Imaging the seizure during surgery with a hyperspectral camera



An epilepsy patient with recurring sensorimotor seizures involving the left hand every 10 min, was imaged with a hyperspectral camera during surgery. By calculating the changes in oxygenated, deoxygenated blood, and total blood volume in the cortex, a focal increase in oxygenated and total blood volume could be observed in the sensory cortex, corresponding to the seizure-onset zone defined by intracranial electroencephalography (EEG) findings. This probably reflects very local seizure activity. After multiple subpial transections in this motor area, clinical seizures abated.

When surgery is considered in patients with refractory focal epilepsy, localization of the onset of habitual seizures is warranted as accurately as possible. Routine investigations to estimate the epileptogenic zone include magnetic resonance imaging (MRI) and scalp electroencephalography (EEG). Difficult cases may also require magnetoencephalography (MEG), positron emission tomography (PET), ictal single-photon emission computed tomography (SPECT), and in some instances, invasive EEG with subdural electrodes (electrocorticography, ECoG) to precisely delineate the ictal-onset zone. All these methods require the focus to be estimated, coregistered to MRI, and relocated to the cortex during surgery. During the last diagnostic stage, ECoG can also be carried out intraoperatively. In this report we describe spectral changes measured by intraoperative imaging using an advanced hyperspectral camera system in a patient with cyclic partial sensorimotor seizures, and we postulate that these changes may be a reflection of focal seizure activity.


The patient is a left-handed woman, who had seizures that started suddenly at the age of 13. These consisted of a tingling sensation in the left hand with finger dystonia, sometimes with a sensory march into the left leg and left-sided hemiclonias. From the start there were daily seizures without loss of consciousness, lasting 30–60 s. In between, the function of the hand was normal. A 24-h EEG showed lack of μ-rhythm and presence of spike-and-waves as well as runs of 8 Hz spike-and-waves over the right parietocentral area, which were sometimes accompanied by a sensation in the left hand. A 3T MRI showed cortical enhancement on fluid-attenuated inversion recovery (FLAIR) images near the hand knob. After having tried seven antiepileptic drugs and chronic vagus nerve stimulation without success, she underwent chronic ECoG monitoring and sensorimotor mapping with an 8 × 8 subdural grid over the right primary sensorimotor area (Fig. 1A). During the recording, recurrent accelerating rhythms occurred with a maximum on electrodes 10, 11, 12, 18, 19, and 20, which were interpreted as electrographic focal seizures during which the patient often reported her typical sensation, followed by left hand dystonia. Figure 1C (and Fig. S1) shows such focal seizure activity which disappears after 7 min (Figs. 1D and S2). These seizures showed a remarkably stable pattern throughout the recordings, with an average cycle of 10 min (Fig. 1B). The epileptogenic zone was finally estimated to involve the sensory hand representation on the postcentral gyrus and a neighboring area in the upper part of the parietal lobe. After 1 week of recordings, grids were removed, and in the same operation she underwent resection of the parietal area just behind the postcentral gyrus, and multiple subpial transections within the functional area. Histology evaluation was normal. Postoperatively she was seizure free for 6 months, and then had a hemiclonic seizure every 4–6 weeks. Left hand functional loss was around 15%.

Figure 1.

(A) A 3D reconstruction of the patient's brain and position and labeling of the electrodes. The trepanation is indicated by a blue line. (B) Trends of power (μV2/Hz) of frequencies (fast Fourier) between 0 and 50 Hz (Y-axis) of a subset of the electrodes (i.e., 10, 20, 44, and 61) in 30 min (X-axis) of ECoG. (C) Ten seconds of ECoG of the same electrodes. Electrodes 10, 20, and to a lesser extent 44, are involved in a seizure, whereas electrode 61 is clearly not. (D) Ten seconds of ECoG of the same electrodes but 7 min later; none of the electrodes show seizure activity. See addenda for the ECoG of the electrodes (1–64) from the 8 × 8 grid at these time points. These addenda confirm that these seizures are focal and restricted to the upper and posterior part of the 8 × 8 grid.

Hyperspectral camera

Intraoperatively, after removal of the subdural grids, but before the actual brain surgery, the hyperspectral camera measurements were undertaken. The hyperspectral camera consists of a liquid crystal tunable filter (Cri VariSpec VIS 10 nm; Cambridge Research & Instrumentation, Inc. [CRi], Woburn, MA, U.S.A.) and a monochrome camera with 1392 × 1024 resolution (PixelFly QE, CRi) that is mounted to the surgical microscope. A polarizer is placed in the incident path, and cross-polarized to the tunable filter, to suppress reflections from the surface. During a 7 min scan in our patient, groups of four wavelength images were captured at 1.5 s intervals. The wavelengths (480, 570, 600, and 660 nm) were chosen such that the oxyabsorption and deoxyabsorption spectrum differed at these wavelengths and that the light was scattered at deeper tissue layers while still within the transmission range of the Cri filter (CRi). The spectral and spatial dependency of the system was normalized using a white and dark reference. The oxygenation and deoxygenation values were calculated by modeling the spectral absorption of the light traveling through the top layers of the brain tissue and scattered back to the camera. Using the image I (t, λ) at t = 0 s as a reference, and assuming the main changing chromophores to be oxygenated and deoxygenated hemoglobin, the fitting formula becomes (Klaessens et al., 2011):

display math(1)

where math formula and cHb represent the change in concentration values of respectively oxygenated and deoxygenated hemoglobin and math formula (λ) and AHb (λ) represent their corresponding absorption spectra. The changes in concentration can then be calculated using least square fitting using pseudo inverse matrix calculation. The sum of both concentration changes gives the change in total blood volume ctot.

Four regions of interest were selected in the postcentral and parietal regions. These regions numbered 10, 20, 44, and 61, respectively, correspond to electrode labels of the implanted grids (Fig. 2).

Figure 2.

Images and time plots of hemoglobin concentrations showing local increase in oxy-hemoglobin and total hemoglobin during epileptic seizure.


By calculating the oxygenated, deoxygenated, and total hemoglobin concentrations over time, a local increase of oxygen was seen in the sensory cortex of the hand (position 20) after 4 min, sustaining over 1 min, and decreasing to the initial base level. The deoxy-hemoglobin concentration decreased during that period, whereas the total hemoglobin concentration showed a similar increase but with a slower decline to the baseline. This increase is not visible for the locations 10, 44, and 61.


During the invasive monitoring, a high frequency of focal electrographic seizures were recorded over the postcentral region extending into adjacent areas (Fig. 1B). Seizures were characterized by recruiting rhythmic activity with increase of frequency and rhythmic spiking. Therefore, seizures could be characterized by an increase of power of all frequencies up to 30–40 Hz (Fig. 1B). ECoG trend analysis (absolute power of frequencies up to 50 Hz) of the preselected electrodes showed the recurring aspect of seizures at the electrodes 10, 20, and 44, but not on electrode 61, and within 30 min of representative ECoG three seizures were captured (Fig. 1B). The ECoG at 10:20 h (see Fig. 1C for preselected and Fig. S1 for all electrodes) shows seizure activity on electrodes 10, 20, and 44, whereas the ECoG at 10h27 (see Fig. 1D for preselected and Fig. S2 for all electrodes) shows no seizure activity. A clear difference can be seen for the electrodes 10, 20, and 44, but not for electrode 61. Therefore, a high seizure frequency of focal seizures on the upper part of the postcentral gyrus is demonstrated. The ECoG recording was continued until the patient went to surgery.


The hyperspectral camera showed a local increase in blood perfusion in the cortical surface that lasted for approximately 1.5 min during brain surgery in a patient with cyclic, repetitive electrographic seizures within a pattern of focal periodic discharges. Although ECoG could not be recorded simultaneously, we hypothesize that the increase of blood perfusion in this patient is related to this cyclic electrographic seizure activity. We have three reasons to think so.

First, we know that increase of blood perfusion is related to seizure activity. Evidence for this comes from visual reports of cortical hyperemia during awake surgery. Ictal SPECT shows increase of blood perfusion in the seizure-onset zone (Van Paesschen, 2004; von Oertzen et al., 2011). EEG–functional MRI (fMRI) reveals an increase in the blood oxygen level–dependent (BOLD) signal in parietal and frontal cortices during seizure activity in some types of generalized epilepsy (Belliveau et al., 1991; Moeller et al., 2009). Second, the temporary increase of blood perfusion during the intraoperative measurement in our patient colocalizes with the ECoG seizure activity in the postcentral gyrus as recorded during the chronic recording in this patient. Third, the constant cycle of electrographic seizures (one per 10 min) during wakefulness and sleep as observed in the days prior to surgery, makes it likely that such an event would be captured during a measurement of 7 min.

The electrographic seizures in chronic ECoG were of longer duration and somewhat more regional than the shorter and more local increases of oxygen and blood perfusion measured with the spectral camera (7 min on electrodes 10, 20, and 44 during ECoG vs. 1–2 min at site 20, only during spectral camera measurement). This may be explained by an effect of the propofol anesthesia during surgery. The ECoG findings might also be subject to volume conduction, but this remains speculative.

Of course, simultaneous recordings using ECoG and a hyperspectral camera would be the ultimate proof that ictal ECoG activity coincides with the observed increase of oxygenation. However, such simultaneous recording would prove difficult, since the electrode grids would interfere the view of the hyperspectral camera.

The multispectral imaging technique is limited by the signal-to-noise ratio in the number of acquired wavelengths and exposure times in acquiring the spectral images. Although the cross-polarization suppresses the superficial signal in favor of signal of deeper tissue layers, about 80% of the light is not used. In addition, the transmission of the tunable filter varies from 25% at 480 nm to 50% at 660 nm. Since 2009, hyperspectral tunable light sources have become available based on laser or light-emitting diode (LED) technology, thereby making faster and interoperative imaging possible. Combined with constant increases in computer power, interoperative and real-time imaging should be possible in the near future.

Intraoperative spectral imaging may be used to localize seizure onset in patients with a high (subclinical) seizure frequency, as can sometimes be seen in the ECoG of focal cortical dysplasia (Ferrier et al., 2006).


Conventional focus localization techniques in presurgical evaluation of epilepsy include scalp EEG, MEG, SPECT, and PET, often coregistered to MRI, and relocated to the cortex during surgery for neuronavigation. Intraoperatively, ECoG can be used for final tailoring of the resection. The hyperspectral camera is a new intraoperative localizing tool with some potential. We have shown the ability to monitor a local epileptic event in the cortex that most likely reflects seizure activity by directly observing local increase of blood perfusion. This shows that we can visualize a very local and subtle brain oxygen increase at high resolution, hopefully in real-time in the near future.


At the time of the clinical measurements in 2009, all authors were paid by the University Medical Center of Utrecht, The Netherlands.


None of the authors has any conflict of interest to disclose. We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.