Towards automated cancer screening: Label-free classification of fixed cell samples using wavelength modulated Raman spectroscopy.

Lana Woolford,C Simon Herrington,Mingzhou Chen,Kishan Dholakia

doi:10.1002/jbio.201700244

Lana Woolford, C Simon Herrington + Show 2 more

Open Access

https://doi.org/10.1002/jbio.201700244

Copy DOI

Abstract

The ability to provide quantitative, objective and automated pathological analysis would provide enormous benefits for national cancer screening programmes, in terms of both resource reduction and improved patient wellbeing. The move towards molecular pathology through spectroscopic methods shows great promise, but has been restricted by spectral quality, acquisition times and lack of direct clinical application. In this paper, we present the application of wavelength modulated Raman spectroscopy for the automated label‐ and fluorescence‐free classification of fixed squamous epithelial cells in suspension, such as those produced during a cervical smear test. Direct comparison with standard Raman spectroscopy shows marked improvement of sensitivity and specificity when considering both human papillomavirus (sensitivity +12.0%, specificity +5.3%) and transformation status (sensitivity +10.3%, specificity +11.1%). Studies on the impact of intracellular sampling location and storage effects suggest that wavelength modulated Raman spectroscopy is sufficiently robust to be used in fixed cell classification, but requires further investigations of potential sources of molecular variation in order to improve current clinical tools.

Highlights

The introduction of national screening programmes such as the Papanicoloau (‘Pap’) test and mammogram have achieved considerable success, and are estimated to prevent, for example, 4500 [1] and 1300 [2] deaths in the United Kingdom every year from cervical and breast cancer, respectively
The primary purpose of cervical screening is to detect abnormal epithelial cells that indicate the presence of cervical pre-cancer or cancer
We demonstrate the use of wavelength modulated Raman spectroscopy (WMRS) [12] as a tool for the automated classification of unlabeled fixed cells in suspension, reflecting its potential utility in diagnosing clinical liquid-based cytology samples

Summary

| INTRODUCTION

The introduction of national screening programmes such as the Papanicoloau (‘Pap’) test and mammogram have achieved considerable success, and are estimated to prevent, for example, 4500 [1] and 1300 [2] deaths in the United Kingdom every year from cervical and breast cancer, respectively. We demonstrate the use of wavelength modulated Raman spectroscopy (WMRS) [12] as a tool for the automated classification of unlabeled fixed cells in suspension, reflecting its potential utility in diagnosing clinical liquid-based cytology samples. The laser wavelength is modulated throughout the acquisition process using a tunable laser [14] This causes the Raman spectral peaks to shift fractionally when measured in terms of scattered wavelength. Methods for post-processing wavelength modulated spectra in order to account for wavelength shift and remove this static fluorescence background have been tested, including standard excitation Raman difference spectroscopy, SD, Fourier filtering, least-squares fit and principal component analysis (PCA), with PCA providing the best signal-to-noise ratio [15]. The aim of this study is to assess the practical potential of WMRS as a method for the automated classification of fixed squamous epithelial cells, modelling cervical cancer diagnosis

| MATERIALS AND METHODS

Findings

| RESULTS

| CONCLUSION