基于神经网络控制的潜用磁罗经自差水下自动校正

为了解决潜用磁罗经不能自动校正自差，且水面校正完全依赖人工操作、校正时间长、校正精度低和易暴露的问题，研制了一种利用神经网络控制技术可实现磁罗经自差自动校正的系统，并提出一种新的剩余自差的测定方法．仿真分析表明该系统降低了校正所需时间，提高了校正精度，避免了潜艇自差校正时的暴露问题。     

Adaptive optimization on ultrasonic transmission tomography-based temperature image for biomedical treatment

Hyperthermia has proven to be beneficial to treating superficial malignancies, particularly chest wall recurrences of breast cancer. During hyperthermia, monitoring the time–temperature profiles in the target and surrounding areas is of great significance for the effect of therapy. An ultrasound-based temperature imaging method has advantages over other approaches. When the temperature around the tumor is calculated by using the propagation speed of ultrasound, there always exist overshoot artifacts along the boundary between different tissues. In this paper, we present a new method combined with empirical mode decomposition(EDM), similarity constraint, and continuity constraint to optimize the temperature images. Simulation and phantom experiment results compared with those from our previously proposed method prove that the EMD-based method can build a better temperature field image, which can adaptively yield better temperature images with less computation for assistant medical treatment control.     

Novel Image Optimization Method for Joint Photoacoustic Tomography

The distribution of speed of sound （SOS） in biomedical tissue and delay compensation （DC） have significant impact on the image quality of photoacoustic tomography （PAT）. When imaging human peripheral joints, using fixed SOS and DC can only ensure that the reconstructed images are focused in a limited depth range, whereas they are defoeused at other depths, which cause severe artifacts and blurring. In this work, a linear-DC based reconstruction approach is proposed to focus the whole PAT image region. It is proved by two in vivo experiments that, compared with traditional delay-and-sum back projection algorithms, the proposed method can effectively optimize the image quality of articular tissues in PAT.     

基于LM算法的神经网络在冠心病诊断中的应用

为了解决冠心病诊断的BP神经网络存在收敛速度慢、容易陷入局部极小以及常出现误诊断等问题．提出一种基于LM算法改进的神经网络诊断系统，包括样本信息选取、病情信息量化、网络学习训练和诊断等过程。临床实验应用表明，这种诊断系统不仅具有算法稳健、样本拟合精度高等优点，而且其诊断效果优于BP算法。     

基于光声谱的生物组织“指纹”光声诊断术*

肿瘤诊断主要采用医学影像结合组织活检病理予以确诊,但尚未实现高效的早期肿瘤筛查,现有影像术的诊断信息不全面和病理的有创、耗时长是关键问题。近年来快速发展起来的新型生物医学光声成像和光声谱分析技术,对生物组织的分子、化学和功能信息高度灵敏,可实现组织"指纹"的量化检测,且实时、无放射性、无电离性,与其他影像技术融合后具有实现无创"质谱"检测的巨大潜力。该文基于光声信号在光学和声学两个维度的谱分析,实现多种肿瘤和正常组织的光声组织"指纹"成像;提取的量化参数对肿瘤、癌前病变和正常组织进行了有效区分。这项技术有望实现高效、无创、低价的在体早期肿瘤筛查。     

一种基于光载波抑制的ROF双工系统

为克服远距离光纤传输系统中色散效应的影响,研究了一种基于光载波抑制的单工ROF(光纤无线电)系统,并提出改进的双工ROF系统传输方案。在中心站,采用20GHz的射频信号将下行链路信号耦合至光载波抑制频段处,通过单模光纤传输50km至基站,并重复利用未调制边带传输上行链路信号。通过仿真得到的系统光谱图和误码率曲线表明:加入上行链路信号传输后,下行链路的光纤能量损耗降低,上、下行链路同时具有较好的抗色散能力。     

一种新型磁罗经系统的误差模型分析及其补偿

给出了一种新型磁通门磁罗经主要误差模型的分析，并基于这种分析讨论了误差的补偿方法。经过试验与仿真分析，这种磁罗经系统补偿后的精度有很大的提高，同时与电罗经相比具有体积小、结构简单、反应快和可靠性高等优点。