Multispectral microscopy for cell differentiation in thyroid cytology

This paper presents a multispectral microscopy system for quantitative cytology. The automated system aims to classify cytology samples of follicular lesions for the purpose of disease diagnosis. While conventional practices rely on the analysis of grey scale or RGB color images, presented system uses thirty one spectral bands for analysis. An intelligent system designed for image acquisition, image segmentation, feature extraction, and classification is presented. Results are presented for the problem of discriminating among four lesion types. In addition, classification performance is compared to the case where multispectral information is not taken into consideration. Results show that the developed system and the use of multispectral information along with morphometric information extracted from spectral images can significantly improve the classification performance and aid in the process of disease diagnosis.