部分还原氧化石墨烯的制备、 表征及对人宫颈癌HeLa细胞的体外杀伤作用

以氧化石墨烯(GO)为原料, 利用温和方法制备了3种不同还原程度的部分还原氧化石墨烯pRGO₁, pRGO₂和pRGO₃(pRGO_1—3)； 利用傅里叶变换红外光谱(FTIR)、 拉曼光谱(Raman)、 X 射线光电子能谱(XPS)、 紫外-可见光谱(UV-Vis)、 透射电子显微镜(TEM)和 EDS能谱对其结构和形貌进行了表征. 细胞实验结果表明, 无激光照射下pRGO_1—3本身的细胞毒性较低; 近红外(NIR)激光照射下pRGO_1—3通过光热和光毒性双重作用杀伤肿瘤细胞. 实验结果显示了pRGO 在肿瘤光热疗法和光动力疗法领域的应用潜力.     

Flower-like titanium dioxide as novel co-reaction accelerator for ultrasensitive “off–on” electrochemiluminescence aptasensor construction based on 2D g-C3N4 layer for thrombin detection

Journal of Solid State Electrochemistry - An ultrasensitive “signal-off–on” electrochemiluminescence (ECL) biosensor is constructed based on f1-TiO2/g-C3N4/PDA for thrombin...     

A microfluidics-dispensing-printing strategy for Janus photonic crystal microspheres towards smart patterned displays

From the implementation point of view, the printable magnetic Janus colloidal photonic crystals (CPCs) microspheres are highly desirable. Herein, we developed a dispensing-printing strategy for magnetic Janus CPCs display via a microfluidics-automatic printing system. Monodisperse core/shell colloidal particles and magnetic Fe₃O₄ nanoparticles precursor serve as inks. Based on the equilibrium of three-phase interfacial tensions, Janus structure is successfully formed, followed by UV irradiation and self-assembly of colloid particle to generate magnetic Janus CPCs microspheres. Notably, this method shows distinct superiority with highly uniform Janus CPCs structure, where the TMPTA/Fe₃O₄ hemisphere is in the bottom side while CPCs hemisphere is in the top side. Thus, by using Janus CPCs microspheres with two different structural colors as pixel points, a pattern with red flower and green leaf is achieved. Moreover, 1D linear Janus CPCs pattern encapsulated by hydrogel is also fabricated. Both the color and the shape can be changed under the traction of magnets, showing great potentials in flexible smart displays. We believe this work not only offers a new feasible pathway to construct magnetic Janus CPCs patterns by a dispensing-printable fashion, but also provides new opportunities for flexible and smart displays.     

基于快速神经网络架构搜索的鲁棒图像水印网络算法

为解决深度学习在图像水印算法中计算量大且模型冗余的问题，提高图像水印算法在抵抗噪声、旋转和剪裁等攻击时的鲁棒性，提出基于快速神经网络架构搜索（neural architecture search，NAS）的鲁棒图像水印网络算法。通过多项式分布学习快速神经网络架构搜索算法，在预设的搜索空间中搜索最优网络结构，进行图像水印的高效嵌入与鲁棒提取。首先，将子网络中线性连接的全卷积层设置为独立的神经单元结构，并参数化表示结构单元内节点的连接，预先设定结构单元内每个神经元操作的搜索空间；其次，在完成一个批次的数据集训练后，依据神经元操作中的被采样次数和平均损失函数值动态更新概率；最后，重新训练搜索完成的网络。水印网络模型的参数量较原始网络模型缩减了92%以上，大大缩短了模型训练时间。由于搜索得到的网络结构更为紧凑，本文算法具有较高的时间性能和较好的实验效果，在隐藏图像时，对空域信息的依赖比原始网络更少。对改进前后的2个网络进行了大量鲁棒性实验，对比发现，本文算法在CIFAR-10数据集上对抵抗椒盐噪声和旋转、移除像素行（列）等攻击优势显著；在ImageNet数据集上对抵抗椒盐高斯噪声、旋转、中值滤波、高斯滤波、JPEG压缩、裁剪等攻击优势显著，特别是对随机移除行（列）和椒盐噪声有较强的鲁棒性。