TY - JOUR
T1 - Discrimination of malignant and normal kidney tissue with short wave infrared dispersive Raman spectroscopy
AU - Haifler, Miki
AU - Pence, Isaac
AU - Sun, Yu
AU - Kutikov, Alexander
AU - Uzzo, Robert G.
AU - Mahadevan-Jansen, Anita
AU - Patil, Chetan A.
N1 - Publisher Copyright:
© 2018 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
PY - 2018/6
Y1 - 2018/6
N2 - Renal mass biopsy is still controversial due to imperfect accuracy. Raman spectroscopy (RS) demonstrated promise as an in vivo real-time, nondestructive diagnostic tool in many malignancies. Short wave infrared (SWIR) RS has the potential to improve on previous RS systems for renal mass diagnosis. The aim of this study is to evaluate a SWIR RS system in differentiating normal and malignant renal samples. Measurements were acquired using a benchtop RS system with excitation wavelength at 1064 nm and an InGaAs array detector. Processed spectra were classified with a Bayesian machine learning algorithm, sparse multinomial logistic regression. Sensitivity and receiver operating characteristic curve analyses evaluated the classifier accuracy. Accuracy of the classifier was 92.5% with sensitivity and specificity of 95.8% and 88.8%, respectively. For posterior probability of malignant class assignment, the area under the ROC curve is 0.94 (95% confidence interval: 0.89-0.99, P <.001). SWIR RS accurately differentiated normal and malignant kidney tumors. RS has the potential to be used as a diagnostic tool in kidney cancer.
AB - Renal mass biopsy is still controversial due to imperfect accuracy. Raman spectroscopy (RS) demonstrated promise as an in vivo real-time, nondestructive diagnostic tool in many malignancies. Short wave infrared (SWIR) RS has the potential to improve on previous RS systems for renal mass diagnosis. The aim of this study is to evaluate a SWIR RS system in differentiating normal and malignant renal samples. Measurements were acquired using a benchtop RS system with excitation wavelength at 1064 nm and an InGaAs array detector. Processed spectra were classified with a Bayesian machine learning algorithm, sparse multinomial logistic regression. Sensitivity and receiver operating characteristic curve analyses evaluated the classifier accuracy. Accuracy of the classifier was 92.5% with sensitivity and specificity of 95.8% and 88.8%, respectively. For posterior probability of malignant class assignment, the area under the ROC curve is 0.94 (95% confidence interval: 0.89-0.99, P <.001). SWIR RS accurately differentiated normal and malignant kidney tumors. RS has the potential to be used as a diagnostic tool in kidney cancer.
KW - Raman spectroscopy
KW - biopsy
KW - renal cell carcinoma
KW - tissue diagnosis
UR - http://www.scopus.com/inward/record.url?scp=85047610045&partnerID=8YFLogxK
UR - https://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=purepublist2023&SrcAuth=WosAPI&KeyUT=WOS:000434641700001&DestLinkType=FullRecord&DestApp=WOS
U2 - 10.1002/jbio.201700188
DO - 10.1002/jbio.201700188
M3 - Article
C2 - 29411949
SN - 1864-063X
VL - 11
JO - Journal of Biophotonics
JF - Journal of Biophotonics
IS - 6
M1 - e201700188
ER -