论文信息 - Detection and grading of human gliomas by FTIR spectroscopy and a genetic classification algorithm

Detection and grading of human gliomas by FTIR spectroscopy and a genetic classification algorithm

A new approach is presented to distinguish cancerous from normal brain tissue via linear discriminant analysis of Fourier transform infrared (FTIR) spectra. FTIR microspectroscopy was used to map various thin-section tumor samples with different malignancy grades (grades II-VI) and non-tumor samples obtained from various patients by surgical removal. Spectral analysis revealed features characteristic of tumors with increasing malignancy. A genetic region selection algorithm combined with linear discriminant analysis was used to derive classifiers distinguishing among spectra of control tissue, astrocytoma grade II, astrocytoma grade III and glioblastoma grade IV. Employing the World Health Organization histopathological diagnostic scheme as the gold standard, the spectra were classified with a success rate of approximately 85 percent. These results demonstrate the potential of the combination of FTIR spectroscopy and pattern recognition routines in providing a more objective method for brain tumour grading and diagnosis.

[1] D A Ramsay,et al. Classification of Astrocytomas and Malignant Astrocytomas by Principal Components Analysis and a Neural Net , 1996, Journal of neuropathology and experimental neurology.

[2] J Hildebrand,et al. Management of malignant brain tumors. , 1997, European neurology.

[3] M. Kates,et al. FTIR study of the monosialoganglioside GM1 in perdeuterated dimyristoylglycerophosphocholine (DMPCd54) multilamellar bilayers: spectroscopic evidence of a significant interaction between Ca2+ ions and the sialic acid moiety of GM1. , 2000, Biochemistry.

[4] B. Scheithauer,et al. The New WHO Classification of Brain Tumours , 1993, Brain pathology.

[5] G. McLachlan. Discriminant Analysis and Statistical Pattern Recognition , 1992 .

[6] B. L. Bauer,et al. Gangliosides: the relevance of current research to neurosurgery. , 1991, Journal of neurosurgery.

[7] R L Somorjai,et al. Near‐optimal region selection for feature space reduction: novel preprocessing methods for classifying MR spectra , 1998, NMR in biomedicine.

[8] P Lasch,et al. FT-IR microspectroscopic imaging of human carcinoma thin sections based on pattern recognition techniques. , 1998, Cellular and molecular biology.