由于该成像系统采用的超连续谱光源, 可以满足所观测样品本源分子在0～4 000 cm-1内的所有拉曼振动活性模式同时共振增强, 在探测光的作用下, 同时产生宽带相干反斯托克斯拉曼散射信号。 然后, 根据不同的化学键进行光谱图像重构, 可以获得反映不同化学键在样品中的分布的图像。 对110 nm的纯的聚苯乙烯珠所形成的具有一定厚度的薄膜, 通过改变探测激光与超连续谱脉冲之间的时间重合度, 测量形成的相干反斯托克斯拉曼散射信号的时间分布迹线图, 通过其中的1 000 cm-1的化学键强度信号进行指数衰减曲线拟合, 得出具体的退相时间, 与文献中已报道的三色CARS的退相时间相比, 判断是否属于三色CARS。 为了检验系统在实际生物学成像中存在的问题, 我们开展了活体小鼠组织生物学应用成像实验, 对记录的数据在2 940 cm-1的CARS信号进行图像重构, 获得CH化学键在组织中的分布, 然后, 对重构图像直接使用小波变换的去噪方式进行图像去噪, 去噪后的图像具有比较清晰的轮廓, 结果表明, 对于对比度比较强的CARS共振信号, 直接使用小波变换的去噪方式就可以获得比较好的图像效果。 但是, 对于信噪比比较差的共振信号, 使用这种处理方法是不合适的, 需要使用别的方法, 先获取好的信号对比度, 再根据感兴趣的化学键进行图像重构, 然后, 再经过小波变换对图像去噪, 图像不仅会变得清晰平滑, 而且, 具有较好的视觉感官效果。

Our imaging system is based on a super-continuum source, which can provide all the samples’ active Raman vibration mode in a range from 0～4 000 cm-1 at the same time. The coming probe pulse interacts with the sample, forming the broadband coherent anti-Stokes Raman scattering (CARS) signal. Then, the image can be reconstructed according to the different chemical bonds, and the distribution of different chemical bonds in the sample can be obtained. In this paper, 110-nm pure polystyrene beads formed a certain thickness of thin films, which can endure high power, by tuning the detection pulse and super-continuum pulse overlay time, measuring the time distribution of the CARS signal and then using exponential decay curve fitting through a chemical bond signal intensity at 1 000 cm-1. This is called the dephasing time. Additionally, it was compared with the CARS dephasing time reported in the literature, and it was determined that it belonged to two-colour or three-colour CARS. To inspect the system application in biological imaging, we carried out the in vivo tissue biology application experiment through recorded data reconstructed in the CARS image at 2 940 cm-1 and the distribution of CH chemical bonds in tissues, and then directly used wavelet transform for image denoising. The denoised image had a clearer outline, and the results showed that the resonant part of the CARS signal contrast was strong and directly using the wavelet transform denoising method can obtain a better image effect. However, the signal-to-noise ratio of the resonant signal was poor, so the use of this approach was not appropriate, and other methods needed to be used to obtain a good signal contrast with the image. Then, according to the interest chemical bond with a good signal contrast to reconstruct image, after the wavelet transform for image denoising, the image will not only become clear and smooth, but have a good vision effect.