Stereovision-based integrated system for point cloud reconstruction and simulated brain shift validation

Xiaochen Yang, Logan W. Clements, Ma Luo, Saramati Narasimhan, Reid C. Thompson, Benoît M. Dawant, Michael I. MigaJournal of Medical Imaging 2017 · DOI

TL;DR

This paper introduces a novel stereovision-based system that uses a tracked surgical microscope to reconstruct 3D point clouds of the brain surface. This system aims to provide a cost-effective way to measure and compensate for brain shift, a critical challenge in neurosurgery, achieving an average displacement error of approximately 2mm in simulated scenarios.

What Problem Does This Paper Address?

Image-guided surgery (IGS) systems are invaluable for neurosurgeons, but their accuracy is compromised by "brain shift." This phenomenon refers to the non-rigid deformation of brain tissue that occurs during surgery due to factors like fluid drainage, swelling, or tissue manipulation. Current IGS systems struggle to account for these dynamic changes, potentially leading to inaccurate targeting. While solutions like intraoperative MRI, CT, and ultrasound exist, they have drawbacks such as radiation exposure, high cost, or poor image quality. This paper addresses the need for a more practical and cost-effective method to capture real-time 3D surface data of the brain to correct for brain shift.

Methods

The core of this work is an integrated system that combines stereovision with microscope tracking.

  • Stereovision for 3D Reconstruction: The system utilizes stereo images captured by a surgical microscope equipped with two cameras. Standard stereovision techniques, including camera calibration (using a chessboard pattern) and rectification, are employed to generate 3D point clouds. The paper specifically mentions using Bouguet's approach for rectification and a block matching (BM) algorithm for disparity computation, with post-filtering for improved edge alignment.
  • Microscope Tracking: To capture point clouds from different viewpoints and track the microscope's movement, a rigid body tracking star is attached to the microscope. This star is tracked by a commercial optical tracking system (Polaris Spectra). A crucial step is the calibration of the microscope's coordinate system with respect to the optical tracking system's coordinate system, achieved using a calibration phantom with fiducial disks.
  • Simulated Brain Shift: A custom-designed craniotomy simulation device is used to mimic realistic brain deformations. This device allows for in-plane stretching (lateral shift) and out-of-plane deformation (sag) of a silicone membrane, simulating various brain shift scenarios.
  • Data Registration and Validation: Point clouds acquired at different microscope positions are registered into a common coordinate space. The system's accuracy is validated by comparing the measured displacements of simulated vessel features on the craniotomy device to ground truth measurements obtained by an optically tracked stylus.

Datasets

The paper does not explicitly mention training datasets. Instead, it focuses on a simulated environment for evaluation:

  • Calibration Phantom: A stair-block phantom with known geometry is used for initial calibration and reconstruction error estimation.
  • Craniotomy Simulation Device: This novel device, featuring a deformable silicone membrane with mock cortical vessels, is used to generate realistic brain shift scenarios (baseline, stretch, and sag) for validation experiments. The ground truth for vessel displacement is obtained using an optical tracking system.

Key Insights and Ideas

  • Stereovision as a Practical Tool: The research demonstrates that readily available surgical microscopes, when augmented with stereovision capabilities and tracking, can be a practical and cost-effective alternative to more complex systems like laser range scanners for intraoperative surface data acquisition.
  • Integrated System for Dynamic Tracking: The key innovation lies in the seamless integration of stereoscopic imaging for 3D reconstruction with real-time optical tracking of the microscope. This allows for the capture of surface data from multiple, arbitrary microscope positions, which is essential for tracking dynamic tissue deformations.
  • Simulated Environment for Validation: The development of a realistic craniotomy simulation device is crucial. It allows for controlled and repeatable generation of brain shift phenomena, providing a valuable benchmark for evaluating the system's performance without the ethical and logistical complexities of in-vivo studies.
  • Reconstruction and Tracking Errors: The study highlights that errors in both the stereo reconstruction process (e.g., disparity computation) and the microscope tracking system contribute to the overall displacement error. Understanding these error sources is vital for future improvements.

Key Results

  • Reconstruction Error: An initial estimation of reconstruction error using a known geometry phantom yielded approximately 1 mm.
  • Microscope Tracking Calibration: The calibration procedure for the microscope tracking system resulted in a reported tracking accuracy of 0.25-0.3mm RMS for the Polaris Spectra system.
  • Point Cloud Registration Error: When registering point clouds from different microscope positions, the mean distance errors were 0.64 mm (x), 0.89 mm (y), and 2.92 mm (z).
  • Simulated Brain Shift Validation: The system achieved an average displacement error of approximately 2 mm when comparing the stereopsis-based measurement of mock vessel displacements to the ground truth measured by the optically tracked stylus.

Limitations and Future Work

  • Fixed Microscope Settings: The current calibration method is limited to fixed microscope settings (focal length and zoom factor). Future work should aim to generate calibration data that accommodates multiple microscope settings during a procedure.
  • Phantom Limitations: The sharp edges of the calibration phantom might lead to higher reconstruction errors compared to the smoother surface of actual cortical tissue.
  • Error Sources: The authors acknowledge that errors stem from stereo reconstruction, fiducial digitization, phantom design, and tracking system inaccuracies.
  • Future Improvements:
    • Refining reconstruction techniques for more accurate calibrations.
  • Designing an optimally calibrated phantom for improved results.
  • Developing methods to handle varying microscope settings.
  • Extending the system to gather cortical data for measuring and compensating for brain shift in actual image-guided surgery.

Summary updated

Abstract

Intraoperative soft tissue deformation, referred to as brain shift, compromises the application of current image-guided surgery navigation systems in neurosurgery. A computational model driven by sparse data has been proposed as a cost-effective method to compensate for cortical surface and volumetric displacements. We present a mock environment developed to acquire stereoimages from a tracked operating microscope and to reconstruct three-dimensional point clouds from these images. A reconstruction error of 1 mm is estimated by using a phantom with a known geometry and independently measured deformation extent. The microscope is tracked via an attached tracking rigid body that facilitates the recording of the position of the microscope via a commercial optical tracking system as it moves during the procedure. Point clouds, reconstructed under different microscope positions, are registered into the same space to compute the feature displacements. Using our mock craniotomy device, realistic cortical deformations are generated. When comparing our tracked microscope stereo-pair measure of mock vessel displacements to that of the measurement determined by the independent optically tracked stylus marking, the displacement error was [Formula: see text] on average. These results demonstrate the practicality of using tracked stereoscopic microscope as an alternative to laser range scanners to collect sufficient intraoperative information for brain shift correction.
BibTeX
@article{yang2017stereovision,
  author = {Xiaochen Yang and Logan W. Clements and Ma Luo and Saramati Narasimhan and Reid C. Thompson and Benoît M. Dawant and Michael I. Miga},
  title = {{Stereovision-based integrated system for point cloud reconstruction and simulated brain shift validation}},
  journal = {Journal of Medical Imaging},
  year = {2017},
  doi = {10.1117/1.jmi.4.3.035002},
}
Figure