A Real-Time Hyperspectral Laparoscopic Stereo Imaging System for Robot-Assisted Surgery

Faster recovery, reduced hospital stays, and a quicker return to normal activities are the most evident advantages of laparoscopic surgery. More than 13 million laparoscopic procedures are performed every year. Laparoscopic surgery has become the technique of choice for virtually every kind of abdominal surgery. Robot‐assisted laparoscopic radical

prostatectomy is the most common surgical approach performed for prostate cancer. However, the current oncological outcome (undetectable prostate‐specific antigen with no evidence of cancer recurrence) and functional outcome

(recovery of continence and erectile function) are only achieved in 62–70% of prostate cancer patients. Unfortunately,

the endoscope used in current laparoscopic robotic surgery does not allow for the intraoperative assessment of tissue to determine if residual tumors remain. Because of positive tumor margin, cancer recurrence inevitably leads to the need for further treatments which are expensive as well as morbid. Though fluorescence imaging via indocyanine green has

provided insights into tissue perfusion and improved tissue contrast during laparoscopic surgery, its drawbacks include a

short half‐life in the bloodstream and the requirement for contrast agent injection. As a result, there is an unmet need to

develop label‐free imaging techniques that can detect cancerous tissue in real time during laparoscopic robotic surgery.

The objective of the proposed research is to develop a label‐free hyperspectral laparoscopic stereo imaging system for

robotic surgery. Hyperspectral imaging (HSI) provides not only high‐resolution spatial images but also spectral data at each

pixel. Both spectral and spatial information can be used to identify various types of tissues including malignant tumors.

This research is based on the success of using our large‐size HSI system in both fresh surgical specimens of human tissue

and in vivo animals. With our advanced machine learning methods, we achieved a sensitivity of 88.9% and a specificity of 97.1% for cancer detection in fresh surgical tissue specimens of more than 200 human patients. However, current HSI

systems are too bulky and their image acquisition speed is too slow for intraoperative use. In this project, we propose to integrate real‐time snapshot HSI cameras with a 3D stereo endoscope and test the system in the prostates of 93 human patients (Aim 1). We will develop and evaluate advanced deep learning algorithms to automatically detect malignant

tissues on hyperspectral images (Aim 2). We will integrate our hyperspectral endoscopic stereo imaging system with a da

Vinci robot from Intuitive Surgical and perform robot‐assisted laparoscopic surgery in a preclinical study (Aim 3). The new

capabilities of the proposed hyperspectral laparoscopic imaging system include: i) the capability to differentiate malignant

from benign tissue during surgery and ii) the capability to map tissue oxygen saturation (StO2), near‐infrared perfusion

index (NIR‐PI), organ hemoglobin index (OHI), and tissue water index (TWI). In this research, the utilization of quantitative

hyperspectral imaging for robot‐assisted laparoscopic surgery represents a major innovation in minimally invasive surgery. The academic‐industrial partnership among UT Southwestern, UT Dallas, and Intuitive Surgical will work in a strategic

alliance and utilize unique resources at each site to ensure the translation of the hyperspectral laparoscopic stereo imaging technique to the commercial da Vinci robot for broad applications and improved surgical outcomes.