Classifying Images of Cell Types in Unstained Tissue

The Spero^®基于激光的红外显微镜用于产生高清晰度，即使用标准化组织学方案制备的未染色薄结肠组织切片的化学图像。接下来，将常规化学计量分析技术与数据一起使用，以便可以识别和分割组织类型。

Solution

薄（10μm）切除的结直肠组织切片的光谱图像数据群安装在红外载玻片上，然后再用第一代获取Spero^®显微镜。This had a spectral range of 900 – 1,800 cm⁻¹（5,500 - 1,100 nm），1.3μm的像素分辨率和由0.7 Na设置的衍射限制性分解性，以及一个单英尺视场（FOV）为650μm×650μm。

录音是由覆盖全频的图像立方体（900 - 1,800厘米）^-1在4厘米处^-1steps) alongside cubes targeting 10 sparsely selected frequencies within the band. These were then analyzed using in ImageLab^®multimodal chemometrics software from Epina Software, based in Vienna, Austria.

To assist with chemometrics data reduction, the raw image cubes were pre-processed using the following sequence:

• Noise reduction using the well-known maximum noise fraction transform
• Rejection of pixels with an amide I peak value (1,656 cm^-1) of less than 0.05 AU (done as a spectral quality test)
• 全带光谱转换为第二个衍生物以去除基线
• 使用Savitzky-Golay算法的9点平滑
• Spectral normalization of all spectral vectors to achieve zero mean and unit standard deviation. This was done to reduce the influence of intensity changes caused by differences in cellular density and tissue thickness.

完成这些预处理步骤后，完整的乐队（900 - 1,800厘米^-1）通过无监督的K均值聚类方法分析了第二个衍生数据集。这是一种非层次迭代方法，可为每个频谱获得“硬”类成员身份。

接下来构建了6个类的错误颜色图像，在整个成像字段中，每个像素都将分配给具有相应颜色的类成员资格。使用这种基于该特定像素的光谱响应的错误着色方法意味着观察者具有快速，简单的方法来“看到”样品中的生化差异。在这种情况下，关键差异对应于细胞和组织结构的基本类型。

Results

从the k-means clustered images, the different structures of the colon tissue are clearly visible. This included the colonic crypts, submucosa, and lamina propria. Within the colonic crypts especially, spectral signatures included combinations of mucin glycosylation bands at 1,044, 1,076, 1,120, and 1,374 cm^-1以及1,740厘米处的透明脂质酯带^-1。

然而，相反，粘膜粘膜是通过许多可以直接链接到具有1,204、1,236、1,280、1,336（酰胺III）和1,452 cm的结构蛋白胶原蛋白胶原蛋白胶原蛋白识别的。^-1respectively.

Spectroscopic differences between the lamina propria and adenocarcinoma are much subtler. These are located at nucleic acid-related vibrations at ca. 964, 1,062, 1,090, and 1,236 cm^-1。

As can be seen, the 10-frequency sparse image cube datasets were used to create high-speed, qualitative chemical maps of collagen, protein and mucin-rich regions which could then be used to guide more detailed exploration and analysis.

This information has been sourced, reviewed and adapted from materials provided by Daylight Solutions Inc.

For more information on this source, please visit日光解决方案公司