Abstract:We present an extension to the widely used spectral angle metric, calculating an azimuthal angle around a reference vector. We demonstrate that it provides additional information, thus improving the classification ability of the spectral angle.

This paper details the development of a camera system that is sensitive to the Blood Oxygen Level Dependant (BOLD) signal for intraoperative delineation of function. The results to date show strong indications that the optical interrogation of this signal is possible in real-time and with minimal change to operating practices and the operating theatre environment. © BackgroundCancer of the brain and CNS account for only 2% of new cases in the UK however it is responsible for 7% of cancer deaths before the age of 70 [1]. Although surgery for cancer falls short of a cure it is the primary method of treatment, the prognosis of surgery is correlated with the extent of resection. This is often limited by the risk of postoperative neurological deficit. This paper reports on a project to reduce this risk by developing an intraoperative tool to detect eloquent areas of the brain. Although functional Magnetic Resonance Imaging (fMRI) and Positron Emission Tomography (PET) are used preoperatively, intraoperative delineation relies on electrocortical stimulation or measuring evoked potentials. These intraoperative techniques are inherently low resolution and take a considerable amount of time to construct a map of the operating field.FMRI uses the differing magnetic properties of oxygenated and deoxygenated haemoglobin to infer function via the haemodynamic response. It is possible to interrogate the oxy/deoxy haemoglobin ratio optically, this method is often referred to as optical imaging of intrinsic signals, this was first observed by [2]. Functional mapping using optical imaging has been investigated for some time, however has not developed sufficiently clinical application. The problems limiting its application as identified in [3] include the limited signal to noise (SNR) principally due to the variation of the signal with pulse and respiratory rate and registration difficulties due to patient movement. This project has developed techniques to overcome these problems. ResultsWe have developed a camera that attaches to a Ziess operating microscope which minimizes the camera movement during surgery. This camera is sensitive to 610nm with a 10nm spectral bandwidth. The camera operates at frame-rates high enough to recover the pulse rate and is therefore immune to degradation of the SNR from this source. Using a high spatial resolution reduces the registration errors and hence further increases the signal to noise. Initial results taken at a low frame-rate (about 2Hz) demonstrate a high correlation with function as shown in Figure 1.

Sentinel Lymph Node Biopsy (SLNB) is an increasingly standard procedure to help oncologists accurately stage cancers. It is performed as an alternative to full axillary lymph node dissection in breast cancer patients, reducing the risk of longterm health problems associated with lymph node removal. Intraoperative analysis is currently performed using touchprint cytology, which can introduce significant delay into the procedure. Spectral imaging is forming a multi-plane image where reflected intensities from a number of spectral bands are recorded at each pixel in the spatial plane. We investigate the possibility of using spectral imaging to assess sentinel lymph nodes of breast cancer patients with a view to eventually developing an optical technique that could significantly reduce the time required to perform this procedure. We investigate previously reported spectra of normal and metastatic tissue in the visible and near infrared region, using them as the basis of dummy spectral images. We analyse these images using the spectral angle map (SAM), a tool routinely used in other fields where spectral imaging is prevalent. We simulate random noise in these images in order to determine whether the SAM can discriminate between normal and metastatic pixels as the quality of the images deteriorates. We show that even in cases where noise levels are up to 20% of the maximum signal, the spectral angle map can distinguish healthy pixels from metastatic. We believe that this makes spectral imaging a good candidate for further study in the development of an optical SLNB.

Sentinel Lymph Node Biopsy (SLNB) is an increasingly standard procedure to help oncologists accurately stage cancers. It is performed as an alternative to full axillary lymph node dissection in breast cancer patients, reducing the risk of longterm health problems associated with lymph node removal. Intraoperative analysis is currently performed using touchprint cytology, which can introduce significant delay into the procedure. Spectral imaging is forming a multi-plane image where reflected intensities from a number of spectral bands are recorded at each pixel in the spatial plane. We investigate the possibility of using spectral imaging to assess sentinel lymph nodes of breast cancer patients with a view to eventually developing an optical technique that could significantly reduce the time required to perform this procedure. We investigate previously reported spectra of normal and metastatic tissue in the visible and near infrared region, using them as the basis of dummy spectral images. We analyse these images using the spectral angle map (SAM), a tool routinely used in other fields where spectral imaging is prevalent. We simulate random noise in these images in order to determine whether the SAM can discriminate between normal and metastatic pixels as the quality of the images deteriorates. We show that even in cases where noise levels are up to 20% of the maximum signal, the spectral angle map can distinguish healthy pixels from metastatic. We believe that this makes spectral imaging a good candidate for further study in the development of an optical SLNB.

This poster presentation will report on all these projects in action with updated results at time of publication.