#1512

In-vivo Fibreoptic Raman Spectroscopy for Rapid Bladder Cancer Diagnosis

W. Tsang¹, A. Kesavan¹, K. Wang¹, T. Law¹, C. Yu², C. Shu², C. Liu², S. Tham¹, T. Thamboo³, Z. Huang², E. Chiong⁴

¹National University Hospital, Urology, Singapore, Singapore
²National University of Singapore, Biomedical Engineering, Singapore, Singapore
³National University of Singapore, Pathology, Singapore, Singapore
⁴National University of Singapore, Surgery, Singapore, Singapore

Introduction:

Diagnosis of bladder cancer is typically done with white light cystoscopy, which has limited sensitivity particularly in detecting flat carcinomas-in-situ. Raman spectroscopy is a label-free optical spectroscopy technique based on Raman scattering, using characteristic molecular vibrational fingerprints to elucidate structural composition of tissues and cells. This study aims to assess the diagnostic utility of in-vivo fibreoptic Raman spectroscopy in bladder cancer detection.

Material and methods:

A customised fibreoptic Raman probe was created to fit the working channel of a rigid cystoscope. During transurethral resection of bladder tumour (TURBT), the probe was placed through the cystoscope to acquire data regarding Raman spectra of bladder tissue in vivo. A proprietary Raman spectral processing framework was designed and programmed to display the processed Raman signals in real time. Histopathological data of the bladder sites evaluated were also collected and correlated with the Raman spectroscopy results.

Results:

A total of 1151 bladder tissue Raman spectra were acquired in vivo from 93 bladder tissue sites in 24 patients (19 male and 5 female patients) with a mean age 68 years. Of the 93 tissue sites evaluated, 46 were found to be malignant on histology, while 47 were benign. Significant tissue Raman spectral differences were observed between malignant and benign tissue sites (Figure 1), indicating specific biomarkers expressed in urothelial carcinoma. Using partial least-squares linear discriminant analysis (PLS-LDA), a 2-class diagnostic model was developed to predict for the presence of bladder cancer. The model was able to achieve 80.43% sensitivity and 87.23% specificity, with an overall diagnostic accuracy of 83.87%.