In this project, we will develop a unique OCT-Raman system based on a selective sampling approach optimised for high-resolution analysis of whole lumpectomy specimens. The aim of using OCT is not to detect the cancer but to identify the adipose tissue, such that the large adipose tissue regions are excluded from any further measurements by Raman spectroscopy.

While OCT has a limited ability to distinguish between tumour and surrounding normal stroma, adipose tissue has a distinctive appearance in the OCT images due to low backscattering within adipose cells (filled with lipids and small/flattened nuclei) compared to the highly scattering benign dense tissue (stroma, ducts and lobules) and malignant tissue. Such specific patterns allow identification of normal adipose tissue from breast tissue (classification models based on reflectivity profiles) with 94% sensitivity and 93% specificity. This will reduce the task of Raman measurements, which can be focused on the smaller remaining regions to discriminate between the benign and malignant tissue. This flexible and adaptable scanning strategy will achieve a much-improved diagnosis accuracy and speed to cover all surgical margins within practical timescales.

The new OCT-Raman system developed in this project will integrate both modules into a single instrument and rely on deep learning algorithms to automatically acquire and analyse data. The OCT module will be designed for fast scanning large lumpectomy specimens (include focus adjustment for irregular 3D surfaces) and machine learning (ML) algorithms will identify the regions of interest (non-adipose tissue" in the OCT images, automatically directing the Raman spectroscopy measurements to these "high-risk areas". A second layer of machine learning models will then classify the Raman spectra to discriminate the cancer (positive margins) from benign tissue. This approach simplifies the use of the instrument, reduces subjectivity and user training: the user will be required only to insert the tissue in the instrument, all steps being afterwards automated (OCT and Raman measurements and analysis) until the display of the final diagnosis map showing any positive margins in red colour. The unique OCT-Raman system will translate the high diagnosis accuracy of Raman spectroscopy from mm-scale to whole lumpectomy level, providing a tool for surgeons to identify positive margins intra-operatively. The Uon team has demonstrated this concept in an instrument based on AF and Raman spectroscopy for the detection of positive margins during Mohs micrographic surgery for skin cancers. The OCT-Raman device that will be used in this study has been developed by the University of Nottingham. This is a proof-of-concept study of a device developed in-house, being undertaken at a single centre only, and the results generated from using the device will not be used to direct or influence the participant's clinical care.

When the machine is already calibrated and trained to differentiate between tumour and normal tissue, we will scan the surface of the wide local excision specimen without handling of the tissue, and return it to the pathologist for routine processing. The tissue slice will not be used for any research purposes. Any identifiable information will only be accessed by members of the clinical care team, and the samples will remain anonymous to the researchers who are not members of the clinical care team.

Inclusion criteria

• Patients undergoing breast surgery (wide local excision).
• Able to give informed consent.
• Any age.

Exclusion criteria

• Patients where there is any doubt regarding the diagnosis from pathologist as ascertained by previous diagnostic biopsy.

Patients undergoing WLE in the breast cancer institute