Feature driven classification of Raman spectra for real-time spectral brain tumour diagnosis using sound

Ryan Stables, Graeme Clemens, Holly J. Butler, Katherine M. Ashton, Andrew Brodbelt, Timothy P. Dawson, Leanne M. Fullwood, Michael D. Jenkinson, Matthew Baker

Research output: Contribution to journalArticle

12 Citations (Scopus)
71 Downloads (Pure)

Abstract

Spectroscopic diagnostics have been shown to be an effective tool for the analysis and discrimination of disease states from human tissue. Furthermore, Raman spectroscopic probes are of particular interest as they allow for in vivo spectroscopic diagnostics, for tasks such as the identification of tumour margins during surgery. In this study, we investigate a feature-driven approach to the classification of metastatic brain cancer, glioblastoma (GB) and non-cancer from tissue samples, and we provide a real-time feedback method for endoscopic diagnostics using sound. To do this, we first evaluate the sensitivity and specificity of three classifiers (SVM, KNN and LDA), when trained with both sub-band spectral features and principal components taken directly from Raman spectra. We demonstrate that the feature extraction approach provides an increase in classification accuracy of 26.25% for SVM and 25% for KNN. We then discuss the molecular assignment of the most salient sub-bands in the dataset. The most salient sub-band features are mapped to parameters of a frequency modulation (FM) synthesizer in order to generate audio clips from each tissue sample. Based on the properties of the sub-band features, the synthesizer was able to maintain similar sound timbres within the disease classes and provide different timbres between disease classes. This was reinforced via listening tests, in which participants were able to discriminate between classes with mean classification accuracy of 71.1%. Providing intuitive feedback via sound frees the surgeons’ visual attention to remain on the patient, allowing for greater control over diagnostic and surgical tools during surgery, and thus promoting clinical translation of spectroscopic diagnostics.
Original languageEnglish
Pages (from-to)98-109
Number of pages12
JournalAnalyst
Volume142
Early online date12 Oct 2016
DOIs
Publication statusPublished - 7 Jan 2017

Fingerprint

Brain Neoplasms
tumor
Raman scattering
brain
Tumors
Brain
Acoustic waves
Tissue
Surgery
Feedback
Frequency modulation
Glioblastoma
Surgical Instruments
Feature extraction
cancer
Classifiers
probe
Sensitivity and Specificity
tissue
sound

Keywords

  • spectroscopic diagnostics
  • disease states
  • Raman spectroscopy
  • tumour margins
  • metastatic braion cancer
  • glioblastoma
  • endoscopic disgnostics
  • sounds
  • cancer

Cite this

Stables, R., Clemens, G., Butler, H. J., Ashton, K. M., Brodbelt, A., Dawson, T. P., ... Baker, M. (2017). Feature driven classification of Raman spectra for real-time spectral brain tumour diagnosis using sound. Analyst, 142, 98-109. https://doi.org/10.1039/C6AN01583B
Stables, Ryan ; Clemens, Graeme ; Butler, Holly J. ; Ashton, Katherine M. ; Brodbelt, Andrew ; Dawson, Timothy P. ; Fullwood, Leanne M. ; Jenkinson, Michael D. ; Baker, Matthew. / Feature driven classification of Raman spectra for real-time spectral brain tumour diagnosis using sound. In: Analyst. 2017 ; Vol. 142. pp. 98-109.
@article{3fb3816b57b1449ab335bff7e372c1eb,
title = "Feature driven classification of Raman spectra for real-time spectral brain tumour diagnosis using sound",
abstract = "Spectroscopic diagnostics have been shown to be an effective tool for the analysis and discrimination of disease states from human tissue. Furthermore, Raman spectroscopic probes are of particular interest as they allow for in vivo spectroscopic diagnostics, for tasks such as the identification of tumour margins during surgery. In this study, we investigate a feature-driven approach to the classification of metastatic brain cancer, glioblastoma (GB) and non-cancer from tissue samples, and we provide a real-time feedback method for endoscopic diagnostics using sound. To do this, we first evaluate the sensitivity and specificity of three classifiers (SVM, KNN and LDA), when trained with both sub-band spectral features and principal components taken directly from Raman spectra. We demonstrate that the feature extraction approach provides an increase in classification accuracy of 26.25{\%} for SVM and 25{\%} for KNN. We then discuss the molecular assignment of the most salient sub-bands in the dataset. The most salient sub-band features are mapped to parameters of a frequency modulation (FM) synthesizer in order to generate audio clips from each tissue sample. Based on the properties of the sub-band features, the synthesizer was able to maintain similar sound timbres within the disease classes and provide different timbres between disease classes. This was reinforced via listening tests, in which participants were able to discriminate between classes with mean classification accuracy of 71.1{\%}. Providing intuitive feedback via sound frees the surgeons’ visual attention to remain on the patient, allowing for greater control over diagnostic and surgical tools during surgery, and thus promoting clinical translation of spectroscopic diagnostics.",
keywords = "spectroscopic diagnostics, disease states, Raman spectroscopy, tumour margins, metastatic braion cancer, glioblastoma, endoscopic disgnostics, sounds, cancer",
author = "Ryan Stables and Graeme Clemens and Butler, {Holly J.} and Ashton, {Katherine M.} and Andrew Brodbelt and Dawson, {Timothy P.} and Fullwood, {Leanne M.} and Jenkinson, {Michael D.} and Matthew Baker",
year = "2017",
month = "1",
day = "7",
doi = "10.1039/C6AN01583B",
language = "English",
volume = "142",
pages = "98--109",
journal = "Analyst",
issn = "0003-2654",

}

Stables, R, Clemens, G, Butler, HJ, Ashton, KM, Brodbelt, A, Dawson, TP, Fullwood, LM, Jenkinson, MD & Baker, M 2017, 'Feature driven classification of Raman spectra for real-time spectral brain tumour diagnosis using sound', Analyst, vol. 142, pp. 98-109. https://doi.org/10.1039/C6AN01583B

Feature driven classification of Raman spectra for real-time spectral brain tumour diagnosis using sound. / Stables, Ryan; Clemens, Graeme; Butler, Holly J.; Ashton, Katherine M.; Brodbelt, Andrew; Dawson, Timothy P.; Fullwood, Leanne M.; Jenkinson, Michael D.; Baker, Matthew.

In: Analyst, Vol. 142, 07.01.2017, p. 98-109.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Feature driven classification of Raman spectra for real-time spectral brain tumour diagnosis using sound

AU - Stables, Ryan

AU - Clemens, Graeme

AU - Butler, Holly J.

AU - Ashton, Katherine M.

AU - Brodbelt, Andrew

AU - Dawson, Timothy P.

AU - Fullwood, Leanne M.

AU - Jenkinson, Michael D.

AU - Baker, Matthew

PY - 2017/1/7

Y1 - 2017/1/7

N2 - Spectroscopic diagnostics have been shown to be an effective tool for the analysis and discrimination of disease states from human tissue. Furthermore, Raman spectroscopic probes are of particular interest as they allow for in vivo spectroscopic diagnostics, for tasks such as the identification of tumour margins during surgery. In this study, we investigate a feature-driven approach to the classification of metastatic brain cancer, glioblastoma (GB) and non-cancer from tissue samples, and we provide a real-time feedback method for endoscopic diagnostics using sound. To do this, we first evaluate the sensitivity and specificity of three classifiers (SVM, KNN and LDA), when trained with both sub-band spectral features and principal components taken directly from Raman spectra. We demonstrate that the feature extraction approach provides an increase in classification accuracy of 26.25% for SVM and 25% for KNN. We then discuss the molecular assignment of the most salient sub-bands in the dataset. The most salient sub-band features are mapped to parameters of a frequency modulation (FM) synthesizer in order to generate audio clips from each tissue sample. Based on the properties of the sub-band features, the synthesizer was able to maintain similar sound timbres within the disease classes and provide different timbres between disease classes. This was reinforced via listening tests, in which participants were able to discriminate between classes with mean classification accuracy of 71.1%. Providing intuitive feedback via sound frees the surgeons’ visual attention to remain on the patient, allowing for greater control over diagnostic and surgical tools during surgery, and thus promoting clinical translation of spectroscopic diagnostics.

AB - Spectroscopic diagnostics have been shown to be an effective tool for the analysis and discrimination of disease states from human tissue. Furthermore, Raman spectroscopic probes are of particular interest as they allow for in vivo spectroscopic diagnostics, for tasks such as the identification of tumour margins during surgery. In this study, we investigate a feature-driven approach to the classification of metastatic brain cancer, glioblastoma (GB) and non-cancer from tissue samples, and we provide a real-time feedback method for endoscopic diagnostics using sound. To do this, we first evaluate the sensitivity and specificity of three classifiers (SVM, KNN and LDA), when trained with both sub-band spectral features and principal components taken directly from Raman spectra. We demonstrate that the feature extraction approach provides an increase in classification accuracy of 26.25% for SVM and 25% for KNN. We then discuss the molecular assignment of the most salient sub-bands in the dataset. The most salient sub-band features are mapped to parameters of a frequency modulation (FM) synthesizer in order to generate audio clips from each tissue sample. Based on the properties of the sub-band features, the synthesizer was able to maintain similar sound timbres within the disease classes and provide different timbres between disease classes. This was reinforced via listening tests, in which participants were able to discriminate between classes with mean classification accuracy of 71.1%. Providing intuitive feedback via sound frees the surgeons’ visual attention to remain on the patient, allowing for greater control over diagnostic and surgical tools during surgery, and thus promoting clinical translation of spectroscopic diagnostics.

KW - spectroscopic diagnostics

KW - disease states

KW - Raman spectroscopy

KW - tumour margins

KW - metastatic braion cancer

KW - glioblastoma

KW - endoscopic disgnostics

KW - sounds

KW - cancer

U2 - 10.1039/C6AN01583B

DO - 10.1039/C6AN01583B

M3 - Article

VL - 142

SP - 98

EP - 109

JO - Analyst

JF - Analyst

SN - 0003-2654

ER -

Stables R, Clemens G, Butler HJ, Ashton KM, Brodbelt A, Dawson TP et al. Feature driven classification of Raman spectra for real-time spectral brain tumour diagnosis using sound. Analyst. 2017 Jan 7;142:98-109. https://doi.org/10.1039/C6AN01583B