Tissue Identification and Photoablation Therapy Control with Contact Diffuse Reflectance Spectroscopy

Abstract

This work explores the capability of contact diffuse reflectance spectroscopy (CDRS) to classify tissue within the coronary arteries as well as control excimer laser coronary atherectomies. The central hypothesis is a transvascular-compatible, fiber optic diffuse reflectance sensing system can distinguish and identify tissue types at the point of therapy, within 25-ms, and while in contact with the tissue layer, therefore enabling precise control of therapy delivery during a surgical intervention. A CDRS based classifier trained to differentiate two-colored synthetic optical phantoms was used to control a clinical excimer laser. To simulate an easy ablation procedure, five procedure experts were given the task of lasing only one color of the phantom. Each user ran through the procedure four times, twice with classifier shared control and twice without. The results show a statistically significant decrease in errors when using CDRS shared control. Demonstrating effective laser control and decreased errors for a simple procedure suggests CDRS shared control in more difficult procedures would lead to a more significant improvement to ablation safety and performance. A custom CDRS collection system for the coronary arteries was also designed and built for ex vivo tissue data collection. Using liquid phantoms and Monte Carlo photon simulations, the effect of the data collection environment, specifically the optical background upon which the arteries are fixed was measured as a function of tissue thickness. Using this data, the CDRS data collection system was tuned to better emulate the in vivo background environment. The simulation data also was used to preliminarily explore the boundary detection capability of CDRS during a simulated excimer laser atherectomy procedure. The same data collection system, now modified with the in vivo-like background was used to take CDRS from plaque and artery within the coronary arteries of six tissue donors. This data was used to determine the CDRS variability within and between donors. The data was also used to test classification performance. The classifiers trained with the CDRS data were able to achieve high area under the curve (AUC) scores (0.94-0.97) across all donors, contact pressures, and plaque layer thicknesses. These combined results support my hypothesis. Further development of a system with these elements might lead to a transvascular-compatible surgical instrumentation and methods that identify tissue in real-time and inform surgical therapy in order to decrease procedural adverse events and increase therapy effectiveness.