Safe and clinically applicable robotized Transcranial Magnetic Stimulation

Transcranial Magnetic Stimulation (TMS) allows for non-invasive and painless stimulation of the (human) brain, in particular of its cortical structures. A strong, rapidly increasing current is driven through a magnetic coil placed on the head of a subject. The generated magnetic field passes through the human skull and induces an electric current inside the cortex which can lead to local stimulation. Currently, neuro-navigated TMS is the state-of-the-art procedure to assist the operator in placing the coil on the head. However, the coil is positioned and held at the target by hand. An optical tracking system visually assists the operator for target localization and coil positioning. Recently, robotized TMS has been introduced, combining neuro-navigation with automation. It is based on a serial robot arm for coil placement and an optical tracking system for head tracking and navigation. In this way, robotized TMS allows for precise and repeatable coil placement. Furthermore, the robotized TMS system employs active motion compensation (MC) to maintain the correct coil position throughout stimulation -- even when the head moves naturally. Even though the system provides increased accuracy, repeatability and comparability, it is not yet mature. In its current state, only experienced and trained users can employ the system without difficulty. This fact is the mainspring of this work, which describes the systematic further development of the current system to a safe and clinically applicable robotized TMS system. To date, a systematic analysis and practical evaluation of the current system is still missing. Therefore, we foremost perform this investigation. We recorded head motion in realistic TMS scenarios and analyzed the impact of head motion on the accuracy of TMS. This analysis fundamentally showed the importance of active motion compensation for accurate TMS. After 30min of stimulation, on average 32% of the induced electric field strength is lost due to head motion in a conventional TMS setup. In contrast, with robotized TMS on average less than 5% of the induced electric field strength is lost. Furthermore, we practically evaluate the robotized TMS system (in its present state) during TMS studies. On the one side, the studies and their outcomes support the special features of robotized TMS for precise and accurate coil positioning. On the other side, however, this practical evaluation shows the deficits of the present implementation. Therefore, there is a need for an user-friendly, safe and clinically applicable system. As primarily researchers, neuroscientists, physicians and medical staff are the operators of TMS, we improve the robotized TMS system for an easy and unproblematic clinical application. The robotized system requires a calibration between tracking system and robot. As the system design is partially mobile, the calibration step might have to be performed frequently, which takes additional time. When the robot and/or the tracking system shift, calibration must be re-performed. It is even worse when such a shift occurs unrecognized during treatment. We develop an online calibration method that is able to update and check the current calibration during application in real-time. For that purpose, a marker is attached to the robot's third link in such a way that it is visible for the tracking system throughout the application. By using a rigid transformation from the marker to the coordinate system in the robot's fourth joint, the calibration can be calculated directly. The marker is tracked by the tracking system and the robot's forward calculation is applied to the fourth joint. A practical evaluation shows that the positioning accuracy of the robotized TMS system is in the same range as with the current calibration method. As an industrial robot is a complex and potentially dangerous system, the allowed robot trajectories are strongly restricted in the current implementation. Any potentially dangerous trajectory for the patient is forbidden by the software control. Therefore, a manual pre-positioning of the robot is required frequently to allow a safe and automatic coil placement afterwards. This is a quite difficult task for unexperienced users. For this reason, we implement a positioning method that allows to position the robot in a hand-guided fashion. To this end, we install a force-torque sensor onto the system. The sensor detects the occurrent forces and torques on the coil and transfers these values into robot movements instantaneously. In this way, the user is able to position the coil easily and fast, without need of the complex, manual pre-positioning with the robot controller. Additionally, the force-torque sensor allows to measure the contact pressure of the coil to the head. In this way, the coil is optimally placed on the patient's head. In combination with motion compensation, the force-torque control maintains the optimal contact pressure during application. This guarantees an optimal stimulation. Due to the direct interaction of robot and patient/operator, safety is a very critical aspect in the development of medical robotic systems. Many systems, such as the robotized TMS system, are based on industrial robots that are adapted to the specific requirements of the application. Safety measures are then purely implemented in the software. Typically, the workspace is restricted and the robot velocity is limited. Even though this implementation is sufficient for most situations, overall system safety cannot be achieved. Programming errors or communication faults can bypass the safety mechanisms. To overcome this, we develop an independent safety layer, named FTA sensor, for the robotized TMS system. It combines a force-torque sensor with an accelerometer for independence from robot input. An embedded system performs the required computations in real-time (roughly 1 ms) and is directly connected to the robot's emergency circuit. In case of an error or collision, the robot is stopped immediately to protect the patient and/or the user from harm. Beside this safety feature, the sensor provides the same functionality as a standard force-torque sensor. As the computations on the embedded systems are much faster as in a pure software implementation, the above presented methods for hand-assisted positioning and contact pressure control were further optimized. Currently, indirect head tracking is the standard technique for neuro-navigated and robotized TMS. This method requires a marker on the patient's head, which is registered to the patient's head. Therefore, the marker must not shift after the registration, as this would lead to inaccuracies and errors in coil positioning. Direct tracking systems, on the contrary, straightly measure the surface of the head or specific facial feature points. This data can then be automatically matched to three-dimensional (3D) head scans of the patient. We test different systems for direct head tracking. First results show that 3D laser scanning systems can be utilized on principle. With advanced technologies, the 3D laser scanning systems will further improve for their application in direct head tracking for TMS. In summary, we implement a safe and clinically applicable robotized TMS system. A key feature is the developed FTA sensor with the optimized hand-assisted positioning method and the contact pressure control. The FTA sensor is now available as an extension to SmartMove (Advanced Neuro Technology B.V., Enschede, The Netherlands), called TouchSense, for the clinical market.