Dense 3D Reconstruction Through Lidar: A Comparative Study on Ex-vivo Porcine Tissue

TL;DR

This study addresses the need for real-time, accurate intraoperative 3D perception by comparing a compact lidar ToF sensor with a state-of-the-art endoscopic stereo system on fresh ex-vivo porcine tissue. Ground-truth 3D surfaces from an Artec scanner are used to quantify signed depth errors through a rigorous camera-to-world registration and post-processing pipeline, including an ART ANOVA statistical framework. The results show lidar offers higher precision, lower latency, and greater robustness to distance and illumination than stereo endoscopy, though tissue-type dependent depth offsets (notably a few millimeters for muscle) emerge. The work highlights lidar as a valuable complement to stereo imaging in a multi-view, real-time reconstruction framework and points toward integrating spectral imaging to compensate tissue-dependent effects.

Abstract

New sensing technologies and more advanced processing algorithms are transforming computer-integrated surgery. While researchers are actively investigating depth sensing and 3D reconstruction for vision-based surgical assistance, it remains difficult to achieve real-time, accurate, and robust 3D representations of the abdominal cavity for minimally invasive surgery. Thus, this work uses quantitative testing on fresh ex-vivo porcine tissue to thoroughly characterize the quality with which a 3D laser-based time-of-flight sensor (lidar) can perform anatomical surface reconstruction. Ground-truth surface shapes are captured with a commercial laser scanner, and the resulting signed error fields are analyzed using rigorous statistical tools. When compared to modern learning-based stereo matching from endoscopic images, time-of-flight sensing demonstrates higher precision, lower processing delay, higher frame rate, and superior robustness against sensor distance and poor illumination. Furthermore, we report on the potential negative effect of near-infrared light penetration on the accuracy of lidar measurements across different tissue samples, identifying a significant measured depth offset for muscle in contrast to fat and liver. Our findings highlight the potential of lidar for intraoperative 3D perception and point toward new methods that combine complementary time-of-flight and spectral imaging.

Figures (14)

• Figure 1: Fresh ex-vivo porcine tissue being imaged by both a standard stereo endoscope and a rigidly attached compact lidar (light detection and ranging) camera. The inset images show the depth maps seen by each camera. This sensor setup was used to capture all study data.
• Figure 2: Ex-vivo porcine tissue samples. a) Abdomen: overall, bloody. b) Abdomen: fatty tissue, bloody. c) Abdomen: muscular tissue, clean. d) Liver: smooth surface, clean. e) Liver: gallbladder, bloody. f) Liver: overall, bloody.
• Figure 3: Experimental setup. a) Main components rigidly attached to the da Vinci robot's camera arm: electromagnetic (EM) sensor, stereo endoscope, and lidar. b) Overall hardware setup. c) Lidar electronics exposed to highlight the infrared (IR) transmitter and receiver, with a five-euro-cent coin for size reference.
• Figure 4: Kinematic transformations (arrows in light blue) between the camera ($\mathcal{F}_{\mathtt{C}}$), the robotic arm ($\mathcal{F}_{\mathtt{A}}$), the rotating tray ($\mathcal{F}_{\mathtt{T}}$), and the GT scan ($\mathcal{F}_{\mathtt{O}}$), with each frame illustrated in red. The center of the illustration also shows a sample manual pairing of all the visible plastic pins selected on the GT scan ($_{\mathtt{O}}\mathbf{p}_i$) and on a corresponding measured point cloud ($_{\mathtt{C}}\mathbf{p}_i$).
• Figure 5: Point-cloud post-processing steps (Section \ref{['metrics']}) applied to a sample lidar frame. a) Raw colored point cloud. b) Signed error field as distance from ground truth. c) Error averaged across multiple static frames. d) Mask to equalize the view field across cameras. e) Mask to extract tissue-specific points. f) Error values pooled (averaged) across space to reduce the resolution for statistical analysis. Point clouds from the stereo endoscope undergo the same process.