Article in press - uncorrected proof Biomed Tech 2009; 54:8–13 2009 by Walter de Gruyter • Berlin • New York. DOI 10.1515/BMT.2009.002
Computer- and robot-assisted stereotaxy for high-precision small animal brain exploration Computer- und robotergestu¨tzte Stereotaxie fu¨r hochpra¨zise Exploration des Kleintierhirns
Lukas Ramrath1,*, Simon Vogt2, Winnie Jensen3, Ulrich G. Hofmann2 and Achim Schweikard1 Institute for Robotics and Cognitive Systems, University of Luebeck, Luebeck, Germany 2 Institute for Signal Processing, University of Luebeck, Luebeck, Germany 3 Department of Health Science and Technology, Aalborg University, Aalborg, Denmark 1
Abstract This contribution introduces a computer- and robotassisted framework for stereotactic neurosurgery on small animals. Two major elements of this framework are presented in detail: a robotic stereotactic assistant and the software framework for placement of probes into the brain. The latter integrates modules for registration, insertion control, and preoperative path planning. Two options for path planning are addressed: (a) atlas-based planning and (b) image-based planning based on computed tomography data. The framework is tested performing robot-assisted insertion of microelectrodes and acquisition of electrophysiological recordings in vivo. Concepts for data analysis pointing towards a mapping of position and neural structure to functional data are introduced. Results show that the presented framework allows precise small animal stereotaxy and therefore offers new options for brain research. Keywords: brain mapping; electrophysiology; medical robotics; microelectrode recordings; neural engineering; stereotaxy.
Zusammenfassung Dieser Beitrag stellt eine computer- und robotergestu¨tzte Umgebung fu¨r stereotaktische Eingriffe am Kleintier vor. Zwei Bestandteile der Umgebung werden im Detail vorgestellt: ein robotischer Assistent und die Softwareumgebung, um Instrumente im Kleintierhirn einzubringen. Letztere integriert dabei Module zur Registrierung, Steuerung des stereotaktischen Assistenten und zur Planung des Eingriffs. Zwei Optionen fu¨r die Planung werden vorgestellt: (a) atlasbasierte und (b) bildbasierte Planung auf *Corresponding author: Lukas Ramrath, Institute for Robotics and Cognitive Systems, University of Luebeck, Ratzeburger Allee 160, 23538 Luebeck, Germany Phone: q49-451-5005693 Fax: q49-451-5005202 E-mail:
[email protected]
Basis von Computertomographiedaten. Die Umgebung wird anhand der Einbringung von Mikroelektroden und der Akquise von elektrophysiologischen Ableitungen in vivo getestet. Konzepte zur Evaluierung der Daten im Hinblick auf eine Zuordnung einer ra¨umlichen Position und einer neuronalen Struktur zu funktionellen Daten werden vorgestellt. Die Ergebnisse zeigen, dass die vorgestellte Umgebung pra¨zise stereotaktische Eingriffe am Kleintier und neue Optionen in der Hirnforschung ermo¨glicht. Schlu¨sselwo¨rter: Elektrophysiologie; Gehirnkartographierung; medizinische Robotik; Mikroelektrodenableitungen; Neuroengineering; Stereotaxie.
Introduction In common neurosurgeries on small animals, manually operated stereotactic frames are used to implant probes (e.g., Lab Standard Stereotaxic, Stoelting Co., Wood Dale, IL, USA). One exemplary field of stereotaxy in small animal brain research is research on the effect of deep brain stimulation for the treatment of cerebral disorders, such as Parkinson’s disease, dystonia, and epilepsy w7x. Target areas of electrical high frequency stimulation include the thalamus, the internal globus pallidus or the subthalamic nucleus (STN) w1x and feature a relatively small size in the rodent model (e.g., the STN covers a volume of approximately 100 mm=100 mm=100 mm). Preoperative path planning is carried out using atlas information, such as the Paxinos atlas w4x. Then, target coordinates for the frame are calculated based on a manual landmark-based registration. Although automated systems have been developed for human stereotaxy w2x, no system has been designed for the rodent model. The main purpose of this work is based on the hypothesis that manual handling of the frame results in positioning inaccuracies. Another disadvantage of the current procedure is that atlas-based planning presumes a morphological similarity between all tested animals. Hence, the localization of measured signals to a spatial position is unreliable and targeting has to be proven by post-hoc histology. For small areas of interest, the current neurosurgical framework implies low repeatability of measurements, imprecise localization, and hardly allows the reliable and individual placement of two or more probes in one animal. This work presents a robot-assisted stereotactic environment for rodents to improve the spatial positioning of, e.g., microelectrodes. It consists of two major compo-
2009/37
Article in press - uncorrected proof L. Ramrath et al.: Computer- and robot-assisted stereotaxy for high-precision small animal brain exploration 9
nents: first, a robotic assistant called Spherical Assistant for Stereotactic Surgery (SASSU). Second, the corresponding software which features (a) atlas-based or (b) individual computed tomography (CT)-based path planning and surgical control. This system allows high-precision placement of probes into the brain and provides an improved framework for brain research. One possible application is robot-assisted microelectrode recordings for which electrical activity is precisely mapped to specific brain structures known from the CT-based planning. This, in a larger scale, stimulates the creation of a functional atlas and adequate methods of neural characterization which allows a comparison of neural signals in different animals.
All stages are controlled by a PCI card MCU-3000 (Roesch & Walter Industrieelektronik GmbH, Schwanau, Germany). A detailed description of the design and the operation can be found in w6x, where the system has been shown to feature a mechanical positioning accuracy of 32 mm. The SASSU additionally integrates a stereotactic frame for the animal which can be repeatedly mounted underneath the SASSU. The frame is made of plastics and can be placed into, e.g., a CT-scanner. This allows to preoperatively obtain a CT scan of the animal which is already fixed into the frame. This supports the planning and registration procedure for an individual animal. Control and path planning framework
Materials and methods Two components will be elaborated in the following section: (I) the stereotactic assistant SASSU and (II) the control and planning environment. The total framework integrates into the surgical workflow by allowing the following functionalities: (a) preoperative calibration, (b) preoperative path planning, (c) intraoperative registration, and (d) control of the tool insertion. Stereotactic assistant SASSU The SASSU manipulator adapts the center-of-arc principle, which is often applied in human applications (e.g., the ZD stereotactic system, Inomed GmbH, Teningen, Germany). These systems typically provide 5 degrees of freedom, which allow placing a tool at any desired point with a desired orientation. The mechanical design of the SASSU is shown in Figure 1A. All stages are equipped with a stepper motor (Nanotec Electronic GmbH, Landsham, Germany), which provides motion resolution -1 mm (e.g., 1 mdeg). To avoid positioning inaccuracies, external optical encoders are integrated into each stage.
Besides the stereotactic system, the control and planning framework is the second major component of the small animal micronavigation system. The framework is software-based and uses JAVA (Sun Microsystems, Palo Alto, CA, USA). Its functionality is adapted to the surgical workflow during stereotactic procedures on small animals. Specifically, it integrates modules for planning, registration, and probe insertion. The planning module allows (a) normalized atlas-based planning on coronal slices of the rat brain based on w4x and (b) individual CTbased planning based on a preoperatively acquired CT of the respective animal (see Figure 1B). Also, both approaches can be combined by, e.g., registration of the CT data to atlas information. Figure 1C shows the layout of the two-dimensional and three-dimensional planning scenario, which allows the interactive specification of multiple target points and the desired angle of insertion (dark gray and light gray trajectory lines). Based on characteristic landmarks on the rat skull, namely the Lambda and Bregma points w4x, the registration module establishes the relation between the robotic assistant and the rat skull. While landmarks are identified by placing the probe tip onto the characteristic points for the atlas-based
Figure 1 Planning scenarios for stereotactic surgery on a rat. (A) The surgical setup with a rat skull fixed into a CT-compatible frame which is removable from the operational site. (B) CT reconstruction of the rat skull fixed into the frame. (C) Two- and three-dimensional planning scenarios.
Article in press - uncorrected proof 10 L. Ramrath et al.: Computer- and robot-assisted stereotaxy for high-precision small animal brain exploration
planning, CT-based planning allows image-based identification of the landmarks. The control module allows the user to actuate each respective axis with a desired step size, velocity, and acceleration.
• Entropy (ENT) defined by:
m
Hs-8hjlog(hj)
Animal experiments and data analysis To test the system setup under real conditions, electrode recordings were performed for four trajectories in the rat brain using the simple atlas-based approach. Before operation, a concentric bipolar microelectrode (FHC Inc, Bowdoin, ME, USA) was attached to the adapter of the stereotactic assistant system. The electrode had a diameter of 250 mm, a pencil point tip configuration, and a length of 30 mm. The two electrode channels were connected to a recording system (Tucker Davis Technologies, Alachua, FL, USA). Microelectrode recordings were acquired using a 10 kHz low-pass filter and an 800 Hz high-pass filter as neural spiking activity is typically within the bandwidth of 1 kHz to 5 kHz. The sampling rate was 24 kHz. Prior to conducting experiments, an approval for all experimental procedures was obtained from the Danish Committee for the Ethical Use of Animals in Research. Each rat was anesthetized using a ketamine (100 mg/kg)-xylazine (5 mg/kg)-acepromazine (2.5 mg/kg) cocktail. Subsequently, it was fixed into the stereotactic frame and registration was performed. The target points and the insertion paths are specified in Table 1. Note that the trajectories Th3 and Th2 were obtained in the same animal but on the laterally mirrored trajectory. After skull opening and removal of the dura, the microelectrode was advanced into the brain with a speed of 0.1 mm/s and different step sizes. Shortly after electrode movement stopped, recording was started. During the microelectrode recordings, all electronic equipments, such as power supply of the heating pad and motor power of the SASSU, were turned off. Analysis of electrode recordings in this work was performed using statistical analysis of the raw extracellular recordings. In contrast to spike-related analysis, the wavetrain-related analysis avoids the user-dependent steps of spike detection and spike sorting. In w3x, the authors show the feasibility of mapping a microelectrode trajectory to the STN with regard to different wavetrain-related measures. Following these results, the following measures for robot-assisted brain exploration are proposed: • Median (med) • Root mean square (RMS) value defined as:
RMSsy(
1 N x(n)2) N n8 s0
(1)
where n denotes a discrete time step.
(2)
js1
where hjsaxj/N is the relative frequency of an event axj with 8N-1 js0axjsN. In this work, the measured voltage amplitudes were binned into Ns100 categories. • Power spectral density (PSD) is given by:
`
Sxx(v)s
8f
(m)exp(-jvmDt).
(3)
xx
ms-`
As computation of the PSD according to Eq. (3) requires an infinite set of autocorrelation coefficients and practically only N signal samples exist, the Welch periodogram has been used as an estimator. The PSD itself is a function of the frequency v and not a scalar value. Using the method by w5x, a scalar value P2500 will be introduced which adds the values of the PSD within an interval of 800 Hz to 2500 Hz, thus providing a scalar value.
Results The following results are presented with the reference position being the brain surface rather than the skull surface, as the brain surface was located approximately 100 mm beneath the computed position in a likely result of brain shift. Contact of the electrode with the brain surface was clearly detected via the audio output of the signal amplifier and may also be carried out automatically by analyzing the characteristic change in the signal amplitude. Figure 2 shows the trajectory Th3 plotted on an anatomic coronal atlas slice taken from the Paxinos atlas. Dark dots indicate the positions of recording. On the right hand side, a 3-s interval of selected recordings after the power supply was switched off is shown. Clearly, a number of different firing patterns can be seen as the probe is inserted deeper into the brain. Although much more data will have to be gathered, this fact motivates an automatic analysis and discrimination of the respective brain region. For each of the wavetrains recorded at the respective depths, statistical analysis measures were computed as discussed previously. Figure 3 shows the three probe trajectories Th3, Th2 (one animal), and We2 (different animal) for each statistical measure. The correlation coefficients for all measures and all trajectories are summarized in Table 2.
Table 1 Trajectory specifications for electrophysiological recordings by electrode insertion into the rat brain in vivo. Name
Animal
Animal weight (g)
AP position (mm)
Lateral position (mm)
Coronal approach angle (deg)
Th3 Th2 Th1 We2
1 1 2 3
280 280 380 290
-3.96 -3.96 -3.96 -3.96
2.97 -2.97 2.97 2.97
0 0 0 0
Article in press - uncorrected proof L. Ramrath et al.: Computer- and robot-assisted stereotaxy for high-precision small animal brain exploration 11
Figure 2 Depth resolved microelectrode recordings. On the left side, the trajectory of the electrode through a coronal section of the rat brain is shown. Recording points are indicated as black dots. The right side shows a 3-s interval of the respective microelectrode recordings, as could be done for an electrophysiological rodent brain atlas.
Discussion The results show that the developed system enables the surgeon to perform robot-assisted, stereotactic neurosurgery for microelectrode recordings in the small animal model. The SASSU system offers multiple benefits over existing stereotactic frames for small animals: (a) a high mechanical positioning accuracy, (b) high repeatability of experiments, and (c) the ability to compensate for tilting of the skull due to inaccurate fixation. Within the course of surgery, the user is able to control the stereotactic robot via an easy-to-use PC planning interface. The positioning accuracy of the SASSU allows to relate the corresponding measurements with a high reliability to a spatial position in the brain. As results are obtained for the atlas-based approach only, results are displayed on the respective coronal slice of the Paxinos atlas. Figure 2 visualizes the concept of depth-resolved microelectrode recordings. Looking at the time-resolved recordings on the right hand side of Figure 2, the recordings at different depths show different waveforms. The results in Figure 3 and Table 2 show that the four chosen measures (a) vary for different depths and (b) correlate between the same tracks in one animal (only laterally mirrored) and in different animals. These results support the hypothesis of a similar and characteristic neural activity obtained on the same tracks in one animal (only laterally mirrored) and
in different animals. Not all measures, however, display the similarity in the same quality. For the RMS (0.66–0.85) and the PSD (0.34–0.89), a stronger correlation than for the ENT (0.0–0.68) and the median (0.03–0.74) value can be observed. The fact that no exact quantitative correspondence is observed is due to different aspects, which could negatively influence the measurements and thus lead to less significant correlation. These effects include the accuracy of the registration process and the shortcomings of planning on the basis of Paxinos atlas information. The results motivate future application of the system in the direction of electrophysiological brain mapping. A premise is the preoperative acquisition of individual small animal magnetic resonance imaging- or CT-data. Then, neural activity can be directly related to neural structures identified from the preoperative data. The robot-assisted framework will provide the required spatial positioning accuracy. Such a mapping would trigger research activities addressing the intelligent characterization and comparison of neural signals. In a large scale, such mapping could provide a functional atlas. Neurophysiologic characteristics, however, are dependent on many factors, such as anesthetics, electrodes, recording systems, etc. Establishing such an atlas therefore requires a careful choice of the surgical setting. Once established, such an atlas could establish a novel modality for neuronavigation for which the specific signal char-
Article in press - uncorrected proof 12 L. Ramrath et al.: Computer- and robot-assisted stereotaxy for high-precision small animal brain exploration
Figure 3 Statistical measures for trajectories shown in a coronal slice of the rat brain at the AP position -3.96 mm. Abbreviations denote different structures and are taken from the respective coronal slice of the Paxinos atlas. (A) Depth resolved median for different trajectories: (top) trajectory Th3 (middle) trajectory Th2 (bottom) trajectory We2. (B) Depth resolved RMS for different trajectories: (top) trajectory Th3 (middle) trajectory Th2 (bottom) trajectory We2. (C) Depth resolved entropy for different trajectories: (top) trajectory Th3 (middle) trajectory Th2 (bottom) trajectory We2. (D) Depth resolved PSD for different trajectories: (top) trajectory Th3 (middle) trajectory Th2 (bottom) trajectory We2.
Article in press - uncorrected proof
0.514 0.825 0.885 1.000 0.467 0.711 1.000 0.885 1.000 0.342 0.467 0.514 -0.002 1.000 -0.482 -0.461 1.000 0.731 0.658 0.769 -0.381 0.145 0.740 1.000
0.731 1.000 0.651 0.847
0.658 0.651 1.000 0.837
0.769 0.847 0.837 1.000
1.000 -0.002 0.688 0.371
0.688 -0.482 1.000 0.632
0.371 -0.461 0.632 1.000
0.342 1.000 0.711 0.825
Th3 Th2 Th1 PSD
We2 Th3 Th2 Th1
ENT
We2 Th3 Th2 Th1
RMS
We2 Th3
acteristics could be compared to the multi-dimensional brain map and thus allow determination of the current position of the probe tip in situ. This would augment existing neuronavigation modalities. In this context, however, multiple questions have to be answered: Is there a one-to-one mapping of spatial position to electrophysiological signal characteristics? Which characteristics of the signal should be evaluated for reliable spatial mapping? How do signal characteristics change from region to region (steady or unsteady)? To summarize, robotassisted small animal stereotaxy offers research options for the development of new methods for neural analysis and for the treatment or diagnosis of, e.g., neural diseases.
Acknowledgements This work was funded by FP Biomedizintechnik at the University of Luebeck, Germany, and the German Federal Ministry of Education and Research (BMBF) project BiCIRTS 13N9190. The authors acknowledge the support of the Department of Neurology (Prof. Dr. med. A. Moser) and thank the staff at the Biomedicinsk Laboratorium, Aalborg Hospital, Denmark for assistance during the animal experiments.
References w1x Benazzouz A, Hallet M. Mechanisms of deep brain stimulation. Neurology 2000; 55: 13–16. w2x Karas C, Chiocca E. Neurosurgical robotics: a review of brain and spine applications. J Robotic Surg 2007; 1: 39–43. w3x Moll C, Struppler A, Engel A. Intraoperative Mikroelektrodenableitungen in den Basalganglien des Menschen. Neuroforum 2005; 1: 11–20. w4x Paxinos G, Watson C. The rat brain. Elsevier Inc., 2007. w5x Pesenti A, Rohr M, Egidi M, et al. Priori, the subthalamic nucleus in parkinson’s disease: power spectral density analysis of neural intraoperative signals. Neurol Sci 2003; 24: 367–374. w6x Ramrath L, Hofmann U, Schweikard A. A robotic assistant for stereotactic neurosurgery on small animals. Int J Med Robotic Comput Assist Surg 2008;4: 295–303. w7x Tronnier V, Fogel W, Krause M, et al. High frequency stimulation of the basal ganglia for the treatment of movement disorders: current status and clinical results. Minim Invas Neurosurg 2002; 45: 91–96.
-0.636 0.309 1.000 0.740 -0.034 1.000 0.309 0.145 1.000 -0.034 -0.636 -0.381 We2 Th1 Th2 Th3
We2
Th1
Th2
Received June 25, 2008; accepted October 16, 2008
Median
Table 2 Correlation coefficients for the median, the root mean square (RMS), the entropy (ENT), and the power spectral density (PSD) of neural activity recorded along three independent trajectories in the rat brain.
L. Ramrath et al.: Computer- and robot-assisted stereotaxy for high-precision small animal brain exploration 13
