A typical robot includes a manipulator, user interface, tracker and registration system, and a controller. These components have many of the same network communication requirements as medical devices.
However, due to the high degree of automation expected from robots, the most important aspect is reliable communications with low latency. A suitable network must guarantee the correctness of transferred data and support real-time messages for fast response in emergencies.
Another reason for using an open network: Accelerating developments in robotics are making the reuse of existing components essential to research. This research economy is supported by open-source software and, for hardware, standard off-the-shelf equipment that allows plug-and-play connection.
The software framework developed at the Surgical Planning Lab of Brigham and Women's Hospital, Boston, combines navigation sensors in a CANopen network, and relies entirely on open-source software components. To increase reliability, the framework is also built into an Image-Guided Surgery Toolkit, an open-source library with many safety features for building image-guided surgery applications.
CanFestival library for CANopen communication is used because it supports most CANopen features and offers high flexibility to designers. To simplify connection of navigation sensors to the CANopen network, this flexibility is narrowed to the smallest number of functions a user would need to handle.
To use the framework in a new system, a user need only add the interface-driver software to get access to the values of the navigation device, define the settings, start the communication, and program a loop or a call-back function to continuously map the newest position values to the right PDO bytes. From that point onward, the navigation device acts as a CANopen NMT slave complying with the CiA 301 specification (CANopen application layer and communication profile) and device profile CiA 401 for generic input/output modules.
As an additional function, the framework can be used as a simple NMT master in a CANopen network. This feature is added to allow initial tests of new systems and the creation of small CANopen networks where only two devices communicate.
The framework demonstrates that CANopen with its underlying CAN-technology offers the necessary features to meet requirements for robotics in medical applications. Most important are its reliable communication and low latencies for high-priority messages. For real-time applications, the nondestructive bus-arbitrary method is also useful for the conclusion and availability of open-source software solutions. CAN functions at a maximum of 1 Mbit/sec, enough for controls and standard sensor data. One caveat: It can become an issue when transmitting real-time image or video data. Here, the integration of another data-line would be necessary.
MRI-compatible robot: One application
Needle biopsies and brachytherapies are typical interventions for diagnosing or treating prostate cancer. In both cases, a needle is inserted into the patient and guided by a surgeon to an exact location. MRI can provide preoperative 3D visualization and intraoperative 2D real-time images, which improve planning, monitoring, and intervention accuracy. However, difficulties using an MRI-scanner for this task include high magnetic fields (of 1.5 Tesla or greater) and a limited operating space for the surgeon.
A robotic system that overcomes these obstacles is under development at the Johns Hopkins University, Baltimore, in cooperation with the Surgical Planning Lab. It is an assistance system placed between the legs of the patient that moves the needle height and angle to a position according to target points set by a surgeon and acquired MRI images. In this position, the surgeon inserts the needle manually transperineal into the patient while the MRI scanner continually produces real-time images of the needletip's position.
Besides the manipulator, the robotic system also includes a controller and a PC running navigation software. The software is a newly developed module for a program called 3D Slicer, and serves as the user control interface. In a preliminary design, Ethernet provided communication between robot controller and navigation module. For the Surgical Planning Lab project, Ethernet was replaced with CAN and the CANopen protocol to evaluate functions of the developed framework. Here, the navigation device is the robot manipulator, which delivers needletip position and orientation data.
The CAN-based framework also performs well in another application, an electromagnetic tracking device called the Aurora, manufactured by NDI Inc., Ontario, Canada. Although the software runs on standard PC hardware with USB-to-CAN-adapters, latency, bus utilization, and the maximum data delivery rate (1,000 Hz) are sufficient for the Aurora's navigation sensors.
FOR FURTHER READING
G. S. Fischer. MRI-Compatible Pneumatic Robot for Transperineal Prostate Needle Placement. IEEE/ASME Transactions on Mechatronics, 3(3), June 2008.
P. W. Mewes. Integrated System for Robot-Assisted Prostate Biopsy in Closed MRI Scanner. In IEEE International Conference on Robotics and Automation, May 2008.
AT A GLANCE
Mounting research for using robots in medical applications is giving rise to numerous new market-ready medical robotic solutions. These developed robotic systems rely on communication technologies to connect components and offer external interfaces. Several open communication standards used in modern hospitals are intended for these advanced medical devices. But are any suitable for the special requirements of robotic systems?
In an effort to find out, researchers at the Surgical Planning Laboratory of Brigham and Women's Hospital, Boston, evaluated the Control Area Network standard or CAN and its higher-layer protocol, CANopen. Specifically, the researchers developed software that connects navigation sensors — such as tracking devices or position-sensors — with a CANopen network. This framework was tested with an electromagnetic tracking-device and in an MRI-compatible robotic system for placing prostate needles.
To make CANopen more suitable for robotics in medical applications, new generic application and device profiles need developing, to extend the network's plug-and-play capabilities to other devices used in medical research. Present profiles such as CiA 401 or specialized CiA 412-1 cannot yet fill this gap. That said, CANopen offers features necessary to meet medical robot requirements. The researchers have successfully developed a framework for CAN-based robot guidance during physician and magnetic resonance or MR image-directed procedures. — H.K.