Show simple item record

AuthorAli M.H.
AuthorRakib F.
AuthorAl-Saad K.
AuthorAl-Saady R.
AuthorLyng F.M.
AuthorGoormaghtigh E.
Available date2020-03-18T08:10:11Z
Publication Date2018
Publication NameJournal of Molecular Structure
ResourceScopus
ISSN222860
URIhttp://dx.doi.org/10.1016/j.molstruc.2018.03.044
URIhttp://hdl.handle.net/10576/13359
AbstractBreast cancer is the second most common cancer after lung cancer. So far, in clinical practice, most cancer parameters originating from histopathology rely on the visualization by a pathologist of microscopic structures observed in stained tissue sections, including immunohistochemistry markers. Fourier transform infrared spectroscopy (FTIR) spectroscopy provides a biochemical fingerprint of a biopsy sample and, together with advanced data analysis techniques, can accurately classify cell types. Yet, one of the challenges when dealing with FTIR imaging is the slow recording of the data. One cm2 tissue section requires several hours of image recording. We show in the present paper that 2D covariance analysis singles out only a few wavenumbers where both variance and covariance are large. Simple models could be built using 4 wavenumbers to identify the 4 main cell types present in breast cancer tissue sections. Decision trees provide particularly simple models to reach discrimination between the 4 cell types. The robustness of these simple decision-tree models were challenged with FTIR spectral data obtained using different recording conditions. One test set was recorded by transflection on tissue sections in the presence of paraffin while the training set was obtained on dewaxed tissue sections by transmission. Furthermore, the test set was collected with a different brand of FTIR microscope and a different pixel size. Despite the different recording conditions, separating extracellular matrix (ECM) from carcinoma spectra was 100% successful, underlying the robustness of this univariate model and the utility of covariance analysis for revealing efficient wavenumbers. We suggest that 2D covariance maps using the full spectral range could be most useful to select the interesting wavenumbers and achieve very fast data acquisition on quantum cascade laser infrared imaging microscopes. ? 2018 The Authors
SponsorThis work was made possible by a NPRP Award [7 - 1267 - 3?328] from the Qatar National Research Fund (a member of The Qatar Foundation). E.G. is Research Director with the National Fund for Scientific Research (Belgium) . The statements made herein are solely the responsibility of the authors.
Languageen
PublisherElsevier B.V.
Subject2D correlation spectroscopy
Breast cancer
Infrared imaging
Infrared spectroscopy
Spectral histopathology
TitleA simple model for cell type recognition using 2D-correlation analysis of FTIR images from breast cancer tissue
TypeArticle
Pagination472 - 479
Volume Number1163
dc.accessType Abstract Only


Files in this item

FilesSizeFormatView

There are no files associated with this item.

This item appears in the following Collection(s)

Show simple item record