In vivo imaging of epileptic foci in rats using a miniature probe integrating diffuse optical tomography and electroencephalographic source localization

SUMMARY

Objective

The goal of this work is to establish a new dual-modal brain mapping technique based on diffuse optical tomography (DOT) and electroencephalographic source localization (ESL) that can chronically/intracranially record optical/EEG data to precisely map seizures and localize the seizure onset zone and associated epileptic brain network.

Methods

The dual-modal imaging system was employed to image seizures in an experimental acute bicuculline methiodide rat model of focal epilepsy. Depth information derived from DOT was used as constraint in ESL to enhance the image reconstruction. Groups of animals were compared based on localization of seizure foci, either at different positions or at different depths.

Results

This novel imaging technique successfully localized the seizure onset zone in rat induced by bicuculline methiodide injected at a depth of 1mm, 2mm and 3mm, respectively. The results demonstrated that the incorporation of the depth information from DOT into the ESL image reconstruction resulted in more accurate and reliable ESL images. Although the ESL images showed a horizontal shift of the source localization, the DOT identified the seizure focus accurately. In one case, when the BMI was injected at a site outside the field of view (FOV) of the DOT/ESL interface, ESL gives false positive detection of the focus while DOT shows negative detection.

Significance

This study represents the first to identify seizure onset zone using implantable DOT. In addition, the combination of DOT/ESL has never been documented in neuroscience and epilepsy imaging. This technology will enable us to precisely measure the neural activity and hemodynamic response at exactly the same tissue site and at both cortical and sub cortical levels.

INTRODUCTION

Epilepsy is a common neurological syndrome, affecting up to 3% of the population.¹ While medication is the first line of treatment, for many patients seizures cannot be adequately controlled with medication alone. Many of these intractable cases are candidates for epilepsy surgery, which is potentially curative if the epileptogenic focus can be identified and safely removed. Surgical options for patients with partial seizures are based on identification and subsequent surgical treatment of the seizure onset zone, or the minimum cortical volume from which seizures arise.^2–4 While, in the most commonly performed surgery, medial temporal lobe epilepsy, cure rates often exceed 70% in carefully selected, cure rates are still much lower in patients with nonlesional neocortical epilepsy.^5–10 One reason for such low cure rates is that localization of the seizure onset zone in neocortical epilepsy is less well understood. Thus, there is unmet need for research focused on identifying the brain regions responsible for generating seizures in neocortical epilepsy syndromes.

In an effort to overcome this limitation, several modalities, such as functional magnetic resonance imaging (fMRI), single-photon emission computed tomography (SPECT), and positron emission tomography (PET), have been employed clinically to identify areas of abnormal blood flow and metabolism associated with epileptic form events including the seizure onset zone.¹¹ Although each of these techniques has its merits, each has either poor temporal or spatial resolution (on the scale of centimeters to several seconds) and do not offer information about seizure onset or propagation patterns. These factors underscore the unmet need for more effective techniques for presurgical mapping of neocortical epilepsy.

Diffuse optical tomography (DOT) has been recently used to in vivo map the onset and spread of epileptic events with excellent spatial and temporal resolution.^{12; 13} Meanwhile, EEG source localization (ESL) is emerging as a useful technique for the study of temporal brain dynamics in animals and in humans.^14–23 Furthermore, using priori information provided by other imaging modalities to constrain ESL reconstruction has been demonstrated to be an effective approach that can improve the accuracy and reliability of the solution.^24–28 By integrating DOT and ESL, a novel technique that can precisely map seizures and localize the seizure onset zone and associated epileptic brain network could be realized. The depth information derived from DOT can be used as a priori information about the location of the sources to enhance the inverse solution in ESL, which is promising approach for better electrical localization of seizures. Here we integrate DOT and ESL through a miniaturized probe, and demonstrate its ability using an acute rat model of focal epilepsy.

METHODS

Animals

Male Sprague–Dawley rats (Harlan Labs, Indianapolis, IN) weighing 240–260g on arrival were allowed one week to acclimate to the 12-h light/dark cycle and given food and water ad libitum. All procedures were approved by the University of Florida Animal Care and Use Committee and conducted in accordance with the National Institutes of Health Guide for the Care and Use of Experimental Animals.

Imaging system

The schematic of the integrated DOT/ESL system is shown in figure 1. This system has been validated using extensive tissue-mimicking phantom experiments.²⁹ As illustrated in figure 1, the computer sends a starting signal to the LED controller to sequentially light up 12 LEDs (780 nm) (Epitex, Inc.). The light beams are delivered to the measuring interface via fiber optic bundles. Diffusing light received by 13 detection fiber bundles are converted to electrical signals and pre-amplified by photodetectors (high sensitivity avalanche photodiode (APD): C5460-01, Hamamatsu). The amplified signals are collected through multi-channel data acquisition (DAQ) boards (NI, PXI-6358). A current driving circuit is designed to drive the LEDs and a Field Programmable Gate Array (FPGA) core board (NI, PCI-7811) is used to control the LED timing sequence. The output power of each LED is adjustable through its independent DC power supply to achieve optimal signal-to-noise ratio (SNR). In the ESL subsystem, 16 channels of EEG signals from the electrodes are pre-amplified and digitalized by a multi-channel pre-amplifier (Tucker-Davis Technologies, RA16PA) and a fast digital signal processor (Tucker-Davis Technologies, RZ5), respectively. Another data acquisition board is used to collect the processed EEG signals. Figure 2(a) shows the photograph of the integrated DOT/ESL interface (17mm×17mm), which is made of aluminum with a weight of about 0.2g. Sixteen homemade copper plates (2mm×2mm) attached to the surface of the interface are used as the EEG electrodes in the ESL subsystem. The diameter of each optical fiber bundle is 1.5mm. The corresponding pattern of source (red) and detector (green) positions for DOT and electrodes (black) positions for ESL is shown in figure 2(b).

Figure 1.

Schematic of the integrated DOT/ESL imaging system

Figure 2.

(a) Photograph of the DOT/ESL interface prototype. (b) The pattern of source (red) and detector (green) positions for DOT and electrode (black) positions for ESL.

For the experiments, the scalp and skull of the rat brain were removed and the DOT/ESL interface was placed right at the surface of the cortex (see the inserted photograph in Fig. 1). DOT/ESL data were recorded immediately before the electrographic seizure onset time (resting state), and after the seizure onset.

Animal procedures

Animals were place in a steraotaxic frame and anesthetized induced with isoflurane (4% for induction and 1.5% for maintenance), then maintained with urethane (1g/kg of body weight, intraperitoneal injection). After the scalp was cut and skull was removed, one, 300µm diameter stainless steel screw electrode (FHC, Bowdoin, ME) was implanted as a reference electrode into the occipital bone. Cortical local field potentials were obtained at 12 kHz, digitized with 16 bits of resolution, and band pass filtered from 0.5 to 6 kHz.

Subsequently, rats (n=4) received 10 µl of 1.9 mM bicuculline methiodide (BMI) and 10 µl saline as control at a depth of 1mm to 3mm below the surface of the right or left parietal cortex covered by DOT/ESL interface, respectively, and other rats (n=2) were injected the same amount of BMI at a site outside the field of view (FOV) of the DOT/ESL interface. Each animal was injected with bicuculline (BMI) into one site of the brain. The infusion was performed at a rate of 0.3 µl/min. The infusion system consisted of a 100 µl gas-tight syringe (Hamilton, Reno, NV) driven by a syringe pump (Cole-Parmer, Vernon Hills, IL). The injector was mounted on a micromanipulator that allowed precise injections at a depth below the surface.

Image analysis

DOT and ESL images were reconstructed by iteratively solving the photon diffusion equation and Poisson’s equation, respectively, using the nonlinear, finite element-based reconstruction algorithms described in Jiang (2010).³⁰ The algorithms use a regularized Newton’s method to update an initial optical property (absorption and scattering coefficients)/current source distribution iteratively in order to minimize an object function composed of a weighted sum of the squared difference between computed and measured optical/local field potential data at the medium surface. In order to obtain more reliable depth information about the location of the sources from ESL, the depth information derived from DOT was used as a hard-priori to constrain the source space of the electrical source imaging.

RESULTS

After the dual-modal DOT/ESL system was validated using extensive phantom experiments,²⁹ we first tested the system accuracy for identifying a focal ictal activity over a time window of 6 min. Whereas a significant increase (t-test, p < 0.05) of optical absorption is seen in the region of the BMI injection, no absorption contrast is observed in the resting state (Fig. 3). These results suggest that the increase in local and surrounding brain tissue absorption was mostly due to the local bicuculline induced seizures. To make sure if this contrast was indeed due to the seizure onset (not because of the injection of a liquid or physical contractions), we have performed controlled measurements where the rat was injected 10µl of saline: no significant contrast (t-test, p > 0.05) in absorption or scattering was observed at the location of the injection. The seizure focus was also clearly identified by the ESL images. The size of the focus detected by ESL, however, was overestimated, and the ESL-identified position of the focus was shifted as well compared to the BMI injection site. In this case, the depth information derived from DOT was used as a priori information to segment the regions in ESL reconstruction.

Figure 3.

Detection of seizure focus. Top panel: EEG recordings at resting state (2 min window) and after BMI injection (6 min window). Bottom panel: time series DOT and ESL images overlaid on a structural MRI in the horizontal plane at resting state (t=0min) and after the BMI injection (t=0.65, 2.25, 3.85 and 5.45min).

To test the sensitivity of detecting seizure focus at different depths, we injected BMI at depths of 1, 2 and 3mm below the surface of the right or left sensory cortex (S1) (n=3),³¹ and the resulting DOT/ESL images are shown in Fig. 4. In ESL, the depth information derived from DOT was also used as a priori information to constrain the ESL inverse solution. The results demonstrated that the constraint derived from DOT effective improved the accuracy and reliability of the ESL solution, especially the depth information of the source. We also see that similar to the results shown in Fig. 3, the position of the focus imaged by ESL is horizontally shifted and its size is overestimated (top panel, Fig. 4) while DOT identified the seizure focus more accurately. The volumetric imaging ability of DOT is better appreciated from the coronal section images shown by the bottom panel of Fig. 4.

Figure 4.

Depth sensitivity of seizure detection. Top panel: DOT and ESL images overlaid on a photograph of the cortex at 1mm (a), 2mm (b) and 3mm (c) depth of BMI injection. Bottom panel: Coronal section DOT and ESL images overlaid on coronal section MRI at 1mm (a), 2mm (b) and 3mm (c) depth of BMI injection.

We also evaluated the specificity of seizure detection for DOT/ESL. In this study, we injected BMI at a site outside the field of view (FOV) of the DOT/ESL interface at the left primary motor cortex [M1(AP: 3.2mm; ML: 2.5mm; DV: 1.5 mm)] (Fig. 5a) and the right caudate putamen [CPu(AP: 3.2mm; ML: 2.5mm; DV: 1.5mm)] (Fig. 5b) (n=2).²⁶ In both cases from the sagittal section DOT/ESL images shown in Fig 5 we see that ESL gives false positive detection of the focus while DOT shows negative detection for both cases. These results show that, in this specific case, DOT can correctly detect negative seizure focus for cases ESL provides false positive detection. We need to point out that we could not impose the constraint to ESL here because the depth information could not be derived from DOT in this case. This case demonstrates the unique advantage of this dual-modal technique for providing the surgeons complementary and complete information to localize the seizure onset zone correctly.

Figure 5.

Specificity of seizure detection. Left of (a) and (b): DOT and ESL images overlaid on a photograph of the cortex at 3mm (a), and 4mm (b) depth of BMI injection. Right of (a) and (b): Sagittal section DOT and ESL images overlaid on sagittall section MRI at 3mm (a), and 4mm (b) depth of BMI injection. The dashed square indicates the field-of-view of the DOT/ESL interface.

DISCUSSION

The main finding of this study is that implantable DOT/ESL imaging is a novel tool for precisely mapping seizures and localizing the seizure onset zone. This study investigated the localization of seizure onset zone at different locations and different depths during focal seizure onset in an acute model of focal epilepsy. This technology will enable us to precisely measure the neural activity and hemodynamic response at exactly the same tissue site and at both cortical and sub cortical levels, providing a new tool for unprecedented investigation of neuro-vascular coupling in epilepsy. To the best of our knowledge, this is the first to identify seizure onset zone using implantable DOT, and also the first to combine DOT and ESL in a miniature probe.

This technique would have a large impact in several areas. The first is that it could complement existing gold standard subdural electrodes techniques. In fact, the multi-channel intracranial EEG data recorded by the dual-modal device coupled with a finite element based reconstruction algorithm allow us to realize EEG source localization for spatial localization of the seizure onset zone.^{18; 22; 32; 33} However, as we pointed out above, some works need to be done to improve the reliability of the ESL results. Therefore, This simultaneous DOT/ESL brain mapping provide not only the ability that would allow for more precisely mapping the seizure onset zone and the brain tissue which is the key to successful epilepsy surgery, but also a unique technique for unprecedented investigation of neuro-vascular coupling in epilepsy since DOT images hemodynamics and ESL measures neural activity. In particular, the results demonstrated that it is an effective approach to improve the accuracy of ESL by imposing the constraints derived from DOT. Second, since our previous DOT study detected early hemodynamic responses with heterogeneous patterns several minutes preceding the electroencephalographic seizure onset, supporting the presence of a ‘‘pre-seizure’’ state, the device could be used for early seizure detection and prediction.³⁴ Optical signal measures could be combined with closed-loop seizure prevention strategies and local treatment. The possibility of seizure prediction has given hope for new warning and therapeutic devices for individuals who cannot be successfully treated with current therapies. Finally, the device could be modified to drive light-sensitive channels (optogenetics) while measuring optical responses in near real-time.³⁵ As a result, an optrode optical mapping device could be used to optically-guide plasticity. Although these studies are beyond the scope of this work, they are directly dependent on this initial step which is to design, build, and test a device for subdural chronic monitoring of optical signals in a realistic animal model of neocortical epilepsy.

We found that the shift of the focus position imaged by ESL occurred even the depth information from DOT was imposed to constrain the space of the source localization in ESL. Although the development of ESL reconstruction algorithm and the use of the constraint is not the focus of this work, we still notice the potentials to improve the EEG source localization accuracy with reconstruction algorithms and constraint-based approaches. In recent years, several algorithms such as sLORETA and MUSCIS, have been developed and compared through extensive simulation studies.^19–23 These studies suggest that some key factors, including the number and distribution of EEG electrodes, and the signal to noise ratio (SNR) of recording signal, significantly affect the localization accuracy. Similar to the discussion by Michel et al.,¹⁹ we believe that the shift of the localization of seizure activity seen in this study was mainly derived from the limited number of the EEG electrodes and the plane structure of the probe used in our DOT/ESL interface.

One of the limitations of this technique in this study is the probe size, although still smaller than in many other reports of DOT system used in the study of epilepsy. In order to really realize an implantable technique, a more miniaturized design is required. Fortunately, with the rapid development of the electronics engineering technology, more and more miniature electronic components, especially the ultra-micro sized version of the surface mount LEDs and Photodiode, appeared and was widely applied in the field of biomedical imaging.^{36; 37} Based on these new techniques, a miniature probe more suitable for implantation in animal and human brain can be achieved. The inflexible configuration of the interface is also one of the limitations of this design, which may affect the accuracy of ESL imaging. In future, a more flexible material such as medical grade silastic films can be used as the substrate of the sensors. Another limitation is the number of EEG electrodes used in ESL. Theoretical and experimental studies have demonstrated that increased spatial sampling of the interictal EEG will considerably increase the output of interictal EEG recordings in epilepsy patients.³⁸ Prior studies also indicate that a minimum of 100 electrodes is needed to properly sample the electric field from the full head surface.³⁹ However, this problem could be easily resolved by decreasing the size of electrode and increasing the electrode number. Despite these limitations, a more sophisticated and implantable dual-model imaging technique is expected in the near future.