基于图卷积的视网膜血管轮廓及高不确定度区域细化框架

针对传统卷积神经网络(convolutional neural network, CNN)受感受野大小的限制,无法直接有效地获取空间结构及全局语义等关键信息,导致宽血管边界及毛细血管区域特征提取困难,造成视网膜血管分割表现不佳的问题,提出一种基于图卷积的视网膜血管分割细化框架。该框架通过轮廓提取及不确定分析方法,选取CNN粗分割结果中潜在的误分割区域,并结合其提取的特征信息构造出合适的图数据,送入残差图卷积网络(residual graph convolutional network, Res-GCN)二次分类,得到视网膜血管细化分割结果。该框架可以作为一个即插即用模块接入任意视网膜血管分割网络的末端,具有高移植性和易用性的特点。实验分别选用U型网络(U-neural network, U-Net)及其代表性改进网络DenseU-Net和AttU-Net作为基准网络,在DRIVE、STARE和CHASEDB1数据集上进行测试,本文框架的Sp分别为98.28%、99.10%和99.04%,Pr分别为87.97%、88.87%和90.25%,证明其具有提升基准网络分割效果的细化能力。     

基于张量鲁棒主成分分析的视网膜眼底图像序列变化检测

在计算机辅助诊断系统中,视网膜眼底图像序列的变化检测是一项重要且具有挑战性的任务。针对视网膜眼底图像序列采样帧少、光照干扰大、难以获得稳健的背景模型,提出了一种基于张量鲁棒主成分分析(tensor robust principal component analysis, TRPCA)的变化检测方法。该方法以TRPCA为模型,通过对序列背景扩充,再利用张量分解而获得变化区域：首先,先选择出序列中最接近正常状态的一张图像作为背景模型;然后,通过预处理将单帧背景模型扩张成多帧背景使得背景模型包含更丰富的光照变化;接着,将整个序列建模为一个3维张量体;最后,利用总变分约束背景模型和变化区域的时空连续性,通过Tucker分解分离出背景模型,获得变化区域。实验结果表明,与基于矩阵的鲁棒主成分分析(matrix robust principal component analysis, Matrix RPCA)方法,Masked-RPCA方法以及不加总变分约束的TRPCA方法相比,基于总变分约束的TRPCA方法能够更准确地分离出变化区域,对血管和光照干扰更具有鲁棒性。     

一种浅层非对称结构的视网膜血管分割网络

针对传统视网膜血管分割网络随着网络深度加深导致微小特征信息丢失,网络分割灵敏度低的问题,提出了一种有别于传统对称编码-解码模块的非对称视网膜血管分割结构。网络权重参数量为7.2MB,以残差注意力模块和多尺度空洞卷积模块作为基础特征提取模块,特征图的最大通道层数只有64层,特征图尺寸减半和反卷积操作都只有两次,能够减少特征图尺寸变化带来的信息丢失现象。本文所提方法在DRIVE和CHASE-DB1数据集上进行测试的准确性分别为96.85%和97.39%,灵敏度分别为84.03%和86.50%,特异性分别为98.08%和98.12%,AUC分别为98.63%和98.99%。     

用于视网膜血管分割的半监督深度学习框架北大核心CSCD

针对目前视网膜血管分割任务中伪标签质量参差不齐,获得高质量的伪标签需要经过筛选的问题,本文提出了一种新的用于视网膜血管分割的半监督深度学习框架。该框架采用分而治之的思想来处理数据,针对有标签数据,采用传统的深度学习方法;针对无标签数据,采用Mean teacher模型,通过对比同一输入的不同形态输出,让模型学习无标签数据之间的共同特征,避免了采用伪标签技术带来的筛选过程。本文将U型网络(u-neural networks,U-Net)、Dense-Net和Ladder-Net 3个基准网络放入该框架,在DRIVE和CHASEDB1数据集上进行训练测试,均取得了较好的分割效果,表明本文框架具有提高网络区分不同阈值像素的能力。     

Comparison of three-dimensional surface scanning techniques for capturing the external ear

Congenital facial anomalies, such as microtia (malformation of the external ear), lead to significant psychosocial effects starting from early childhood. Three-dimensional (3D) scanning and advanced manufacturing are being investigated as a cheaper and more personalised method of fabricating reconstructive treatments for patients compared to traditional approaches. To date, most case studies have used expensive 3D scanners, yet, there is potential for low-cost devices to provide comparable results. This study aimed to investigate these different approaches. Both ears of 16 adult participants were scanned with three devices: Artec Spider (Artec Group), Intel^® RealSense? (Intel), and the Apple iPhone^® 7 (Apple Inc.) combined with photogrammetry using 90, 60 and 30 photographs. The scanning time, processing time, accuracy, completeness, resolution and repeatability of each technique were assessed using the Artec Spider as a reference scanner. Our results show that the iPhone had the longest processing time however, this decreased nine-fold when reducing the number of photos from 90 to 30. There was no significant difference in the accuracy, completeness or repeatability of the iPhone scans with 90 photographs (1.4?±?0.6?mm, 79.9%, 1.0?±?0.1?mm), 60 photographs (1.2?±?0.2, 79.3%, 0.9?±?0.2?mm) or 30 photographs (1.2?±?0.3?mm, 74.3%, 1.0?±?0.2?mm). The Intel RealSesne performed significantly worse in each parameter (1.8?±?03?mm; 46.6%, 1.4?±?0.5). Additionally, the RealSense had significantly lower resolution with not enough detail captured for the application. These results demonstrate that the ear can be accurately 3D scanned using iPhone photographs. We would recommend capturing between 30 and 60 photographs of the ear to create a scan that is accurate but without the downfall of long processing time. Using these methods we aim to provide a more comfortable setting for the patient and a lower-cost and more personalised ear prosthesis.     

An effective deep learning model for grading abnormalities in retinal fundus images using variational auto-encoders

Diabetic retinopathy (DR) and Diabetic Macular Edema (DME) are severe diseases that affect the eyes due to damage in blood vessels. Computer-aided automated grading will help clinicians conduct disease diagnoses at ease. Experiments of automated image processing with deep learning techniques using CNN produce promising results, especially in the medical imaging domain. However, the disease grading tasks in retinal images using CNN struggle to retain high-quality information at the output. A novel deep learning model based on variational auto-encoder to grade DR and DME abnormalities in retinal images is proposed. The objective of the proposed model is to extract the most relevant retinal image features efficiently. It focuses on addressing less relevant candidate region generation and translational invariance present in images. The experiments are conducted in IDRID dataset and evaluated using accuracy, U-kappa, sensitivity, specificity and precision metrics. The results outperform compared with other state-of-art techniques.     

Differentiating types of muscle movements using a wavelet based fuzzy clustering neural network

Abstract: The electromyographic signals observed at the surface of the skin are the sum of many small action potentials generated in the muscle fibres. After the signals are processed, they can be used as a control source of multifunction prostheses. The myoelectric signals are represented by wavelet transform model parameters. For this purpose, four different arm movements (elbow extension, elbow flexion, wrist supination and wrist pronation) are considered in studying muscle contraction. Wavelet parameters of myoelectric signals received from the muscles for these different movements were used as features to classify the electromyographic signals in a fuzzy clustering neural network classifier model. After 1000 iterations, the average recognition percentage of the test was found to be 97.67% with clustering into 10 features. The fuzzy clustering neural network programming language was developed using Pascal under Delphi.     

基于OpenGL的三维牙颌模型可视化研究

为了实现牙颌模型的三维数据可视化,通过计算机逼真的显示牙列形态,并进行交互测量及变形等操作.在口腔修复体计算机辅助设计中,以OpenGL为渲染工具,结合自主设计的牙颌模型三维激光扫描仪获取的点云数据,采用基于B样条曲面的三维重建技术,以Visual C 为开发工具实现数字化牙颌模型的三维可视化.以Visual C 和OpenGL为工具,减少了程序设计难度,提高了执行效率,图形显示更加逼真流畅.本系统的设计满足了口腔医生的需求,给患者带来了便利,并促进了口腔修复体的开发制作.     

国内外人工腿(假肢)研究的进展及发展趋势

人工腿(假肢)一直是机器人学和生物医学工程领域 的一个前沿性研究课题．在这方面的研究中,英、日两国于90年代中期研制成功的智能型 人工腿反映了该研究的最新进展．本文共分五个部分,第一部分是引言;第二部分重点介绍 国外在90年代初以前人工腿的研究概况;第三部分重点介绍国外从90年代初至今人工腿 的研究概况,特别是智能人工腿的研究情况;第四部分重点介绍国内人工腿的研究现状;最 后一部分介绍人工腿研究的发展趋势．     

基于背景估计和同态滤波的眼底图像归一化

针对眼底图像的归一化问题,提出一种基于背景估计和同态滤波的眼底图像归一化方法。从图像观测模型的角度出发,利用背景取样估计和插值技术实现图像背景的估计,通过边界奇异值处理和同态滤波增强图像细节,使图像的非均匀照度和对比度得到归一化增强。实验结果表明,与其他方法相比,该方法的归一化性能较好。