Intra-operative Detection of Positive Margins and Lymph Nodes in Breast Surgery

Overview

In this project, the investigators will develop novel optical coherence tomography (OCT)-Raman spectroscopy and autofluorescence (AF)-Raman spectroscopy systems based on a selective sampling approach optimised for high-resolution analysis of whole lumpectomy specimens and sentinel lymph node (SLN) biopsies, respectively. The aim of using optical coherence tomography is not to detect cancer directly, but rather to identify adipose tissue so that large adipose regions can be excluded from subsequent Raman spectroscopy measurements. Although OCT has limited ability to distinguish tumour tissue from the surrounding normal stroma, adipose tissue exhibits a distinctive appearance in optical coherence tomography images because of its low backscattering properties, resulting from adipocytes that are filled with lipids and contain small, flattened nuclei. In contrast, benign dense tissue (stroma, ducts, and lobules) and malignant tissue produce much stronger backscattering signals. These characteristic patterns enable adipose tissue to be distinguished from other breast tissues using classification models based on optical coherence tomography reflectivity profiles, achieving 94% sensitivity and 93% specificity. Excluding adipose tissue from further analysis reduces the number of Raman spectroscopy measurements required, allowing the remaining, smaller tissue regions to be examined to discriminate between benign and malignant tissue. This flexible and adaptable scanning strategy is expected to improve both diagnostic accuracy and scanning speed, enabling complete assessment of surgical margins within clinically practical timescales. In addition, the investigators will develop a novel (AF)-Raman spectroscopy system based on a selective sampling approach optimised for high-resolution analysis of sentinel lymph node specimens. The purpose of incorporating autofluorescence imaging is to identify the optimal sampling locations for subsequent Raman spectroscopy measurements, thereby improving the efficiency of tissue interrogation while maintaining diagnostic accuracy.

The new optical coherence tomography (OCT)-Raman spectroscopy system developed in this project will integrate both modalities into a single instrument and employ deep learning algorithms for automated data acquisition and analysis. The OCT module will be designed for rapid scanning of large lumpectomy specimens, including automatic focus adjustment for irregular three-dimensional tissue surfaces. Machine learning (ML) algorithms will identify regions of interest (non-adipose tissue) in the OCT images and automatically direct Raman spectroscopy measurements to these high-risk areas. A second layer of machine learning models will then classify the Raman spectra to distinguish cancerous tissue (positive margins) from benign tissue. This integrated approach simplifies operation, reduces user subjectivity, and minimises training requirements. The user will only need to place the specimen into the instrument, after which all subsequent steps-including OCT imaging, Raman spectroscopy measurements, data analysis, and image reconstruction-will be performed automatically. The final output will be a diagnostic map highlighting any positive surgical margins in red. By combining rapid OCT imaging with the molecular specificity of Raman spectroscopy, the system aims to translate the high diagnostic accuracy of Raman spectroscopy from millimetre-scale sampling to whole-specimen assessment, providing surgeons with a practical tool for intra-operative margin evaluation. The OCT-Raman device used in this study has been developed by the University of Nottingham. This is a single-centre proof-of-concept study of an in-house developed device. The results generated will be used solely to evaluate the performance of the device and will not be used to direct or influence participants' clinical care. In addition, the project will be extended to include the analysis of sentinel lymph nodes using an autofluorescence (AF)-Raman spectroscopy system. This system integrates autofluorescence imaging and Raman spectroscopy into a single device for lymph node assessment, combining the high imaging speed and spatial resolution of autofluorescence with the molecular specificity of Raman spectroscopy. Data acquisition and analysis will be performed using automated deep learning algorithms. The University of Nottingham team has previously demonstrated this concept in an autofluorescence-Raman spectroscopy instrument developed for detecting positive margins during Mohs micrographic surgery for skin cancer. In a proof-of-concept study conducted at Nottingham University Hospitals NHS Trust, the device achieved greater than 95% sensitivity and greater than 95% specificity, with total scanning times of 20-30 minutes, while preserving tissue integrity for subsequent histopathological examination. Once the system has been calibrated and trained to distinguish tumour tissue from normal tissue, the surface of each specimen will be scanned without direct handling of the tissue before being returned to the pathologist for routine clinical processing. The tissue specimens used for clinical diagnosis will not be used for research purposes. Any identifiable patient information will be accessible only to members of the clinical care team, and all samples will remain fully anonymized to researchers who are not involved in the participants' clinical care.