CURRENT INTRAOPERATIVE METHODS used to maximize the extent of tumor removal are limited to intraoperative biopsies, ultrasound, and stereotactic volumetric resections. A new technique involving the optical imaging of an intravenously injected dye has the potential to localize tumors and their margins with a high degree of accuracy. In a rat glioma model, enhanced optical imaging was performed and indocyanine green was used as the contrast-enhancing agent. In all 22 animals, the peak optical change in the tumor was greater than in the ipsilateral brain around the tumor and the contralateral normal hemisphere. The clearance of the dye was significantly delayed to a greater extent in the tumor than in the brain around the tumor and the normal brain. After attempts were made at complete microscopic resection, enhanced optical imaging of the tumor margins and the histological samples demonstrated a specificity of 93% and a sensitivity of 89.5%. Enhanced optical imaging was capable of outlining the tumor even when the imaging was done through the cranium. The optical imaging of rat gliomas with a contrast-enhancing dye is able to differentiate between normal brain and tumor tissue both at the cortical surface and at the tumor margins. The application of these studies in an intraoperative clinical setting may allow for the more accurate determination of tumor margins and may increase the extent of tumor removal.
The discrepancy between intraoperative observations of glial tumor margins and diagnostic imaging studies has been emphasized in recent years with correlative studies of tissue stereotactically obtained from regions adjacent to the imaged tumor nidus (17,28–30). Tumor cells have been found several centimeters beyond the contrast-enhancing lesion on both computed tomography and magnetic resonance images in high-grade glial tumors. Likewise, pleomorphic nonreactive astrocytes have been identified in biopsies taken from areas beyond T2-weighted hyperintense magnetic resonance regions surrounding noncontrast-enhancing, low-grade astrocytic gliomas.
Because of the diffuse and infiltrative nature of glial tumors, it is often difficult to rely on the gross appearance and consistency of the lesion at the resection margins to achieve a radical tumor resection. Intraoperative ultrasound provides an accurate real-time assessment of the location and consistency of both high- and low-grade glial tumors (4,9,15). However, as the resection proceeds, spurious echogenic signals are produced by the contused brain and artifacts created by the irrigation fluid at the resection cavity interface (32). In addition, previously irradiated brain and long-standing edema may cause reactive astrocytosis, which produces widespread echogenic signals, making intraoperative ultrasound unreliable as a means to define the tumor margin.
Additional methods used with some success in an attempt to enhance tumor localization during surgery involve frozen-section biopsies and smear preparations, intravenous dye administration with fluorescein (39) or porphyrin derivatives (3,5,26,27,31,40,43,44,49,51,53) to demarcate a disrupted blood-brain barrier, and intraoperative computed tomographic imaging (7). Recently, an experimental study completed at our institution suggested that indocyanine green (ICG) (23) (Becton Dickinson Co., Cockeysville, MD), injected intravenously at one to two times the 50% lethal dose, visibly stains a rat glioma dark green (23). Because ICG is tightly bound to serum albumin, the tumor staining is possibly commensurate with the disruption of the blood-brain barrier.
The purpose of this study was to take advantage of the sensitivity of a charge-coupled device camera and enhanced optical imaging of rat gliomas using ICG to determine whether normal and edematous brain could be distinguished from tumor tissue. We hypothesized that the dynamic nature of the perfusion of an absorption dye through the brain would allow the differentiation of normal brain from tumor tissue. With a spatial resolution of the optical images below 20 μm2/pixel (22), even small areas of residual tumor could potentially be identified. The optical imaging studies were able to differentiate normal brain from tumor tissue with doses 1/100th that used for the previous animal studies and were even able to localize tumors through the cranium. Optical imaging in the resection cavities accurately determined positive and negative biopsies with a high sensitivity and specificity. Portions of this work have been presented previously (19,20,24).
MATERIALS AND METHODS
Rat glioma model
The rat glioma model, as approved by the Animal Care Committee at the University of Washington, has been previously described in detail (14,46–48). Briefly, an ethyl-nitrosourea-induced F-344 rat cell line (36B-10) was developed from a clonal population of a spinal malignant astrocytoma. The glioma cells were maintained in vitro in Weymouth's medium containing 10% fetal calf serum. For the stereotactic implantation, the rats were anesthetized with intraperitoneal injections of 25 to 30 mg/kg of sodium pentobarbital. Viable cells (5 × 104) were implanted stereotactically into the right frontal hemisphere of 30 syngeneic, 140- to 160-g female rats (Simonsen Labs, Gilroy, CA). The stereotactic coordinates for the right frontal lobe implantation were 4.5 mm anterior to the frontal-zero plane, 3 mm right from the midline, and 6 mm deep. The cells were injected through a 27-gauge needle, which remained in place for 30 seconds postinjection; the needle was then removed, and the hole was covered with bone wax. The scalp was then sutured, and the animals were observed for 3 to 4 hours until they returned to normal activity and feeding. Animals were studied between 10 and 14 days after tumor implantation. From previous studies (16,46–48), the animals begin to show clinical symptoms from mass effect by 16 to 19 days, including decreased activity and diminished feeding.
The optical imaging study comprised results obtained from 22 animals. Animals were initially anesthetized with 2% isoflurane. The femoral vein was cannulated for the administration of intravenous drugs, including the dye ICG. Anesthesia was maintained with an intraperitoneal injection of α-chloralose (50 mg/kg) and urethane (160 mg/kg), and then, usually, one subsequent dose at 20% the initial concentration was needed. After the animals were placed in a stereotactic holder, the scalp was anesthetized with 1% lidocaine with epinephrine and reflected. Imaging studies were then carried out either before or after the removal of the cranium. The dura was reflected over the bony opening, which extended from in front of the coronal suture, anteriorly, to the lambdoidal suture, posteriorly. The tumor typically occupied the anterior one-half to two-thirds of the right hemispheric exposure. The compressed brain without any tumor infiltration by histopathological examination was defined as the brain around the tumor (BAT) to separate it from the normal hemisphere on the contralateral side.
After the completion of the cortical surface imaging, the operating microscope was used to attempt the gross total removal of the tumor. After the resection, the optical imaging was repeated and the images were processed without knowledge of the extent of resection. Sites were then chosen on the basis of optical imaging that should be positive or negative for tumor. These sites were then biopsied and placed in numbered vials for later blinded histological analysis (see Materials and Methods). The rats were then sacrificed with a lethal bolus of pentobarbital.
The intravenous dye used for these experiments was ICG (Cardio-Green; Becton-Dickinson), which is a tricarbocyanine dye with a molecular weight of 775 g/mol. After intravenous injection, ICG is rapidly bound to plasma proteins, including albumin and globulins (α-lipoproteins) (6,12,42). ICG is nearly completely eliminated from the liver by hepatic parenchymal cells, with very minimal if any extrahepatic uptake, enterohepatic circulation, or renal excretion (6) (for a review, see Reference (42)). ICG is eliminated in healthy human subjects at a rate of 18 to 24% per minute, with a normal half-life of 2.5 to 3.0 minutes. In human subjects, only 4% of the dye remains in the serum 20 minutes after injection (42). In plasma or blood, the maximal absorption of ICG is shifted from 770 to 780 to 790 to 805 nm. In both animals and humans, ICG has a 50% lethal dose after intravenous administration of 50 to 75 mg/kg. For these studies, the ICG was made up fresh for each experiment with sterile water, pH 5.5 to 6.5, and the typical dose was 1 mg/kg. A 1-ml aliquot of the dye was always injected over 1 second during the storage of the last control image.
The experimental setup is shown in Figure 1 (21,22). Light from a 100-W tungsten-halogen bulb regulated by a direct current power supply was passed through a longpass filter (690 nm), and, with a right-angled prism, it was reflected through a 50- to 200-mm objective lens onto the cortical surface. The reflected light was then collected by the same objective lens and focused by a projection lens onto the surface of a charge-coupled device camera (COHU 6300; San Diego, CA). The imaging apparatus was attached to the stereotactic frame, which was rigidly fixed to a vibration isolation table. Synchronization of the image acquisition to heartbeat and respiration was not necessary secondary to the stability of the experimental setup and specially designed automatic warping algorithms, which compensated for small amounts of movement.
Images (512 × 480 pixels) were acquired at 30 Hz and digitized at 8 bits (256 grey levels) (Imaging Technology Inc., Woburn, MA). Every 2 seconds, a single image comprising 30 averaged frames was collected (1 s) and then stored (1 s). Control images were collected before the injection of the dye and then for 2 to 10 minutes after dye injection. An intertrial interval of 20 minutes allowed the optical changes to return to baseline. The initial control images of each trial were subtracted from each other to ensure that the baseline starting point of each trial was equivalent.
A single control image was chosen, and then the rest of the controls (four to six images) and each of the post-dye injection images were subtracted from the chosen control. The resultant image was then divided by the original control image and multiplied by 100 to give a relative percentage difference image for the entire sequence before and after the dye injection. The optical change that occurred between separate control images was 0.2 to 0.7%, whereas the peak changes resulting from the dye injection were in the range of 5 to 40%. These relative percentage difference changes, which are calibrated after the analog offset has been adjusted, are 6- to 10-fold greater than the absolute percentage difference changes, as calibrated by neutral density filters. The control percentage difference image and that of a post-dye injection percentage difference image are represented by a pseudocolor scale showing linear changes in the magnitude of optical change from near 0%, represented as black, to peak changes between 8 and 40%, shown as red.
The spatial resolution of an individual pixel in the images averaged 13.5 × 11.7 μm2. Boxes measuring from 15 to 30 pixels per side were then drawn on the images. The boxes correlated with biopsy-proven tumor, BAT, or nontumor tissue, as determined by histopathological examination. The average percentage change in the individual boxes was calculated from the raw data and was used to graphically demonstrate the optical changes over time in the different types of tissue.
The biopsy specimens were fixed in 10% paraformaldehyde, embedded in paraffin, sectioned, stained with Nissl substance, hematoxylin, and eosin, and mounted. All specimens were blindly read by one of the authors (AMS) without knowledge of the animal or biopsy site. The histological specimens were then labeled either positive or negative for tumor. Correlation of the imaging results and the histological examination were made for both the sensitivity and the specificity of the optical imaging identification of residual tumor. χ2 or Student's t-tests were performed to determine significance.
Although the majority of the imaging studies were done with dye concentrations of 1 mg/kg, the concentration of the dye injected was varied in two experiments to test whether different dosages of dye changed the characteristics of the tumor identification. A comparison during a single trial of dye injections over a 10-fold range is shown in Figure 2. At both dosages (0.1 and 1.0 mg/kg), a 1-ml aliquot of dye was injected over 1 second. The percentage differences of optical changes from baseline for the tumor tissue (tumor), brain around the tumor (BAT), and normal brain (norm) are shown. At an injection concentration of 0.1 mg/kg, there is a much higher peak optical change in the tumor tissue compared with that in the BAT and that in the normal brain (Fig. 2, A and C). When the dosage of the injected dye was increased to 1.0 mg/kg (Fig. 2, B and D), there was a 10-fold increase in the peak optical change in the tumor tissue compared with that with the 0.1 mg/kg dose but only a 6-fold rise in the peak optical change in the surrounding tumor and only a 3-fold increase in the normal brain.
Evidence for the significance of these optical changes is present when the absolute magnitude of the percentage difference optical change is compared for the three areas (tumor, BAT, and normal). In Figure 3A, the 0.01 mg/kg dosage shows no significant optical changes but does act as a baseline to demonstrate the inherent noise in the system. The changes from baseline are less than 1.0%. When the concentration of ICG was increased to 0.1 mg/kg, the peak optical changes were near 6% in the tumor compared with less than 4% for the BAT and only 2.3% in the normal brain. At the dosage typically used for the rest of the experiments, 1.0 mg/kg, there is a large increase in the optical change in the tumor but less so in the BAT and normal brain (Fig. 3B).
Rat glioma imaging studies
The time course of the dye perfusion through the tissue had a dynamic nature. All of the dye injections for these experiments were done at a dosage of 1 mg/kg. The optical imaging of indocyanine dye perfusion in 16 separate trials from the cortical surface in nine different animals demonstrates the dynamic nature of the optical changes. The cortical surface of the tumor, the BAT (upper left), and the normal hemisphere (right) are shown in Figure 4A. Sequential images obtained before injection and 2, 6, 30, and 90 seconds after injection are shown in Figure 4, B through F, respectively. The peak optical changes occur 6 seconds after injection in all three regions (Fig. 4C), but, after the normal hemisphere begins to clear the dye, the tumor tissue maintains a large optical signal due to lack of dye clearance. These optical changes anatomically localize the tumor site (Fig. 4F).
The optical signal begins to change within the first 2 to 3 seconds after injection and peaks 6 seconds after injection in all three areas: the tumor tissue, the BAT, and the normal brain (Fig. 5). However, the three different tissue types are differentiated by the rate of rise over the first 4 seconds, the peak optical change reached, and the eventual plateau that occurs after the first 30 seconds. The tumor tissue has a significantly greater peak optical change (40.5 ± 9.6%) than either the BAT (16.4 ± 6.8%) or the normal brain (9.7 ± 4.7%). A significantly higher peak optical change was found in the tumor tissue compared with that in the BAT or the normal brain (P < 0.001).
The key feature for the differentiation of normal from abnormal tissue is the dynamic distribution of the dye into three different tissue compartments (normal, BAT, and tumor). In all 16 animals, there were prolonged increases (>2 min) in the optical signal in the tumor after normal brain and BAT tissue had returned toward baseline. Finally, the normal and BAT regions had a significantly different dye uptake (rise time: normal, 2.4%/s; BAT, 4.0%/s; P < 0.01). The dynamic features, specifically the different rise times of the dye uptake and the time needed for clearance, are critical for determining the type of tissue involved both in imaging the cortical surface and in imaging resection margins.
Resection margins in rat gliomas
These animal studies provided the opportunity to look at the feasibility of imaging resection margins once all visible tumor had been removed. In a single study from one animal, the optical image obtained 2 minutes after dye injection demonstrates regions of delayed dye clearance (red areas in Fig. 6, C and D). The boxes in Figure 6A were the areas chosen for biopsy. The more rapid rate of rise seen in the cortical surface imaging of tumors is still present for the resection margins that were positive for tumor compared with the normal brain (Fig. 7). Again, significant differences between the tumor and normal brain existed for the rate of rise, the peak optical change, and the plateau 60 seconds after dye injection (all P < 0.001).
The increase in the magnitude of the error bars for the tumor tissue over time is the result of two different plateau characteristics for the tumor specimens. The first type of curve of optical changes was much like that of tumor tissue imaged from the cortical surface, with a very flat plateau for 2 to 4 minutes after the dye injection before the return to baseline. The second type of optical change at the tumor margin was one of increasing signal changes that did not return to baseline for 5 to 10 minutes and, in two cases, never reached the original baseline values.
The sensitivity and specificity of the optical imaging were determined for the 34 samples (n = 12 animals). Of 15 biopsy sites deemed negative for tumor by the optical imaging, 14 were clear of tumor by histological analysis (specificity, 93%). Most of the specimens that were negative for tumor were taken from the posterior wall of the tumor resection cavity or the depth of the cavity. Of 19 biopsy sites deemed positive for tumor by the optical imaging, 17 of the biopsy specimens were read as positive for tumor (sensitivity, 89.5%). The two sites that were negative for tumor on histology but positive for tumor by the optical imaging had increased cellularity on histopathological examination. Even though the average biopsy site was quite small (0.5 to 2.5 mm2), the results of the comparison of tumor and normal brain biopsies were clearly significant (P < 0.001).
Imaging through intact cranium
Far-red wavelengths are known to penetrate through bone. The imaging of tumor tissue was attempted through the intact cranium of the rat (Fig. 8A). The extent of tumor identified is not as accurate as with the cortex exposed; however, the area lying beneath the cranium with tumor tissue was easily localized and continued to concentrate the dye over minutes. Initially, after dye injection, the area of the tumor demonstrated a much larger signal than than that of the normal brain of the contralateral hemisphere (Fig. 8B), and the overall optical change peaked 8 seconds after dye administration (Fig. 8C). One minute after dye injection, the dye had been cleared from the normal brain and the only residual activity remained in the tumor tissue and the sagittal and transverse sinuses (Fig. 8D).
The time course of the optical changes imaged through the cranium from 10 trials in four animals, as determined by the average optical change in a box placed directly over the tumor and over the normal hemisphere, is shown in Figure 9. The peak optical changes for the tumor imaged through the cranium were 13.9 ± 3.9% and were significantly greater compared with those of the normal brain of 7.8 ± 2.3% (P < 0.01). The plateau phase 60 seconds after dye injection was also significantly greater in the tumor tissue (4.8 ± 0.8%) compared with that in the normal brain (1.8 ± 0.6%) (P < 0.001).
Dynamic changes in the optical signal
The dynamic differences in the dye perfusion through normal brain, BAT, and tumor tissue were easily distinguished by the use of optical imaging. The difference in optical changes in the three tissue types that allow differentiation could be multifactorial, including slower transit times through tumor tissue, greater extravasation of protein-bound dye through leaky tumor capillaries, more rapid uptake of the dye by the tumor tissue, preferential uptake of the dye by tumor cells, and increases in the absorptive properties of the dye when bound in the extravascular spaces of the tumor tissue.
A number of studies have looked at rat and human gliomas and their microvasculature compared with normal cortex (1,2,10,11,16,25,38,41). In both experimental and clinical studies, blood flow in tumor tissue is slower and more variable than in normal tissue. The differences between the tumor and normal brain have been attributed to tumor location, degree of neoplastic infiltration, and necrosis (25). In the rat glioma model used in this study, previous work has shown that blood flow is slower in the tumor tissue than in the contralateral normal hemisphere (16). The blood flow in the BAT was intermediate between that in the tumor tissue and that in the normal brain. With cultured spheroids of C6 astroglial cells transplanted into rat brain, blood flow was also slower in tumor tissue than in normal brain tissue (11). Further studies have shown that the microvessel volume fraction was equivalent between the tumor and normal brain; however, because only 50% of the tumor was actively perfused, the surface area of perfused microvessels in the tumor was one-half that of the normal brain (25). These changes could account for a slower flow of the dye through the tumor compared with that in the normal brain and could also lead to more rapid clearance of the dye by the normal brain in contrast to the tumor.
The permeability of tumor capillaries is much higher than in the normal brain (31,33,34,55). The leakiness of these capillaries leads to the extravasation of larger particles, resulting in edema and an increase in the interstitial pressure surrounding tumor microvessels. Because tumor microvessels do not contain the normal arteriole smooth muscle (25,45), they lack the local control of pressure gradients. These two characteristics lead to a stasis of flow in the tumor tissue. The overall effect on the dye perfusion is longer transit times than in the normal brain, potentially prolonging the duration of the large optical signal from tumor tissue. These findings support our dynamic changes in optical signals from the tumor and normal brain during dye perfusion. The optical changes show a more rapid entrance of the dye in the tumor compared with that in the normal brain, but the slower transit time in the tumor tissue results in prolonged increases in the optical signal in the tumor. Also, because the tissue surrounding the tumor would be expected to have increases in interstitial pressures above that in the normal brain but without the leaky capillaries and other microvasculature changes, it is not surprising that the BAT has an intermediate duration of optical changes.
Whether a more rapid clearance mechanism of the dye from the normal brain occurs or whether the dye is preferentially sequestered by tumor cells is unclear. Other dyes have been tested in animal models involving blood-brain barrier breakdown, e.g., cold cortical lesion and air embolism (54). Sodium fluorescein, which remains essentially free in serum, was contrasted to Evans blue dye, which is bound almost entirely to plasma proteins, primarily albumin. In both animal models, the fluorescein was rapidly cleared from both the normal brain and the regions with the blood-brain barrier breakdown, whereas the Evans blue remained in the extravascular space for a prolonged period of time. Because the elimination of ICG from the blood is rapid, 18 to 24%/min, the gradual increase in the optical signal above even the initial early peaks suggests that the dye is remaining in the extravascular compartment.
In the blood, ICG is more than 95% bound to plasma proteins, much like Evans blue. The Evans blue-plasma protein complex is actually taken up by the astrocytes in the region of blood-brain barrier breakdown (54). If the same is true of ICG, this mechanism would adequately explain the delayed peak in the optical signal in the rat glioma studies. However, these comparisons can still not answer the question of possible increases in the absorptive capacity of ICG when the dye is bound in the tumor extracellular matrix.
Comparison with other techniques
Photodynamic therapy has been applied to human tumors with some promise (26,27,31,37,40,53). The potential use of hematoporphyrins in photodynamic therapy depends on the preferential sequestration of the dye by tumor cells. With specific wavelength lasers, tumor cells that contain the dye are selectively destroyed. Imaging techniques have been used to look at laser-induced fluorescence differences between tumor and normal tissue in a rat glioma model (44). These studies, like this report, rely on optical differences between the normal and tumor tissue enhanced with a dye. In this report, the ICG is used as an absorption dye imaged at far-red wavelengths, whereas the other study used a fluorescent dye and much shorter wavelengths. Although both studies have their respective advantages, the previous report relies on a point source (>500 μm2 spatial resolution) for acquiring data and this report uses a charge-coupled device camera (<20 μm2/detector spatial resolution) attached to the operating microscope. In our studies, the camera provides excellent spatial resolution and the potential for real-time information. When combined with a heads-up display to overlay the optical image on the image being viewed by the neurosurgeon through the operating microscope, the exact location of the residual tumor can be identified with optical imaging and resected in a standard fashion (19,20,24,28). The fluorescent dyes, although having the advantage of being used as a photosensitizer and capable of analyzing tissue samples, are dependent on the unique uptake by the tumor tissue of hematoporphyrin, which is not homogenous in animal brain tumor models (3) and has been variable in human astrocytomas (37).
Imaging of tumors
The optical imaging of far-red and infrared light is currently developing into a useful tool for noninvasively monitoring physiological changes in the brain (35,36). Optical imaging through the shaved-down cranium of a monkey has shown the functional architecture of the ocular dominance columns in the visual cortex (13,18,52). Infrared light is already being clinically used to monitor cerebral oxygen saturation and oxygen delivery (36). Light penetration into tissue is determined by the optical characteristics of the tissue and the wavelength of the light (8,50). The penetrance of light has been measured at different wavelengths in human brain tissue. The penetrance is defined as the depth where the light intensity falls to 1/e or 37% of its initial value. The penetrance of light at 660 nm into the human cortex is 1.2 to 1.6 mm; at 710 nm, it is 1.5 to 1.7 mm. At infrared wavelengths out to 1100 nm, the light (650 to 1100 nm) even penetrates the human cranium to depths of several centimeters.
Recent experimental data have shown the ability to noninvasively image the time course of ICG flow through the vasculature of the cortical surface (35,36). Even though light scattering limits the spatial resolution when the optical imaging is performed through the cranium, our findings demonstrate the potential for the noninvasive preoperative imaging of tumors. These optical imaging techniques, when combined with the optical imaging of neuronal activity, may also allow the noninvasive monitoring of functional activity as it relates to the growth of tumors (19,20,24).
Through the use of optical imaging techniques, the diagnosis, surgery, and management of patients with primary brain tumors may be improved. Preliminary results from human patients with different grade astrocytomas demonstrate the ability of enhanced optical imaging to grade astrocytomas and identify residual tumor in the resection cavity (19,20,24). Enhanced optical imaging also has the ability to provide real-time feedback in the operating room. Therefore, this new methodology may become a critical adjunct in guiding the resection of high- and low-grade gliomas and metastases throughout the central nervous system.
This work was supported by the American Association of Neurological Surgeons Research Foundation Fellowship, a Klingenstein Fellowship in the Neurosciences, and a Washington Technology Center Initiative Grant (to MMH) and by National Institutes of Health Grant KO8 NS01253-01 (to MSB, American Cancer Society Professor of Clinical Oncology 071). Cardiogreen was kindly provided by Theodore Carski, M.D., of Becton Dickinson, Inc. We are grateful for the technical and photographic support provided by Julie Schoenfield, Etorre Lettich, Carol Robbins, Janet Clardy, and Paul Schwartz. We thank H. Richard Winn, M.D., for continued support. MMH and DWH are involved in developing the optical imaging of tumors for commercial application.
The authors herein describe a technique by which to enhance the excision of cerebral tumors, in particular, cerebral gliomas. It is recognized that there is still debate regarding the oncological value of cytoreductive surgery and whether maximizing tumor resection will improve patient survival. It is known that most gliomas have approximately 1011 cells at the time of diagnosis (about 100 g), and, therefore, even a 99% tumor excision will result in only 2 logs of tumor reduction, leaving approximately 109 tumor cells. I do think that every possible method should be used to maximize the safe removal of tumors so that adjuvant therapies can be used on a minimal tumor burden. The technique described is yet another way to achieve a maximal tumor resection, and I am sure that improvements can be expected in both the localizing dye and the imaging techniques, which will allow the method to be widely used.
It is unlikely that this technique will be of value in removing those “pockets” of tumor that are missed by conventional surgery. The technique, however, will not enhance tumor resection in the “brain adjacent to tumor” region, in which the normal brain is invaded by microscopic islands of tumor. The challenge in neuro-oncology is to find therapies that will target these cells that are now largely responsible for tumor recurrence.
Andrew H. Kaye
Parkville, Victoria, Australia
No imaging study shows the histological margins of intra-axial glial tumors. Computed tomography does not, magnetic resonance imaging does not, and the method described in this article does not. All of these modalities and methods show an epiphenomenon from which we infer information about tumor boundaries. Contrast enhancement on computed tomography and magnetic resonance imaging demonstrates disruption of the blood-brain barrier as the result of neovascularity; newly formed blood vessels do not form tight junctions. Indocyanine green bound to serum albumin also shows us blood-brain barrier breakdown, as does fluorescein, which was suggested for a similar purpose over 20 years ago. Then, a neurosurgeon needed a Wood's & by the method presented in this article, a liquid crystal display camera and heads-up display on the microscope would be required.
In a glial neoplasm, methods that target blood-brain barrier breakdown localize the tumor tissue component (which induces neovascularization). This can be removed by a number of methods. In human gliomas, unfortunately, that is only part of the story.
The problem in most human glial tumors is isolated tumor cells that invade the surrounding intact parenchyma, which cannot be selectively extracted by surgical means with any technology known at this time. This parenchyma is usually, but not always, edematous (resulting in hypodensity on computed tomography and prolongation of T1 and T2 on magnetic resonance imaging), but the blood-brain barrier is usually intact and is not associated with neovascularity. The method described in this article would, therefore, not detect this. Resecting this part of the “tumor” would be resecting intact, functioning, albeit sick, brain parenchyma.
It is always interesting to see which methods get picked up by the neurosurgical community and become the standards of practice. The fluorescein method did not, although it was similar to the method described here and was available at a time when computed tomography- and magnetic resonance imaging-based volumetric stereotactic methods were not available and when few neurosurgeons questioned the benefit of aggressive resections in glial neoplasms. Nonetheless, the method and technology described in this article must still be tested in human gliomas.
Patrick J. Kelly
New York, New York
This work by Haglund and colleagues uses and demonstrates the power of the image processing technology available today. The extreme sensitivity (and specificity) of the methodology has been shown in functional applications and is now illustrated with structural pathology. Although, in this study, optical imaging may be detecting indirect evidence of tumor, its potential use in clinical neurosurgery for tumors is nonetheless considerable.
In addition to being able to use sophisticated signal processing algorithms, which enable the success described, optical imaging is conceptually powerful, given the inherently visual nature of most neurosurgery. Our field already uses and relies on superb optical and imaging equipment, as in our operating microscopes and camera accessories, and methodology as described in this report may eventually be easily incorporated into everyday practice. Heads-up display technology in the operating room can integrate previously obtained imaging studies as well as both retrospective and real-time physiological data, and, as the authors have noted, this will provide seamless feedback to the surgeon.
David W. Roberts
Lebanon, New Hampshire