Share Email Print
cover

Proceedings Paper

Incorporating machine learning with Raman spectroscopy to differentiate bone types
Format Member Price Non-Member Price
PDF $17.00 $21.00

Paper Abstract

Raman spectroscopy permits label-free molecular quantitation of biological samples in situ in a non-destructive manner. Combining machine learning with Raman spectroscopy has increased its potential for use in molecular imaging and discrimination of living cells and tissues in biological research fields. In this work, Raman spectroscopy was paired with machine learning techniques to classify specimen of similar tissues. Raman spectra of rat long bone, rabbit long bone, and rabbit crania were collected and classified into their respective categories. The spectra were truncated to the range of 400 to 1800 wavenumbers. To train and validate the machine learning algorithms, the data were randomly split such that 80% (n = 499) of the data were used for training, and 20% (n = 125) were used for validation. Three approaches were taken to prepare the data for classification. The first approach utilized all Raman intensities between 400 and 1800 wavenumbers to perform the classification. The second approach reduced the dimensions of the dataset using Principal Component Analysis (PCA) prior to performing classification. The third approach also reduced the dimensions of the dataset by extracting intensities of peaks that are of interest for bone analysis and using these peaks for classification. The peaks chosen were Amide I, Amide III, Proline, CH2 wag, and Carbonate. Raman spectra were classified using supervised learning techniques for each data preparation approach. The supervised methods include Support Vector Machine (SVM), Decision Tree, Random Forrest, and Naïve Bayes. The three groups were successfully sorted into their respective classes by the applied classification algorithms. The most successful classification models were achieved by reducing the dataset to peaks of interest, and performing classification utilizing Support Vector Machine achieving a validation accuracy up to 98.40%. This proof of concept has potential to be applied to numerous research applications that require sensitive discrimination between similar tissues.

Paper Details

Date Published: 21 February 2020
PDF: 8 pages
Proc. SPIE 11252, Advanced Chemical Microscopy for Life Science and Translational Medicine, 1125217 (21 February 2020); doi: 10.1117/12.2546463
Show Author Affiliations
Michael Sieverts, PolarityTE, Inc. (United States)
Kendall Stauffer, PolarityTE, Inc. (United States)
Caroline Garrett, PolarityTE, Inc. (United States)
Pratima Labroo, PolarityTE, Inc. (United States)
Nikolai Sopko, PolarityTE, Inc. (United States)


Published in SPIE Proceedings Vol. 11252:
Advanced Chemical Microscopy for Life Science and Translational Medicine
Ji-Xin Cheng; Wei Min; Garth J. Simpson, Editor(s)

© SPIE. Terms of Use
Back to Top
PREMIUM CONTENT
Sign in to read the full article
Create a free SPIE account to get access to
premium articles and original research
Forgot your username?
close_icon_gray