By Greg Blackman
Surgical navigation systems are still a relatively new technology. However, according to Claudio Gatti, co-CEO of Claron Technology, which develops vision software for medical applications, the technique has become the standard of care when dealing with neurological or spinal injuries.
Claron Technology’s MicronTracker is a navigational tool for surgeons and uses Point Grey Research’s Bumblebee2 stereovision cameras to track where a surgeon’s instrument is inside the patient. The system operates by overlapping 3D Computed Tomography (CT) or Magnetic Resonance Imaging (MRI) scans, generated prior to surgery, with real-time image data produced by the Bumblebee2 cameras during the procedure. For instance, neurosurgeons will use CT scans to pre-plan a surgical procedure for resecting a brain tumour. During the procedure, the cameras from Point Grey track small, chequered marker patterns, called Xpoints, on the external portion of the tool; these report the position and orientation of the instrument inside the body. The system then overlays the real-time data with the preoperative 3D model from the CT scans, to give the surgeon a virtual view of what is ahead or around the tool.
FIGURE 1. The system can overlay the real-time data with the preoperative 3D model from CT scans, to give the surgeon a virtual view of what is ahead or around the tool.
There are other tracking systems available using infrared light, but the MicronTracker uses vision and passive lighting to identify the marker. According to Gatti, the navigation systems currently in use are large, complex and costly systems. With the computer vision-based approach, the navigation technology is being scaled down towards simpler and more compact systems.
‘There were two main challenges in designing the system,’ explains Gatti. ‘The robustness of the recognition and being able to do it in real time, and, secondly, the precision of the measurement – an accuracy of 0.3mm rms across a field of view up to two metres in diameter.’
The robustness of the recognition is dependent on the detection algorithm and the design of the pattern placed on the surgical tool. In addition, the software needs to be able to adapt to the lighting conditions. ‘Surgical lighting can be very powerful, measuring hundreds of thousands of lux,’ says Gatti. ‘However, this light is often focused on a specific area, so some areas can be very bright while others will be in almost complete darkness.’
Point Grey Bumblebee2 cameras have the functionality to cycle through different exposure modes depending on the lighting conditions. If there is a large range of illumination, the MicronTracker’s high dynamic range mode is designed to handle it.
The precision of the measurement was addressed by a very accurate detection algorithm and high precision custom calibration of the camera. ‘The resulting detection algorithm allows the identification of the Xpoint on the x-y plane with an accuracy of 1-2 per cent of a pixel,’ comments Gatti.
MicronTracker can also use multiple stereoscopic cameras to extend the field of measurement, and each camera can support either binocular or trinocular vision depending on the surgical procedure and the degree of accuracy required.
FIGURE 1. Field of measurement can be easily increased by adding more cameras
Gatti notes that this type of computer vision-based product is driven largely by advances in software. In addition, with a steady stream of faster CCD sensors becoming available and higher communication bus bandwidth, this technology could become much more common within hospitals.
Reprinted with permission from the imveuroper article, "Vision is a cut above " in the August/September 2009 issue