡汤加减方为基础,采购3个批次中药饮片样本,采集并标注1 524张饮片图像,制作中药饮片数据集,基于Faster R-CNN算法训练目标检测模型并进行系统构建。结果 经测试Faster R-CNN模型均值平均精度(mean average precision, mAP)达到了95.10%,利用训练好的目标检测模型结合文字识别和预警算法构建预警系统,对中药饮片调剂过程进行识别,系统能对调剂错误行为进行准确自动报警。结论 该系统可实时、主动进行中药饮片调剂检测和预警,为中药饮片调剂智能化提供新思路,提高中药饮片调剂科技水平,促进中医药与人工智能相结合。     

MHRA and NHS PASA Publications Update

Abstract

Background: Reliable step counting is a critical part of locomotion research. Current counting methods can be inaccurate, time consuming, expensive or encumbering to the subject. Here, we present a camera-based optical method for automatically counting steps.

Methods: Fifteen healthy adults walked, jogged and ran on a treadmill at three different constant speeds (1.21, 2.01, 2.68?m/s) and once at varying speed (1.21–2.68 m/s) for 90?s. Subjects had visual marker affixed to their left foot while walking. Video was recorded synchronously at low- and high-resolution during trials. The step count found manually from the video was compared to an automated video analysis system using the two configurations of the optical system.

Results: Bland–Altman plots, Intra-class correlation coefficients (ICC) and relative error comparison were used for quantitative assessment of device reliability. Reliability of optical method was high (ICC ≥0.98).

Conclusions: The method produces accurate step count results for the range of speeds tested. They use customisable open-source software and off-the-shelf hardware. The method has a low cost of implementation compared to many consumer products and grants researchers access to the raw sensor data.     

Tissue engineering scheming by artificial intelligence

Tissue engineers are often confused when seeking the most effective, economical and secure scheme for tissue engineering. The aim of this study is to generate tissue engineering schemes with artificial intelligence instead of human intelligence. The experimental data of tissue engineered cartilage were integrated and standardized with a centralized database, and a scheme engine was developed using artificial intelligent methods (artificial neural networks and decision trees). The scheme engine was trained with existing cases in the database, and then was used to generate tissue engineering schemes for new experimental animals. Following the schemes generated by the artificial intelligent system, we cured 18 of the 20 experimental animals. In conclusion, artificial intelligence is a powerful method for decision making in the tissue engineering realm.     

Coping Versus Mastery Modeling Intervention to Enhance Self-efficacy for Exercise in Patients with COPD

Abstract

The literature suggests self-efficacy is a determinant of physical activity and management of Chronic Obstructive Pulmonary Disease (COPD). The purpose of this study was to (1) test the effects of two vicarious experience interventions, coping versus mastery modeling, on self-efficacy in COPD patients performing a cardiopulmonary exercise test (CPET), and (2) determine the type of self-efficacy most strongly related to physical activity in COPD patients. After a baseline assessment of self-efficacy (task, coping for exercise, coping for breathing, scheduling, and walking) and potential moderators, 120 COPD patients watched a mastery model or coping model CPET video, or received usual care verbal instructions. Then, self-efficacy was assessed, followed by a CPET, and another assessment of self-efficacy. Fitbits tracked participants’ step count the week following contact. Repeated measures MANOVAs assessed the intervention effects and multiple regressions assessed the contribution of self-efficacy subtypes to step count. All self-efficacy subtypes improved in the mastery and coping conditions, although greater improvement of self-efficacy for coping with exercise barriers was observed in the coping condition. Self-efficacy did not improve in the control condition and no moderators were identified. Self-efficacy for coping with exercise barriers was the self-efficacy subtype most strongly related to step count. This research suggests modeling is a useful intervention technique to enhance self-efficacy in COPD patients, although coping models may be more beneficial than mastery models for enhancing capability beliefs during complex tasks. Future interventions in COPD patients should target self-efficacy for coping with exercise barriers.     

Decision support for tendon tissue engineering

The aim of this study is to provide decision support with artificial intelligence for tendon tissue engineering strategies. The experimental data of tissue-engineered tendons were integrated and standardized with a centralized database, and a decision support system was developed using both artificial neural networks and decision trees. The decision support system was trained with existing cases in the database, and then was used to generate tissue engineering schemes for new experimental animals. Following the schemes generated by the artificial intelligent system, we cured 28 of the 30 experimental animals. In conclusion, artificial intelligence is a powerful method for decision support in the tendon tissue engineering realm.     

Decision support for tendon tissue engineering

The aim of this study is to provide decision support with artificial intelligence for tendon tissue engineering strategies. The experimental data of tissue-engineered tendons were integrated and standardized with a centralized database, and a decision support system was developed using both artificial neural networks and decision trees. The decision support system was trained with existing cases in the database, and then was used to generate tissue engineering schemes for new experimental animals. Following the schemes generated by the artificial intelligent system, we cured 28 of the 30 experimental animals. In conclusion, artificial intelligence is a powerful method for decision support in the tendon tissue engineering realm.     

Aerobic exercise affects sleep,psychological wellbeing and immune system parameters among subjects with chronic primary insomnia

BackgroundChronic primary insomnia is characterized by long-term difficulties in maintaining and initiating sleep, too early waking up, poor mood, fatigue, impaired concentration and poor quality of life. Exercise training is recommended to prevent and alleviate sleep disorders.ObjectiveThe aim of the study was to investigate the influence of aerobic exercise training on quality of sleep, psychological wellbeing and immune system among subjects with chronic primary insomnia.Material and methodsEighty previously sedentary subjects with chronic primary insomnia subjects enrolled in this study, their age ranged from 35–56 years. All participants were randomly assigned to supervised aerobic exercise intervention group (group A, n=40) or control group (group B, n=40). Polysomnographic recordings for sleep quality assessment, Beck Depression Inventory (BDI), Profile of Mood States(POMS), Rosenberg Self-Esteem Scale (RSES), number of CD3+, CD4+, CD8+ T cells count and CD4/CD8 ratio were measured before and at the end of the study after six months.ResultsThere was a significant increase in the total sleep duration, sleep efficiency and sleep onset latency in group(A) after six months of aerobic exercise training, while, wake time after sleep onset and rapid eye movement (REM) latency significantly reduced after six months of aerobic training compared with values obtained prior to aerobic exercise training. Also, the mean values of BDI, POMS, CD3 count, CD4 count and CD8 count decreased significantly and the mean value of RSES significantly increased in group (A) after the aerobic exercise training, while the results of the control group were not significant. Moreover, there were significant differences between both groups at the end of the study.ConclusionExercise training can be considered as a non-pharmacological modalty for modifying sleep quality, psychological wellbeing and immune system among subjects with chronic primary insomnia.     

人工智能三维术前规划在全髋关节置换术中的应用和准确性分析

目的 探讨人工智能三维规划系统在全髋关节置换术中假体型号选择的准确性。方法 选取2019年11月至2020年12月住院期间的27例（32髋）行初次全髋关节置换手术患者，其中男10例，女17例。术前完成患侧髋关节X线及CT影像资料采集，分别使用传统模板测量法和人工智能三维术前规划进行假体型号预测，通过与全髋关节置换术中实际所用型号进行对比，比较两种预测方法的吻合率。结果 人工智能三维术前规划对髋臼及股骨假体型号预测的吻合率分别为90.6%和81.3%，传统模板测量对髋臼及股骨假体型号预测的吻合率分别为56.3%和46.9%。两种预测方法吻合率比较，差异有统计学意义（P<0.05）。人工智能三维规划系统预测髋臼杯型号同实际应用型号之间存在相关性（r=0.915，P<0.001），人工智能三维规划系统预测股骨假体型号同实际应用型号之间存在相关性（r=0.941，P<0.001）。结论 人工智能三维术前规划系统较传统模板测量方法更能准确预测假体型号。     

Web-video-mining-supported workflow modeling for laparoscopic surgeries

MotivationAs quality assurance is of strong concern in advanced surgeries, intelligent surgical systems are expected to have knowledge such as the knowledge of the surgical workflow model (SWM) to support their intuitive cooperation with surgeons. For generating a robust and reliable SWM, a large amount of training data is required. However, training data collected by physically recording surgery operations is often limited and data collection is time-consuming and labor-intensive, severely influencing knowledge scalability of the surgical systems.ObjectiveThe objective of this research is to solve the knowledge scalability problem in surgical workflow modeling with a low cost and labor efficient way.MethodsA novel web-video-mining-supported surgical workflow modeling (webSWM) method is developed. A novel video quality analysis method based on topic analysis and sentiment analysis techniques is developed to select high-quality videos from abundant and noisy web videos. A statistical learning method is then used to build the workflow model based on the selected videos. To test the effectiveness of the webSWM method, 250 web videos were mined to generate a surgical workflow for the robotic cholecystectomy surgery. The generated workflow was evaluated by 4 web-retrieved videos and 4 operation-room-recorded videos, respectively.ResultsThe evaluation results (video selection consistency n-index ≥0.60; surgical workflow matching degree ≥0.84) proved the effectiveness of the webSWM method in generating robust and reliable SWM knowledge by mining web videos.ConclusionWith the webSWM method, abundant web videos were selected and a reliable SWM was modeled in a short time with low labor cost. Satisfied performances in mining web videos and learning surgery-related knowledge show that the webSWM method is promising in scaling knowledge for intelligent surgical systems.     

基于睡眠心电数据的人工智能抑郁评估系统

睡眠障碍是抑郁症的评估因素之一，可从心电数据中提取特征值用于抑郁评估，辅助临床诊断。设计开发一种基于睡眠心电数据的人工智能抑郁评估系统，其中包含心电数据的分析、心电滤波和心电R波的定位。其间并进行数图转换，生成心电RR间期图。利用TensorFlow进行图像的增强处理后，构建卷积神经网络，最终得出睡眠心电数据的人工智能抑郁评估模型。     

Movement-Based Interaction Applied to Physical Rehabilitation Therapies

BackgroundHealth care environments are continuously improving conditions, especially regarding the use of current technology. In the field of rehabilitation, the use of video games and related technology has helped to develop new rehabilitation procedures. Patients are able to work on their disabilities through new processes that are more motivating and entertaining. However, these patients are required to leave their home environment to complete their rehabilitation programs.ObjectiveThe focus of our research interests is on finding a solution to eliminate the need for patients to interrupt their daily routines to attend rehabilitation therapy. We have developed an innovative system that allows patients with a balance disorder to perform a specific rehabilitation exercise at home. Additionally, the system features an assistive tool to complement the work of physiotherapists. Medical staff are thus provided with a system that avoids the need for them to be present during the exercise in specific cases in which patients are under suitable supervision.MethodsA movement-based interaction device was used to achieve a reliable system for monitoring rehabilitation exercises performed at home. The system accurately utilizes parameters previously defined by the specialist for correct performance of the exercise. Accordingly, the system gives instructions and corrects the patient’s actions. The data generated during the session are collected for assessment by the specialist to adapt the difficulty of the exercise to the patient’s progress.ResultsThe evaluation of the system was conducted by two experts in balance disorder rehabilitation. They were required to verify the effectiveness of the system, and they also facilitated the simulation of real patient behavior. They used the system freely for a period of time and provided interesting and optimistic feedback. First, they evaluated the system as a tool for real-life rehabilitation therapy. Second, their interaction with the system allowed us to obtain important feedback needed to improve the system.ConclusionsThe system improves the rehabilitation conditions of people with balance disorder. The main contribution comes from the fact that it allows patients to carry out the rehabilitation process at home under the supervision of physiotherapists. As a result, patients avoid having to attend medical centers. Additionally, medical staff have access to an assistant, which means their presence is not required in many exercises that involve constant repetition.