Diffuse Reflectance Spectroscopy for Intraoperative Tumor Margin Assessment

Abstract

The aim of this graduation project is to assess the feasibility of the photonic technology called diffuse reflectance spectroscopy (DRS) to aid in intraoperative margin assessment with the goal of obtaining clear margins for liver, breast and soft tissue tumors. The idea is to integrate optical fibers in the existing and widely used electrosurgical knife, which consists of a pencil holder and a blade electrode. The work consists of workflow analyses and experiments. Based on the former the role of DRS in clinical practice can be specified. The main object of the experiments is to find out if DRS is still usable when a layer of coagulated tissue due to electrosurgical activity is present. Lumpectomy workflows using four techniques are investigated. For palpable tumors intraoperative ultrasound (US) and palpation are used, while non-palpable tumors require iodine seeds or placement of a guidewire. The most important limitations of current methods include: no information about the actual margin for iodine seeds and guidewire procedures, complex patient preparation, expertise required to understand US images and lack of reinforced learning for palpation. One advantage of all techniques is that they provide extra information, allowing the surgeon to create a more detailed mental image of the tumor. Ultrasound provides continuous information on the margin, while palpation is very intuitive. DRS integrated into the electrosurgical instrument provides distinct benefits in the local assessment phase, including direct feedback on whether or not to make a cut without losing information. DRS sensing should be directed forward, parallel to the axis of the pencil. In the experimental phase the shape and size of the tissue area influenced by electrosurgical treatment is inves- tigated first. It is found that layers of coagulation up to 1.4 mm may occur in surgery. Next, experiments are performed to identify the effect of coagulation on DRS spectra. Protein denaturation and reduced water con- tent are clearly visible in the diffuse reflectance spectra, resulting in a decreased slope and increased intensities of peaks in the near-infrared (NIR) range. Spectral alterations are severe, requiring further investigation using Monte Carlo (MC) simulations. Two-layered MC models that consist of coagulated tissue on top of normal tissue are build for coagulation up to 1.5 mm. Using the slope and peaks described above a predictive model is created, based on a sigmoid fit of these so-called predictors. Coagulation depth is predicted with R2 values up to 0.99, indicating a valid model. Two-layered MC models of normal and tumor tissue show that breast tumors can be detected up to 4 mm, liver tumors are visible at 2 mm and lipoma in skeletal muscle can be seen at 5 mm when the fiber distance (FD) = 6 mm. Finally, three-layered MC models of coagulated, normal and tumor tissue yield predictors that strongly differ from tissue without tumor presence. No predictive model for tumor depth is created. The results from this project show that accurate depth determination of coagulation is possible as well as sensing the presence of a tumor within the sensing zone. The latter ensures that DRS may bring at least the same qual- ities as iodine seeds and the guidewire. An important benefit is that DRS provides information on the complete margin, rather than a discrete point. Quantitative depth prediction is required to compete with intraoperative US for invasive tumors. Overall, it was shown that if DRS is successfully developed it can bring important benefits to clinical practice. The effect of tissue coagulation on diffuse reflectance spectra is large, but can be accounted for with a mathemat- ical predictive model. The next crucial step is further investigation of tumor depth sensing and investigation of a complete predictive model. The depth sensing requirements are only completely met for breast tumors. For liver resection, tumors are only detected up to 2 mm. This means that the desired 1 cm margin cannot be safeguarded with DRS. Close margins however are often accepted in clinical practice and do improve patient outcome. DRS integration should not be considered as a replacement for current margin assessment techniques but rather as a complement. Overall, integration of DRS into electrosurgical instruments seems promising. More research is required to see if DRS can truly perform in the OR.