Patient motion tracking and compensation in IGS
A major focus of my research targets patient motion tracking. To my knowledge, there is very little data published on this issue (e.g. Westermann2000) This post is a short introduction to the topic to raise interest and to give a general overview. If you know of more than the referenced papers below, please let me know!Image-guided surgery requires trackable markers, used as references. The event of patient motion occurs when the body’s position moves relative to the base frame of the device executing the surgical plan. The fundamental problem with patient motion is that without proper identification and compensation, the whole surgical plan may be obsolete, and the treatment potentially harmful. From the clinical point of view, maximum a few mm of error could be tolerated, depending the speed of the tool, it might mean 0.5–2 s delay. If it is noticed in time, re-registration is recommended to avoid damaging the patient. However, re-registration is usually time consuming (and might be cumbersome), therefore it should be avoided, whenever possible. From the technical point of view, many sources of errors can be represented as patient motion. The main sources of external (i.e., excluding physiological) patient motion during surgery include:
- large forces applied by surgeon (e.g., bone milling),
- bumping into the operating table,
- leaning against the patient,
- inadequate fixation,
- equipment failure.
Dynamic correction for unforseen events with the use of typically deployed intraoperative navigation systems remains a significant challenge. While significant effort has been invested to describe the surgical workflow with mathematical models [Jannin2007, Lahiri2010], relatively few projects have dealt with the modeling of the OR setup and environment in general.
The robot’s position information and the tracking data must be kept consistent throughout the operation, especially in the case of neurosurgical or orthopedic procedures, where the accuracy is absolutely crucial. Practically, this can be achieved with a rigid mechanical fixation between the device and the patient. Smaller robots, such as the SmartAssist (Mazor Surgical Technologies Inc., Caesarea, Israel) [Plaskos2005] or the Mini Bone-Attached Robotic System (MBARS, ICAOS and Carnegie Mellon University, U.S.) [Wolf2005] may be bone-mounted. This requires more invasive fixation on the patient side (bone screws), and large forces may still cause relative motion between the patient and the tool. In orthopedics, there are significant contact forces, making it necessary to use stronger screws.
Employing a large, powerful robot may lead to serious tissue damage. The ROBODOC system (Curexo Technology Corporation, Fremont, CA) [Kazanzides2008] was the first FDA approved automated bone milling robot for hip replacement, bone screws and a bone motion sensor to detect fixation failures. If the bone moves more than 2 mm despite the fixation, the system halts, and calls for re-registration.
One option to reduce tissue trauma is to use multiple dynamic reference bases to follow the motion of the robot base and the patient separately. Unfortunately, not every tracking system supports this, and it may cause difficulties to maintain the line-of-sight without disturbing the physician. Extending the active workspace of a tracking system may result in higher inherent accuracies due to the inhomogeneity of its field. Some commercially available systems combine surface-mounted and in-body fiducials to track external and physiological organ motion, though it requires a separate operation placing the markers. A successful example is the CyberKnife radiation therapy system (Accuray Inc., Sunnyvale, CA) that can track skin motion through a special suit and organ motion by taking bi-plane x-ray images and locating fiducials (gold beads) that were implanted pre-operatively [Saito2009]. Other groups tried different filtering approaches with limited effectiveness [Baron2010].
Robotic setups could incorporate accelerometers and gyroscopes, primarily to detect sudden changes; however these require electrical coupling and their resolution might not be sufficient for proper compensation. Besides, these would increase the costs and complexity of the system. Charge-coupled device (CCD) cameras can survey the OR, and image processing techniques could solve the localization problem, but the resolution may not be high enough, and it might have significant hardware requirements.Dynamic registration and correction for patient motion has been implemented with PET/SPECT scans [Fulton2002, Bruyant2005, Rahmim2007] to improve image quality through compensated reconstruction. However, these setups only considered rigid environment, where neither the camera, nor the PET gantry move.
The robot’s position information and the tracking data must be kept consistent throughout the operation, especially in the case of neurosurgical or orthopedic procedures, where the accuracy is absolutely crucial. Practically, this can be achieved with a rigid mechanical fixation between the device and the patient. Smaller robots, such as the SmartAssist (Mazor Surgical Technologies Inc., Caesarea, Israel) [Plaskos2005] or the Mini Bone-Attached Robotic System (MBARS, ICAOS and Carnegie Mellon University, U.S.) [Wolf2005] may be bone-mounted. This requires more invasive fixation on the patient side (bone screws), and large forces may still cause relative motion between the patient and the tool. In orthopedics, there are significant contact forces, making it necessary to use stronger screws.
Employing a large, powerful robot may lead to serious tissue damage. The ROBODOC system (Curexo Technology Corporation, Fremont, CA) [Kazanzides2008] was the first FDA approved automated bone milling robot for hip replacement, bone screws and a bone motion sensor to detect fixation failures. If the bone moves more than 2 mm despite the fixation, the system halts, and calls for re-registration.
One option to reduce tissue trauma is to use multiple dynamic reference bases to follow the motion of the robot base and the patient separately. Unfortunately, not every tracking system supports this, and it may cause difficulties to maintain the line-of-sight without disturbing the physician. Extending the active workspace of a tracking system may result in higher inherent accuracies due to the inhomogeneity of its field. Some commercially available systems combine surface-mounted and in-body fiducials to track external and physiological organ motion, though it requires a separate operation placing the markers. A successful example is the CyberKnife radiation therapy system (Accuray Inc., Sunnyvale, CA) that can track skin motion through a special suit and organ motion by taking bi-plane x-ray images and locating fiducials (gold beads) that were implanted pre-operatively [Saito2009]. Other groups tried different filtering approaches with limited effectiveness [Baron2010].
Robotic setups could incorporate accelerometers and gyroscopes, primarily to detect sudden changes; however these require electrical coupling and their resolution might not be sufficient for proper compensation. Besides, these would increase the costs and complexity of the system. Charge-coupled device (CCD) cameras can survey the OR, and image processing techniques could solve the localization problem, but the resolution may not be high enough, and it might have significant hardware requirements.Dynamic registration and correction for patient motion has been implemented with PET/SPECT scans [Fulton2002, Bruyant2005, Rahmim2007] to improve image quality through compensated reconstruction. However, these setups only considered rigid environment, where neither the camera, nor the PET gantry move.
Comments