Thomas J.E. Hubbard, PhD1,2; Alexander P. Dudgeon, PhD2; C Murray; Douglas J. Ferguson1,2; Angela C. Shore1,2; Nicholas Stone1,2
1Royal Devon and Exeter Hospital, Exeter, UK
2University of Exeter Medical School, Exeter, UK
Background
Accurate intraoperative margin analysis (IMA) would reduce the need for surgical re-excision in breast-conserving surgery (BCS). High-wavenumber Raman spectroscopy (HWN RS) is a vibrational spectroscopy highly sensitive to changes in the protein and lipid environment and water content–biochemical differences found between tumor and normal breast tissue. We determined the diagnostic accuracy of a HWN RS system to differentiate between tumor and nontumor breast tissue for future IMA.
Materials and Methods
Thawed fresh-frozen human breast specimens were obtained with ethical approval from the Exeter Clinical Research Facility Tissue Bank (CRF Ref: CRF320; Tissue bank ethics number 16/ SC/0162). Paired biopsy specimens of tumor and nontumor tissue were measured with a handheld probe with a 785 nm laser excitation and a spectrometer with an indium-gallium-arsenide, or InGaAs, camera. Sections from specimens underwent histopathological diagnosis and Raman microspectrometry measurement with a Renishaw inVia system. Spectra were analyzed with a principal component analysis (PCA)–fed linear discriminant analysis (LDA) model, with leave-one-out-cross-validation (LOOCV) and calculation of the water/total HWN area ratio (W/TAR); a diagnostic threshold was determined by receiver operating characteristic curve analysis from a binomial logistic regression with k-fold 5 cross-validation.
Results
Ninety-six patients were included (invasive ductal carcinoma, n = 64; invasive lobular carcinoma, n = 21; ductal carcinoma in situ, n = 3; other carcinoma, n = 8). Matching of histopathology to HWN microspectrometry Raman spectra confirmed the spectral differences between tumor (protein, water-rich) and nontumor (lipid-rich, water-poor) tissue. Handheld probe spectra PCA analysis revealed the changes in protein, lipid, and water content were predominant spectral features, accounting for greater than 95% variance. The LDA analysis demonstrated an accuracy of 93.2% (sensitivity 93.8%; specificity 92.7%), and W/TAR analysis with a diagnostic threshold of 0.66 gave an overall accuracy of 92.2% (sensitivity 94.8%; specificity 89.6%) for differentiating between tumor and nontumor tissue. Diagnostic accuracy was unaffected by tissue staining with patent blue dye for concurrent sentinel lymph node biopsy, indicating its potential for intraoperative analysis.
Conclusions
We present an emerging technology that can differentiate between tumor and nontumor breast tissue with 93.2% accuracy. This system has the potential to provide IMA in BCS and reduce re-excision rates.