科研之友
微信
新浪微博
Facebook
分享链接
摘要
Mass-spectrometry imaging (MSI), the combination of molecular mass analysis and spatial information, providing visualization of molecules on complex biological surfaces, is currently receiving a significant amount of attention among the mass-spectrometry community. One important problem in this research field concerns the development of an effective method for the classification and identification of MSI data, especially for both differentiating a cancerous tissue from adjacent normal tissues and classifying the different functional regions in a complex biological tissue. For this purpose, we developed a new method, which involved image reconstruction from raw mass-spectral data; preprocessing of MSI data; classification of tissue regions with reference to the background regions using self-organizing feature maps; and identification of regions of special interest in whole-tissue samples by learning-vector quantization. The MSI data of samples of six pairs (12 tissue samples) of human cancerous and adjacent normal bladder tissues were used to test the efficacy of this method. The results showed an error rate of less than 23.38% for identification of cancerous regions and an error rate of less than 9.08% for identification of the adjacent normal regions. The method was also tested with reference to classification of the regions of the white matter and gray matter of three adjacent slices of mouse brain tissue. The slice in the middle was used to establish an identification model, and the other two slices were used to test the model. The inconsistency rate of the results obtained by identification using a self-organizing feature map was less than 4% compared with the results using learning-vector quantization. This indicated that the method could be carried out simply and efficiently to extend the capability of MSI and underlined its potential to be a regular tool in the studies related to clinical applications.
