Raman spectroscopy is a promising tool towards biopsy under vision as it provides label-free image contrast based on intrinsic vibrational spectroscopic fingerprints of the specimen. The current study applied the spectral unmixing algorithm vertex component analysis (VCA) to probe cell density and cell nuclei in Raman images of primary brain tumor tissue sections. Six Raman images were collected at 785 nm excitation that consisted of 61 by 61 spectra at a step size of 2 micrometers. After data acquisition the samples were stained with hematoxylin and eosin for comparison. VCA abundance plots coincided well with histopathological findings. Raman spectra of high grade tumor cells were found to contain more intense spectral contributions of nucleic acids than those of low grade tumor cells. Similarly, VCA endmember signatures of Raman images from high grade gliomas showed increased nucleic acid bands. Further abundance plots and endmember spectra were assigned to tissue containing proteins and lipids, and cholesterol microcrystals. Since no sample preparation is required, an important advantage of the Raman imaging methodology is that all tissue components can be observed - even those that may be lost in sample staining steps. The results demonstrate how morphology and chemical composition obtained by Raman imaging correlate with histopathology and provide complementary, diagnostically relevant information at the cellular level.

No abstract

In this paper, we introduce an end-to-end Amharic text-line image recognition approach based on recurrent neural networks. Amharic is an indigenous Ethiopic script which follows a unique syllabic writing system adopted from an ancient Geez script. This script uses 34 consonant characters with the seven vowel variants of each (called basic characters) and other labialized characters derived by adding diacritical marks and/or removing parts of the basic characters. These associated diacritics on basic characters are relatively smaller in size, visually similar, and challenging to distinguish from the derived characters. Motivated by the recent success of end-to-end learning in pattern recognition, we propose a model which integrates a feature extractor, sequence learner, and transcriber in a unified module and then trained in an end-to-end fashion. The experimental results, on a printed and synthetic benchmark Amharic Optical Character Recognition (OCR) database called ADOCR, demonstrated that the proposed model outperforms state-of-the-art methods by 6.98% and 1.05%, respectively.

No abstract