Key Laboratory of Optoelectronic Devices and Systems of Ministry of Education and Guangdong Province, College of Physics and Optoelectronic Engineering, Shenzhen University, Shenzhen 518060, China
Fund Project:Project supported by the National Natural Science Foundation of China (Grant Nos. 61975131, 61775144, 61525503, 61620106016, 61835009), the Natural Science Foundation of Guangdong Province, China (Grant No. 2018A030313362), the Key Project of Department of Education of Guangdong Province, China (Grant No. 2016KCXTD007), and the Shenzhen Basic Research Project, China (Grant Nos. JCYJ20170818141701667, JCYJ20170818144012025, JCYJ20170412105003520)
Received Date:14 June 2020
Accepted Date:11 July 2020
Available Online:25 November 2020
Published Online:05 December 2020
Abstract:Laser scanning confocal microscope (LSCM) is one of the most important tools for biological imaging due to its strong optical sectioning capability, high signal-to-noise ratio, and high resolution. On the basis of LSCM, line-scanning fluorescence microscopy (LSFM) uses linear scanning instead of point scanning to improve the speed of image acquisition. It has the advantages of simple system structure, fast imaging speed, and weak phototoxicity, and in addition, it is more suitable for high-resolution and fast imaging of living thick samples. It is of great significance for studying the life science, biomedicine, and others. However, the current LSFM technology still faces many urgent problems in terms of system flexibility, imaging speed, resolution and optical sectioning capabilities. Therefore, based on the existing multifocal structured illumination microscopy (MSIM) in our laboratory, a digital line-scanning fluorescence microscopy (DLSFM) based on digital micromirror device(DMD) is presented in this paper. In the illumination path, a high-speed spatial light modulator DMD is adopted to realize multi-line parallel scanning excitation, which simplies the optical system and improves the flexibility and scanning speed of the system. A DLSFM image reconstruction algorithm based on the standard deviation of fluorescence signal is proposed, which is combined withthree-dimensional (3D) Landweber deconvolution algorithm to achieve 3D high-resolution optical slice image reconstruction. On this basis, the imaging experiments on fluorescent beads and standard samples of mouse kidney section are carried out by using DLSFM. The experimental results show that the resolution of DLSFM in the x, y and z directions is 1.33 times, 1.42 times and 1.19 times that of wide field microscope, respectively, and the fast 3D high-resolution optical sectioning imaging of biological samples is realized, which lays a technical foundation for further developing the rapid high-resolution imaging of the whole cells and tissues in vivo. Keywords:line-scanning microscopy/ digital micromirror device/ fluorescence microscopy
其中, $N$为DMD的微反射镜总数量, $n$为开状态下的DMD微反射镜数量. 首先, 通过对均匀的罗丹明荧光溶液进行线扫描成像来分析F与扫描线数的关系, 如图2所示. 通过采集单帧不同密度线条纹激发下的罗丹明荧光溶液图像, 如图2(a)所示, 并对采集的荧光溶液图像不同帧的同一平行线位置的横截面宽度进行计算, 如图2(b)所示. 图 2 DMD不同填充因子的数字线扫描成像分析 (a)不同密度线条纹激发下的罗丹明荧光溶液图像; (b) 不同帧的同扫描线位置的强度轮廓线 Figure2. Digital line scan imaging analysis of DMD with different fill factor: (a) The rhodamine fluorescence solution images excited by different density scanning lines; (b) the intensity profiles through the dotted line in (a).
从图2(b)中看出, 随着F的增大, 不同线密度条件下激发样片产生荧光线条的横截面宽度几乎接近于一恒定值, 但当F增大到0.0566时, 可以看出线条的横截面强度除了中心的主峰强度外, 有一些旁瓣信号, 这主要是由于高线密度下平行线间的散射光和离焦信号导致的. 因此, 为了获取理想F情况下平行线阵列的激发数目, 需对图像SNR作进一步分析, 如图3所示. SNR定义如下: 图 3 DMD不同填充因子的信噪比分析 (a) 焦面信号和背景信号强度分布图; (b) 焦面信号和背景信号强度与DMD填充因子的关系曲线; (c) DMD不同填充因子的信噪比曲线 Figure3. SNR analysis of DMD with different fill factor: (a) The intensity distributions of focal plane signal and background signal intensity; (b) the intensity curves of focal planne signal (black) and background signal (red); (c) the curve of the SNR versus fill factor of DMD.
为了验证DLSFM的成像能力, 可通过测量直径为100 nm的荧光珠对该系统的分辨率进行标定, 对视场范围内25颗荧光珠的图像进行统计平均, 得到了系统在$x, y, z$方向的平均分辨率, 用荧光珠的半高全宽(full width at half maximum, FWHM)值来表征, 如图4所示. 图 4 100 nm荧光珠标定的系统三维空间分辨率 (a)宽场图像; (b)虚拟狭缝重构图像; (c)标准差重构图像; (d)虚拟狭缝与LW解卷积组合重构图像; (e)标准差与LW解卷积组合重构图像. 其右边对应的直方图为25颗荧光珠在沿$x, y, z$方向的FWHM分布, $\overline X, \overline Y, \overline Z $分别为其平均值 Figure4. 3D spatial resolution of the system calibrated by the 100 nm fluorescent bead: (a) WF image; (b) VS image; (c) STD image; (d) VS + DE image; (e) STD + DE image. The corresponding histograms on the right show the measured distribution of $x, y, z$ FWHM values from 25 beads.
接下来, 利用搭建的DLSFM系统, 选用荧光标记的老鼠肾切片(FluoCellTM prepared slide#3 mouse kidney section, *AF 488 WGA, AF 568 phalloidin, Product of US)作为实验样片, 选取其中厚度为11.2 μm的肾小球作为实验对象, 进行扫描成像, 样品荧光发射光波长为568 nm, 样品面像素尺寸大小为108 nm. 在成像采集过程中, DMD上7条激发线横向并行扫描的步数为48步, 扫描速率为250 Hz, 三维电动纳米位移台轴向扫描间隔为200 nm. 因此, 共获得2688幅线扫描源图像. 这些采集的样品图像经VS、STD, 以及LandweberDE算法处理后获得样品的不同深度的光切片层析图像, 如图5所示. 图 5 肾小球轴向光切片的重构图像 Figure5. Reconstructed slice images of the same glomerulus.
从图5中可以看出, WF模糊不清, 大致能分辨外轮廓, 对于边缘细节几乎难以分辨开来, 经VS图像重构方法处理后, 离焦信号得到了一定的去除, 边缘细节得到了一定的凸显; 然而, VS图像重构方法与STD图像重构方法成像相比, 其对比度与分辨率相对较低. 利用Landweber解卷积算法在二者基础之上进行处理, 其对比度及分辨率得到了进一步改善, 尤其是对于标准差图像重构, 其经解卷积算法处理后, 样品细节结构得到了显著的凸显, 图像分辨率得到了进一步的提升. 为了细致观察其重构效果, 选取轴向深度处于8 μm处的肾小球切片的一区域进行放大, 如图6所示, 其中, 图6(a)—(f)分别为WF图像、VS图像和STD图像; 图6(a)—(c)分别经Landweber DE算法处理成像(图6(d)—(f)); 图6(i)—(n)分别为图6(a)—(f)黄色线框区域的放大. 图 6 不同算法的图像重构细节比较 (a)?(c) 分别为WF图像、VS图像和STD图像; (d)?(f) 分别由(a)?(c)经Landweber DE算法处理成像; (i)?(n) 分别为图(a)?(f)黄色线框区域的放大图 Figure6. Comparison of image reconstruction details of different algorithms: (a)?(f) WF image, VS image, and STD image; (d)?(f) the Landweber DE images for (a)?(c); (i)?(n) magnified view of the yellow rectangular areas in (a)?(f).
从图6可以看出, 经由图6(a)—(f)黄色线框分别放大后的图6(i)—(n), 其中, 图6(l)—(n)在图6(i)—(k)的基础之上分辨率得到了进一步提升, 尤其是对于图6(n), 与图6(m)相比, 其样品结构凸显得尤为明显. 因此, 说明了数字线扫描WF、VS图像重构、STD图像重构经Landweber算法解卷积后图像分辨率得到进一步的提升, 其中STD图像对比度以及分辨率明显比VS图像重构的要高, 如图7所示, 其中, 图7(a)—(c)分别为WF图像、VS图像和STD图像; 图7(d)—(f)分别为图7(a)—(c)经Landweber解卷积算法处理成像; 图7(g)和7(h)分别为图7(a)—(f)白色虚线位置的归一化强度拟合. 图 7 不同算法重构图像的对比度比较 (a)?(c) 分别为WF图像、VS图像和STD图像; (d)?(f) 分别为图(a)?(c)经Landwebe解卷积算法处理成像; (g)?(h) 分别为图(a)?(f)白色虚线位置的归一化强度轮廓图 Figure7. Contrast comparison of reconstructed images with different algorithms: (a)?(c) WF image, VS image, and STD image; (d)?(f) images obtained from (a)?(c) processed by Landwebe deconvolution algorithm; (g)?(h) the normalized intensity profiles through the white dotted line position in (a)?(f).