Intraoperative imaging of brain tumors using spectral signatures of tissue, based on injected fluorescent dye such as 5-ALA, has enabled surgeons to target residual malignant tissue near the boundaries of the tumor cavity where extent of resection is most difficult. This paper presents a novel approach to intraoperative tumor boundary detection based on a moving excitation laser crossing a tumor boundary while measuring spectral signatures generated. In prior work, we have characterized the intrinsic spectral signatures of glioblastoma and healthy brain tissue from in vivo mouse models within the 400 to 700 nm range given a 405 nm excitation source at a single spot, without the use of injected dye. In this work, we present a theoretical model of expected spectral signature observations for a moving excitation laser across a tumor boundary based on discretized contribution of known spectral signatures (i.e. GBM, healthy) within the region of the laser spot on the surface of a tissue. This approach allows for improved intraoperative boundary detection despite having a laser spot size larger than the desired resolution of detection.
The removal of tissue with a laser scalpel is a complex process that is affected by the laser incidence angle on the surface of the tissue. Current models of laser ablation, however, do not account for the angle of incidence, assuming that it is always normal to the surface. In order to improve ablation modeling in soft tissue, this work characterizes photoablation crater profiles at incidence angles ranging from 0 degrees to 45 degrees off perpendicular. Simulated results, based on a discretized steady-state ablation model, are generated for comparison based on the assumption that material removal occurs in the direction of the laser. Experiments in an agarose-based, homogeneous soft tissue phantom are performed with a carbon dioxide (CO2) laser. Surface profiles of the craters are acquired using a micro x-ray computed tomography scanner (Micro-CT) and compared to results from the simulation. The difference of the simulated and experimental results are measured and the error analysis is reported.
The ability to differentiate healthy and tumorous tissue is vital during the surgical removal of tumors. This ability is especially critical during neurosurgical tumor resection due to the risk associated with removing healthy brain tissue. In this paper, we present an epifluorescence spectroscopy guided device that is not only capable of autonomously classifying a region of tissue as tumorous or healthy in real-time–but is also able to differentiate between different tumor types. For this study, glioblastoma and melanoma were chosen as the two different tumor types. Six mice were utilized in each of the three classes (healthy, glioblastoma, melanoma) for a total of eighteen mice. A “one-vs-the-all” approach was used to create a multi-class classifier. The multi-class classifier was capable of classifying with 100% accuracy. Future work includes increasing the number of mice in each of the three tumor classes to create a more robust classifier and expanding the number of tumor types beyond glioblastoma and melanoma.