Artikel
Classification of human glioma by infrared spectroscopy
Klassifikation humaner Gliome durch Infrarot-Spektroskopie
Suche in Medline nach
Autoren
Veröffentlicht: | 4. Mai 2005 |
---|
Gliederung
Text
Objective
Optical spectroscopic imaging techniques have the potential to expand the ability to diagnose tissue beyond that of standard pathological methods because they characterise tissue at a microstructural molecular level with a spatial resolution in the single cell range. With the aim to achieve an intraoperative glioma grading, infrared spectroscopy was used to characterise and classify specimens of human glioma patients according to the WHO grading system.
Methods
Spectroscopic maps of native specimens of patients with human gliomas were collected ex vivo by infrared (IR) spectroscopy. Variations of tissue composition within each spectroscopic map were demonstrated by cluster analysis and were compared to histopathological findings. In certain glioma patients with inhomogeneous tumours showing regions of different tumour grades, multiple specimens were collected using neuronavigation. Spectroscopic maps were compared to histopathological and radiological findings.
Results
Accurate classification was achieved by a supervised multivariate algorithm called soft independent modelling of class analogies (SIMCA) and a genetic optimal region selection routine and linear discriminant analysis. A classification success rate of up to 96% was obtained for the glioma grading depending on the complexity of data analysis. Normal tissue and low-grade or high-grade regions of glioma could be reliably distinguished by IR spectroscopy.
Conclusions
Infrared spectroscopy has a great potential for identification and grading of cerebral glioma.