Neurosurg Rev (2015) 38:59–70 DOI 10.1007/s10143-014-0578-y
REVIEW
Resection of supratentorial gliomas: the need to merge microsurgical technical cornerstones with modern functional mapping concepts. An overview Giannantonio Spena & Pier Paolo Panciani & Marco Maria Fontanella
Received: 13 January 2014 / Revised: 22 April 2014 / Accepted: 22 June 2014 / Published online: 21 October 2014 # Springer-Verlag Berlin Heidelberg 2014
Abstract Although surgery is not curative for the majority of intracranial gliomas, radical resection has been demonstrated to influence survival and delay tumor progression. Because gliomas are very frequently located in eloquent or more generally critical areas, surgeons must always balance the maximizing resection with the need to preserve neurological function. In this overview, we tried to summarize the recent literature and our personal experience about (1) the benefits and limits of using preoperative anatomical and functional neuroimaging (anatomical MRI, DTI fiber tracking, and functional MRI), (2) the issues to consider in planning the surgical strategy, (3) the need to thoroughly understand microsurgical techniques that enable a maximal resection (subpial dissection, vascular manipulation, etc.), (4) the importance of individualizing surgical strategy especially in patients with gliomas in eloquent areas (the role of neuropsychological evaluation in redefining eloquent and non-eloquent areas), and (5) how to use intraoperative mapping techniques and understand why and when to use them. Through this paper, the reader should become more familiar with a comprehensive panel of techniques and methodologies but more importantly become aware that these recent technical advances G. Spena : P. P. Panciani : M. M. Fontanella Neurosurgery Department, Spedali Civili and University of Brescia, Brescia, Italy P. P. Panciani e-mail: [email protected] M. M. Fontanella e-mail: [email protected] G. Spena (*) Neurosurgery Department, Spedali Civili of Brescia, Piazzale Spedali Civili 1, 25120 Brescia, Italy e-mail: [email protected] G. Spena e-mail: [email protected]
facilitate a conceptual change from classical surgical paradigms toward a more patient-specific approach. Keywords Brain neoplasms . Diagnostic imaging . Diffusion tensor imaging . Electric stimulation . Quality of life . Magnetic resonance imaging . Gliomas
Introduction Surgical treatment of supratentorial gliomas has changed markedly in the last decades. It has been demonstrated that surgery influences survival of high-grade gliomas (HGGs) especially in association with adjuvant treatment (chemotherapy and radiotherapy) [1–3]. Very recently, a larger resection was associated with longer survival or delayed relapse in low-grade glioma (LGGs) patients [4–8]. Globally, a policy of more extensive resection has evolved, and many authors have advocated widening the indications for surgery. Consequently, neurosurgeons are often asked to perform the widest possible resection. Moreover, the widespread use of early postoperative MRI follow-up has led to a more objective evaluation of the extent of resection (EOR) that no longer relies on only the surgeon’s perception of the EOR [9, 10]. The neurosurgeons can be aware of the actual EOR, evaluate those areas with surgically induced damage (i.e., ischemia), better correlate the intraoperative anatomical and functional boundaries of the resection with a detailed postoperative image, and gain insights about the anatomo-functional organization of a single patient. The goal of wider resection is easily achievable in some locations; however, in other sites, such as eloquent area tumors (EATs) or more generally critical areas, a wide resection is more difficult. Still, an oncologically desirable resection always should be balanced with the need to preserve neurological function, in both in HGGs and LGGs, although for different reasons. In HGGs, postoperative impairment and the potentially detrimental effect of adjuvant therapy
60
(especially radiotherapy) were demonstrated to be crucially related to shorter survival [11–13]. On the other hand, patients with LGGs are often young, and their mean survival is far longer than in patients with HGGs; therefore, they are usually expected to recover their normal function after surgery. For EATs, it has been shown that cortical and subcortical electrical stimulation (CSES) can guarantee a maximal resection while minimizing neurological sequelae [14–19, 82]. The planning of this approach requires the correct interpretation of preoperative functional imaging such as functional MRI (fMRI) and diffusion tensor imaging fiber tracking (DTI-ft) that can add useful information about individual anatomical and functional characteristics. In addition, the neurological and neuropsychological status of the patient as well as his/ her expectations and needs must be clarified preoperatively to tailor the surgical resection and thereby achieve the best quality of life for that specific patient. Regardless of whether the tumor is located in eloquent areas or not, it is of highest importance that the manipulation of vessels, arachnoid, and pial planes is conducted with mastery. The concept of “debulking” the tumor has progressively been abandoned in favor of a more tailored resection based on accurate knowledge of the brain-tumor interface anatomy. Consequently, an accurate analysis of the preoperative anatomical imaging is warranted. The aim of this paper is to elucidate those aspects of the surgical planning and operative technique that are required to achieve a large but safe resection and to show how to maximally exploit functional mapping concepts in order to treat more complex EATs.
Surgical planning: what can we get from anatomical and functional MRI? Currently, neurosurgeons have the many options to preoperatively evaluate brain anatomy and function, tumor morphology, and brain/tumor interface that enable them to thoroughly plan most of the surgery. The possibility to preoperatively define the tumor growth pattern (diffuse or bulky, the presence of a contrast enhancing rim, and the degree of the distortion of surrounding brain) and to reliably preview the histology (slow or rapidly evolving tumor) is important because these characteristics influence the type of interface with the healthy brain [20]. In our recent paper [18], we were able to demonstrate that EATs that extend toward subcortical tracts were less amenable to gross total resection and this situation was not forcedly related to the tumor volume. Similarly, other investigators demonstrated that the presence of infiltrated or displaced fascicles on the preoperative DTI-ft was predictive of a lower probability of total resection, even in tumors with a smaller preoperative volume, in which an extensive resection was expected [21].
Neurosurg Rev (2015) 38:59–70
In addition, morpho-structural factors related to the biology of the tumor are determinants of the EOR. The tumor-brain interface seems to influence the resection by creating better dissection planes, either for non-enhancing than for enhancing tumors. Although gliomas are infiltrating tumors, the growth characteristics of tumors that infiltrate or dislocate white matter tracts can influence the ease of resection. Similarly, Talos et al. [22] found that a large tumor volume associated with diffuse tumor margins and oligodendroglioma or oligoastrocytoma histopathologic type was predictive of incomplete resection. The strategy of surgeons is to perform the largest resection possible; therefore, the tumor’s tridimensional shape must be considered. Modern MRI can clearly delineate the sulcal, gyral, and vascular anatomy and help create a map of the tumor and surrounding brain. Several authors demonstrated how specific MR sequences, such as the T2-reversed and STIR, can add very reliable details of cortical and sulcal anatomy [23, 24]. The arterial and venous system of the tumor and the principal arterial and venous branches passing by the tumor are exactly displayed on T2 or T1 postgadolinium sequences (Fig. 1). Neuronavigation devices are very useful instruments to precisely locate brain lesions and help define the craniotomy size and position. Moreover, these devices can locate some structures at the very beginning of the procedure (vessels and sulci). Unfortunately, this tool cannot function once CSF outflows and the tumor is debulked, because the framework is definitively altered. Hence, most of the anatomical details are obtained from preoperative MRI or, when available, from intraoperative MRI. If the tumor is sited near a major cistern, such as the sylvian fissure, ambient cistern, or crural cisterns (tumors of the mesiotemporal region), a T2 sequence can help to understand whether a pial plane is preserved and whether the tumor has a close relationship with the vessels or nerves. When dealing with a tumor that is located purely in the white matter, sulcal anatomy has to be clearly defined by MRI to determine the most direct pathway to the tumor. This route can be either transgyral or transsulcal because no strong evidence has proven the safest route (read below). Neuronavigation can be used to choose the most direct path to the tumor. Esposito et al. [25] described a useful and less expensive method to locate cortical structures that they named MRI-based corticotopography. They exploited volumetric MRI reconstruction of the cortical anatomy of the subject that was uploaded to a laptop and matched with an intraoperative image. Despite the method utilized to locate subcortical tumors, if the mass is sited under eloquent areas, it is advisable to choose the transcortical route with direct mapping [26] (see below).
Neurosurg Rev (2015) 38:59–70
61
Fig. 1 a–d Preoperative MR showing a postrolandic tumor invading the subcortical white matter with poorly defined margin (blue arrow). Turquoise arrows point out two rolandic veins passing over the tumor. e, f
DTI-ft reconstructs a CST in a very close position to the tumor. g, h, i, l, m, n Postoperative MR shows gross total resection of the tumor. Note the veins that were spared during resection (green arrows)
fMRI and DTI-ft
area. The spots of activation during fMRI scanning are greatly affected by the statistical threshold chosen for data evaluation. Even with the use of one or more fixed statistical thresholds, blood oxygenation level-dependent (BOLD) signal intensities and cluster sizes differ significantly from one patient to another and between different paradigms (e.g., foot movement, hand movement), even when examinations are standardized. Studies that compared presurgical fMRI findings with a reference procedure, such as CSES performed in patients with lesions near the central sulcus [27–29], reported high concordance; percent agreement ranged from 83 to 92 % [19, 30, 31, 83, 85]. However, for evaluation of language areas, the utility of fMRI is diminished as demonstrated by our results (42.8 %) as well as by previous reports of variable sensitivities and specificities ranging from 59 to 100 % and from 0 to 97 %, respectively [32, 33]. Among fMRI tasks, verb generation demonstrated a higher concordance when compared to intraoperative language mapping [34]. One explanation is that
The advent of functional imaging has advanced our understanding of brain function organization and has changed the clinical management of brain tumors in critical areas. The innovative imaging tool allows the surgeon to replace the a priori conception of brain function organization with a more individual and pathology-based approach. In brain tumor patients, neuroanatomy and functional localization are altered by the space-occupying tumors that deform imaging landmarks (central sulcus, triangular gyrus, etc.) and by the plastic reorganization of cortical functional maps. Moreover, the tumor can infiltrate the gyri nearby an eloquent area or can deeply infiltrate these areas. Consequently, the surgeon has to plan the surgical strategy based on the necessity to remove the tumor from the eloquent tissue without safe margins. Hence, the most relevant concern in every mapping technique is the reliability of the spatial positioning of a suspected eloquent
62
fMRI maps the entire cortical network involved in a specific task and usually is not able to differentiate between essential and substitutable epicenters. Furthermore, since a BOLD signal is generated by increased blood flow, infiltration by a vascularized tumor can completely alter the local microvascular pattern and potentially hamper the reliability of the BOLD signal [6, 35–37]. In our practice, we perform fMRI preoperatively to understand the activation pattern, the location of the precentral or postcentral gyrus, and the approximate distance to the tumor. If the distance from the activation spot is greater than one gyrus or the subcortical infiltration is minimal, we may even choose not to perform an awake intraoperative CSES. For tumors in language areas, we calculate the lateralization index that, together with neuropsychological testing, can indicate the dominant hemisphere. However, it has to be kept in mind that, as shown by some authors, in patients with tumor in presumed non-eloquent hemisphere, fMRI can underestimate the presence of language areas compared to direct stimulation [38]. The recent introduction of DTI-ft has enhanced the ability to reconstruct major white matter pathways. DTI-ft requires the acquisition of images after application of diffusionsensitizing gradients along at least six different spatial orientations and computation of the diffusion tensor and reconstruction of maps of the mean diffusivity and of the white matter anisotropic properties, usually in terms of fractional anisotropy [39]. DTI-ft can provide good tridimensional representation of major tracts, such as corticospinal tract (CST) or arcuate fasciculus, with impressive anatomical accuracy that is comparable to postmortem dissections [40]. Because the images generated by DTI-ft are the result of complex mathematical modeling aimed at resolving the hypercomplex structure of white matter, several authors have addressed some practical questions [19, 41]. For instance, to what degree do the images correspond to the actual anatomy of the bundles? What is the relationship of the bundle(s) to the tumor (displaced, infiltrated, interrupted)? And, most importantly, what is the function of the bundle(s)? Must the bundle be spared or can it be sacrificed? DTI-ft is currently intended to virtually reconstruct white matter tracts, but it cannot evaluate the function related to a tract. Moreover, the actual rate of correspondence between a deterministic DTI-ft and intraoperative subcortical stimulation is no more than 80 % [41]. It is certain that DTIft will help locate the position of given tracts and thereby improve subcortical mapping (Fig. 1). In order to improve the reliability of DTI-ft, very recently, Bucci et al. used preoperative high angular resolution diffusion MRI (HARDI) data to delineate the corticospinal tract in patients with gliomas and compared these data with the reference of intraoperative cortical and subcortical direct mapping. They found an evident advantage in terms of accuracy and precision of this probabilistic approach over the classical deterministic model [42]. Nowadays, DTI-ft can be useful for teaching white matter anatomy from a surgical perspective in combination with
Neurosurg Rev (2015) 38:59–70
cadaveric dissection. It can also be used to speculate on the function of white matter tracts by comparing postoperative DTI-ft with the position of subcortical mapping sites.
Surgical approach and resection strategies There is still debate about the optimal size of the craniotomy; some authors prefer a “minimally invasive” approach and advocate a small craniotomy that exposes the least cortex. Although this latter approach could be understandable in non-EATs, some authors demonstrated that even in these tumors, it is possible and convenient to expose and remove more cortex and, thanks to the CSES, to achieve larger resection (so-called supra-total resection) with an increased impact on the natural history of the disease [43]. However, for EATs, direct mapping requires that the presumed eloquent cortex be accessible. In fact, it is important to expose as much cortex as needed to obtain responses following stimulation and to expose the specific motor or sensory cortex where the current has to be set; in this way, false negative mapping can be reduced [44–46]. Moreover, tumors that infiltrate large areas of subcortical brain tissue sometimes require a larger cortical exposure to provide sufficient space to perform subcortical stimulation. Actually, for small tumors near the central sensory-motor area, we also use a circumscribed craniotomy that encompasses the gyri that require stimulation. Once the dura is opened and reflected, the first step is to accurately examine the cortex and try to recognize all the gyral, sulcal, and vascular anatomy. In this early stage, navigation can help to accurately locate cortical and vascular anatomy. Although gliomas are infiltrating tumors, it is very useful to understand which sulcus or gyrus surrounds the tumor [81]. This strategy is intended to avoid intratumoral debulking that could sometimes leave tumor behind and to achieve, when possible, the resection of the whole infiltrated gyri. This concept is particularly relevant to LGGs that typically do not trespass the pia of the sulcus and hence this pia can be used as a topographical limit. In such cases, tumor resection starts with subpial dissection (see below). Conversely, HGGs can sometimes extend past sulci and obviously distort the anatomy. Moreover, HGGs can present with central necrosis and thick peripheral walls. If these conditions are present and the tumor is large, it is sometimes preferable to first perform an intratumoral debulking then continue the resection until apparently healthy brain is encountered. However, HGGs do not have distinct margins between the tumor mass and the surrounding brain, so gross total resection is challenging. Numerous surgical technologies have been developed to detect residual tumor tissue. Recently, imageguided resection (intraoperative MRI, ultrasound, and CT) has been surpassed by fluorescence-guided surgery with 5-
Neurosurg Rev (2015) 38:59–70
63
aminolevulinic acid (5-ALA). 5-ALA is a precursor in the hemoglobin synthesis pathway that leads to the accumulation of fluorescent porphyrins in HGG cells. Oral administration of 5-ALA several hours before surgery and a modified neurosurgical microscope are required. In our experience, decisionmaking based on 5-ALA fluorescence positivity had a sensitivity of 91.1 % and a specificity of 89.4 % [47]. The impact of fluorescence in surgery for gliomas was clearly shown by a large, multicenter, phase III randomized controlled trial in which 5-ALA-guided resection resulted in a significantly higher resection rate that translated into a longer progression-free interval, as well as a longer median survival [37]. The major venous system and terminal arteries, especially in delicate areas such as perirolandic/perisylvian region and parasagittal region, must always be respected because violation of these vessels is a primary cause of postoperative infarction and unsatisfying results (Figs. 2 and 3). In fact, if large caliber arteries are safely handled, both with or without microsurgical technique, venous trunks and terminal arteries are more prone to rupture or thrombose. Recently, a study reported that postoperative diffusion-weighted MRI indicated that ischemic lesions could account for 30 % of postoperative deficits after surgery for primary gliomas and up to 80 % of deficits in recurrent gliomas [48]. For tumors in the parasagittal area (infiltrating SMA or prerolandic and postrolandic area), the dura must be opened with great care because large bridging veins very often enter the sagittal sinus few millimeters away from the midline, and hence, there is always a risk of cutting them inadvertently. In this situation, it is preferable to preserve the dura along the vein and continue the opening anteriorly and posteriorly. Subpial dissection Subpial dissection is a very familiar technique, performed in medio-temporal lobe surgery (epilepsy, tumors) [49] where the dissection and resection of the uncus to the parahippocampal gyrus can be safely performed by respecting the pial layer that separates these structures from the neurovascular contents of the ambient and crural cistern. The first steps are the following: coagulate the pia just in front of the sulcus delimiting the tumor, open it with scissors, and detach the brain from the internal surface of the pia with a dissector or bipolar tips (Figs. 1 and 4). Following this dissection plan, it is possible to rapidly reach the bottom of the sulcus from its cortical layer (Figs. 2 and 4). This maneuver can be performed either microscopically or macroscopically and avoids any violation of the intrasulcal vascularization. This strategy is particularly advantageous, because intrasulcal dissection could damage vascularization of the closest eloquent gyrus and should therefore be avoided when manipulating infiltrated brain just in front of an eloquent gyrus. By
Fig. 2 a, b Preoperative MRI in T1 postgadolinium and T2. The tumor completely invades the left frontal operculum and is in close contact with the sylvian vasculature. Although the margins are not well-defined, a sulcus with a draining vein delimitates posteriorly the tumor (green arrow). c Intraoperative picture. Language areas were found only on the gyrus posteriorly (namely the foot of the frontal ascending gyrus), while the tumor area was unresponsive. In the depth of the cavity, all the sylvian arteries were dissected and separated from the tumor. Resection stopped in the postero-medial aspect of the tumor as stimulation disrupted articulatory disturbances (green arrow). d, e Postoperative gadolinium enhanced and T2 MRI demonstrating gross total resection
following these pial layers, it is possible to quickly delimitate the relevant cortical surface of the tumor and, when in noneloquent areas, to continue dissecting subcortically until noninfiltrated brain is apparent. Conversely, in those cases where the tumor infiltrates subcortical functional pathways (i.e., corticospinal tract), our strategy is to stop the subpial dissection at the bottom of the sulcus where subcortical connections start (u-fibers). Due to the absence of any anatomical reference in the white matter, resection is alternated with stimulation to
64
Neurosurg Rev (2015) 38:59–70
Fig 3 A right-handed, 37-yearold male presenting with seizures. a, b T2-FLAIR MRI demonstrated a large frontotemporo-insular tumor. c Intraoperative picture. Cortical mapping demonstrated the absence of language sites in the anterior frontal and temporal opercula and motor area of the mouth posteriorly (numbered tags 2, 3, and 4). The tumor was resected through a trans-opercular approach. Green arrow represents the sylvian fissure. d–f A near total resection is displayed on postoperative T2 MRI
define functional limits which will represent the boundaries of the surgical cavity (Figs. 3, 4, and 5). This part of the resection is complex because there is no uniform surface, such as cortex, Fig. 4 The artist’s drawing depicts the fundamental stages of the subpial dissection. See text for explanation
to stimulate and to delineate resection margins. Very recently, Bello et al. demonstrated how the choice of the subcortical stimulation protocol (monopolar or bipolar) adapted to the
Neurosurg Rev (2015) 38:59–70
patient’s clinical history, and tumor characteristics can improve resection of gliomas and impact directly on the extent of resection and patient functional integrity [50]. To define functional limits, we both request patients to perform specific tasks or to spontaneously speak or move to anticipate the approaching to critical fascicles. This “active” mapping provides continuous feedback on the effect of the resection and additional insights into brain functioning. Regarding motor function, many methods (evoked potential and direct mapping in an asleep patient) can readily detect primary motor cortex, but spontaneous movement in an awake subject provides more information about coordination and other aspects of movement than does simple muscle contraction. Many authors have long postulated the need to maintain a safe distance, at least 1 cm, from a functional site [45, 46, 51]. More recently, this recommendation has changed because accumulated experiences have clearly demonstrated that continuous cortical and subcortical stimulations enable the surgeon to identify and preserve eloquent cortex and white matter bundles (Fig. 5). Because replacing “safe margins” with
65
functional boundaries increases the EOR, this recent strategy is believed to affect the natural history of the tumor. This more aggressive strategy is related to a higher frequency of transient postoperative neurological deficits, but it has also led to very satisfying long-term neurological outcomes [52]. When the tumor is deep-seated and completely hidden in the white matter, surgeons must find the route to the tumor that is both most direct and that sacrifices the least normal brain. Classically, the most frequently used routes were trans-sulcal, trans-fissural, or trans-gyral depending on the individual anatomy [53–56]. In those brain areas presumptively defined as “non-eloquent”, navigation devices can precisely localize a purely subcortical lesion, and the most direct path is then exploited. Conversely, when the overlying cortical area and white matter encapsulating the mass are eloquent, simple localization systems or microsurgical techniques cannot guarantee a safe route and hence low morbidity. Stimulation is applied over the gyrus identified as the best entry point to reach the tumor. If mapping detects no responsive cortical sites in that gyrus, the corridor is started exactly there. If some eloquent area is encountered along the predetermined route, the surgeon must modify the route based on the mapping results. If a trans-sulcal approach is chosen, stimulation of deep cortex is still performed to exclude the presence of functional cortex. Similarly, if a trans-gyral approach is chosen, after cortectomy, subcortical stimulation is again performed. In doing so, the dissection through the brain is completed, leaving apparently functional white matter untouched.
CSES: why, when, and what to expect Why?
Fig. 5 a Preoperative T2 MRI of a LGG presenting sharp margins. b Intraoperative pictures. Numbered tags refer to responsive sites during stimulation (1, 2, and 5, motor and sensory of left hand; 3 and 4, sensory area of the mouth; 6, dysarthria). Blue arrows represent the arachnoid plan delimiting the sulci and the veins passing over the sulci. This plane together with pia contouring the sulcus is used for the subpial dissection. c At the end of the resection, the pia of the sulcus (white arrow) posteriorly is untouched as well as the veins (green arrow). White tags refer to subcortical pathways that represent the deep boundaries of the resection (movements of the left hemiface and left conjugated eyes deviation). d Postoperative MR showing complete resection of the tumor
Although the direct electrical stimulation of the brain has been used since the first decades of the past century [57–61], the refinement of technical equipment and availability of ultrashort acting anesthetics and new analgesics are compelling reasons to revive this technique. The goals of CSES are to detect the areas of the brain that are required for a given function and to continuously check the integrity of both the cortical and the subcortical circuitry. Different from functional imaging, CSES enables the surgeon to map, point-by-point, the entire cortico-subcortical area of interest with excellent reproducibility, reliability, and safety. In EATs, it was demonstrated that postoperative sequelae were more frequent (19 vs 4 %) in case series without CSES than in series that performed intraoperative mapping [62]. A study comparing LGGs that had been resected with or without CSES in the same institution demonstrated that with CSES, 62 % of gliomas selected for surgery were located within an eloquent area. In contrast, without CSES, only 35 % were located within an eloquent
66
Neurosurg Rev (2015) 38:59–70
area. The proportion of resections that were subtotal and total were 37 and 6 % (with no signal abnormality), respectively, without CSES, whereas these proportions were greater, 50.8 and 25.4 %, respectively, with CSES (P
REVIEW
Resection of supratentorial gliomas: the need to merge microsurgical technical cornerstones with modern functional mapping concepts. An overview Giannantonio Spena & Pier Paolo Panciani & Marco Maria Fontanella
Received: 13 January 2014 / Revised: 22 April 2014 / Accepted: 22 June 2014 / Published online: 21 October 2014 # Springer-Verlag Berlin Heidelberg 2014
Abstract Although surgery is not curative for the majority of intracranial gliomas, radical resection has been demonstrated to influence survival and delay tumor progression. Because gliomas are very frequently located in eloquent or more generally critical areas, surgeons must always balance the maximizing resection with the need to preserve neurological function. In this overview, we tried to summarize the recent literature and our personal experience about (1) the benefits and limits of using preoperative anatomical and functional neuroimaging (anatomical MRI, DTI fiber tracking, and functional MRI), (2) the issues to consider in planning the surgical strategy, (3) the need to thoroughly understand microsurgical techniques that enable a maximal resection (subpial dissection, vascular manipulation, etc.), (4) the importance of individualizing surgical strategy especially in patients with gliomas in eloquent areas (the role of neuropsychological evaluation in redefining eloquent and non-eloquent areas), and (5) how to use intraoperative mapping techniques and understand why and when to use them. Through this paper, the reader should become more familiar with a comprehensive panel of techniques and methodologies but more importantly become aware that these recent technical advances G. Spena : P. P. Panciani : M. M. Fontanella Neurosurgery Department, Spedali Civili and University of Brescia, Brescia, Italy P. P. Panciani e-mail: [email protected] M. M. Fontanella e-mail: [email protected] G. Spena (*) Neurosurgery Department, Spedali Civili of Brescia, Piazzale Spedali Civili 1, 25120 Brescia, Italy e-mail: [email protected] G. Spena e-mail: [email protected]
facilitate a conceptual change from classical surgical paradigms toward a more patient-specific approach. Keywords Brain neoplasms . Diagnostic imaging . Diffusion tensor imaging . Electric stimulation . Quality of life . Magnetic resonance imaging . Gliomas
Introduction Surgical treatment of supratentorial gliomas has changed markedly in the last decades. It has been demonstrated that surgery influences survival of high-grade gliomas (HGGs) especially in association with adjuvant treatment (chemotherapy and radiotherapy) [1–3]. Very recently, a larger resection was associated with longer survival or delayed relapse in low-grade glioma (LGGs) patients [4–8]. Globally, a policy of more extensive resection has evolved, and many authors have advocated widening the indications for surgery. Consequently, neurosurgeons are often asked to perform the widest possible resection. Moreover, the widespread use of early postoperative MRI follow-up has led to a more objective evaluation of the extent of resection (EOR) that no longer relies on only the surgeon’s perception of the EOR [9, 10]. The neurosurgeons can be aware of the actual EOR, evaluate those areas with surgically induced damage (i.e., ischemia), better correlate the intraoperative anatomical and functional boundaries of the resection with a detailed postoperative image, and gain insights about the anatomo-functional organization of a single patient. The goal of wider resection is easily achievable in some locations; however, in other sites, such as eloquent area tumors (EATs) or more generally critical areas, a wide resection is more difficult. Still, an oncologically desirable resection always should be balanced with the need to preserve neurological function, in both in HGGs and LGGs, although for different reasons. In HGGs, postoperative impairment and the potentially detrimental effect of adjuvant therapy
60
(especially radiotherapy) were demonstrated to be crucially related to shorter survival [11–13]. On the other hand, patients with LGGs are often young, and their mean survival is far longer than in patients with HGGs; therefore, they are usually expected to recover their normal function after surgery. For EATs, it has been shown that cortical and subcortical electrical stimulation (CSES) can guarantee a maximal resection while minimizing neurological sequelae [14–19, 82]. The planning of this approach requires the correct interpretation of preoperative functional imaging such as functional MRI (fMRI) and diffusion tensor imaging fiber tracking (DTI-ft) that can add useful information about individual anatomical and functional characteristics. In addition, the neurological and neuropsychological status of the patient as well as his/ her expectations and needs must be clarified preoperatively to tailor the surgical resection and thereby achieve the best quality of life for that specific patient. Regardless of whether the tumor is located in eloquent areas or not, it is of highest importance that the manipulation of vessels, arachnoid, and pial planes is conducted with mastery. The concept of “debulking” the tumor has progressively been abandoned in favor of a more tailored resection based on accurate knowledge of the brain-tumor interface anatomy. Consequently, an accurate analysis of the preoperative anatomical imaging is warranted. The aim of this paper is to elucidate those aspects of the surgical planning and operative technique that are required to achieve a large but safe resection and to show how to maximally exploit functional mapping concepts in order to treat more complex EATs.
Surgical planning: what can we get from anatomical and functional MRI? Currently, neurosurgeons have the many options to preoperatively evaluate brain anatomy and function, tumor morphology, and brain/tumor interface that enable them to thoroughly plan most of the surgery. The possibility to preoperatively define the tumor growth pattern (diffuse or bulky, the presence of a contrast enhancing rim, and the degree of the distortion of surrounding brain) and to reliably preview the histology (slow or rapidly evolving tumor) is important because these characteristics influence the type of interface with the healthy brain [20]. In our recent paper [18], we were able to demonstrate that EATs that extend toward subcortical tracts were less amenable to gross total resection and this situation was not forcedly related to the tumor volume. Similarly, other investigators demonstrated that the presence of infiltrated or displaced fascicles on the preoperative DTI-ft was predictive of a lower probability of total resection, even in tumors with a smaller preoperative volume, in which an extensive resection was expected [21].
Neurosurg Rev (2015) 38:59–70
In addition, morpho-structural factors related to the biology of the tumor are determinants of the EOR. The tumor-brain interface seems to influence the resection by creating better dissection planes, either for non-enhancing than for enhancing tumors. Although gliomas are infiltrating tumors, the growth characteristics of tumors that infiltrate or dislocate white matter tracts can influence the ease of resection. Similarly, Talos et al. [22] found that a large tumor volume associated with diffuse tumor margins and oligodendroglioma or oligoastrocytoma histopathologic type was predictive of incomplete resection. The strategy of surgeons is to perform the largest resection possible; therefore, the tumor’s tridimensional shape must be considered. Modern MRI can clearly delineate the sulcal, gyral, and vascular anatomy and help create a map of the tumor and surrounding brain. Several authors demonstrated how specific MR sequences, such as the T2-reversed and STIR, can add very reliable details of cortical and sulcal anatomy [23, 24]. The arterial and venous system of the tumor and the principal arterial and venous branches passing by the tumor are exactly displayed on T2 or T1 postgadolinium sequences (Fig. 1). Neuronavigation devices are very useful instruments to precisely locate brain lesions and help define the craniotomy size and position. Moreover, these devices can locate some structures at the very beginning of the procedure (vessels and sulci). Unfortunately, this tool cannot function once CSF outflows and the tumor is debulked, because the framework is definitively altered. Hence, most of the anatomical details are obtained from preoperative MRI or, when available, from intraoperative MRI. If the tumor is sited near a major cistern, such as the sylvian fissure, ambient cistern, or crural cisterns (tumors of the mesiotemporal region), a T2 sequence can help to understand whether a pial plane is preserved and whether the tumor has a close relationship with the vessels or nerves. When dealing with a tumor that is located purely in the white matter, sulcal anatomy has to be clearly defined by MRI to determine the most direct pathway to the tumor. This route can be either transgyral or transsulcal because no strong evidence has proven the safest route (read below). Neuronavigation can be used to choose the most direct path to the tumor. Esposito et al. [25] described a useful and less expensive method to locate cortical structures that they named MRI-based corticotopography. They exploited volumetric MRI reconstruction of the cortical anatomy of the subject that was uploaded to a laptop and matched with an intraoperative image. Despite the method utilized to locate subcortical tumors, if the mass is sited under eloquent areas, it is advisable to choose the transcortical route with direct mapping [26] (see below).
Neurosurg Rev (2015) 38:59–70
61
Fig. 1 a–d Preoperative MR showing a postrolandic tumor invading the subcortical white matter with poorly defined margin (blue arrow). Turquoise arrows point out two rolandic veins passing over the tumor. e, f
DTI-ft reconstructs a CST in a very close position to the tumor. g, h, i, l, m, n Postoperative MR shows gross total resection of the tumor. Note the veins that were spared during resection (green arrows)
fMRI and DTI-ft
area. The spots of activation during fMRI scanning are greatly affected by the statistical threshold chosen for data evaluation. Even with the use of one or more fixed statistical thresholds, blood oxygenation level-dependent (BOLD) signal intensities and cluster sizes differ significantly from one patient to another and between different paradigms (e.g., foot movement, hand movement), even when examinations are standardized. Studies that compared presurgical fMRI findings with a reference procedure, such as CSES performed in patients with lesions near the central sulcus [27–29], reported high concordance; percent agreement ranged from 83 to 92 % [19, 30, 31, 83, 85]. However, for evaluation of language areas, the utility of fMRI is diminished as demonstrated by our results (42.8 %) as well as by previous reports of variable sensitivities and specificities ranging from 59 to 100 % and from 0 to 97 %, respectively [32, 33]. Among fMRI tasks, verb generation demonstrated a higher concordance when compared to intraoperative language mapping [34]. One explanation is that
The advent of functional imaging has advanced our understanding of brain function organization and has changed the clinical management of brain tumors in critical areas. The innovative imaging tool allows the surgeon to replace the a priori conception of brain function organization with a more individual and pathology-based approach. In brain tumor patients, neuroanatomy and functional localization are altered by the space-occupying tumors that deform imaging landmarks (central sulcus, triangular gyrus, etc.) and by the plastic reorganization of cortical functional maps. Moreover, the tumor can infiltrate the gyri nearby an eloquent area or can deeply infiltrate these areas. Consequently, the surgeon has to plan the surgical strategy based on the necessity to remove the tumor from the eloquent tissue without safe margins. Hence, the most relevant concern in every mapping technique is the reliability of the spatial positioning of a suspected eloquent
62
fMRI maps the entire cortical network involved in a specific task and usually is not able to differentiate between essential and substitutable epicenters. Furthermore, since a BOLD signal is generated by increased blood flow, infiltration by a vascularized tumor can completely alter the local microvascular pattern and potentially hamper the reliability of the BOLD signal [6, 35–37]. In our practice, we perform fMRI preoperatively to understand the activation pattern, the location of the precentral or postcentral gyrus, and the approximate distance to the tumor. If the distance from the activation spot is greater than one gyrus or the subcortical infiltration is minimal, we may even choose not to perform an awake intraoperative CSES. For tumors in language areas, we calculate the lateralization index that, together with neuropsychological testing, can indicate the dominant hemisphere. However, it has to be kept in mind that, as shown by some authors, in patients with tumor in presumed non-eloquent hemisphere, fMRI can underestimate the presence of language areas compared to direct stimulation [38]. The recent introduction of DTI-ft has enhanced the ability to reconstruct major white matter pathways. DTI-ft requires the acquisition of images after application of diffusionsensitizing gradients along at least six different spatial orientations and computation of the diffusion tensor and reconstruction of maps of the mean diffusivity and of the white matter anisotropic properties, usually in terms of fractional anisotropy [39]. DTI-ft can provide good tridimensional representation of major tracts, such as corticospinal tract (CST) or arcuate fasciculus, with impressive anatomical accuracy that is comparable to postmortem dissections [40]. Because the images generated by DTI-ft are the result of complex mathematical modeling aimed at resolving the hypercomplex structure of white matter, several authors have addressed some practical questions [19, 41]. For instance, to what degree do the images correspond to the actual anatomy of the bundles? What is the relationship of the bundle(s) to the tumor (displaced, infiltrated, interrupted)? And, most importantly, what is the function of the bundle(s)? Must the bundle be spared or can it be sacrificed? DTI-ft is currently intended to virtually reconstruct white matter tracts, but it cannot evaluate the function related to a tract. Moreover, the actual rate of correspondence between a deterministic DTI-ft and intraoperative subcortical stimulation is no more than 80 % [41]. It is certain that DTIft will help locate the position of given tracts and thereby improve subcortical mapping (Fig. 1). In order to improve the reliability of DTI-ft, very recently, Bucci et al. used preoperative high angular resolution diffusion MRI (HARDI) data to delineate the corticospinal tract in patients with gliomas and compared these data with the reference of intraoperative cortical and subcortical direct mapping. They found an evident advantage in terms of accuracy and precision of this probabilistic approach over the classical deterministic model [42]. Nowadays, DTI-ft can be useful for teaching white matter anatomy from a surgical perspective in combination with
Neurosurg Rev (2015) 38:59–70
cadaveric dissection. It can also be used to speculate on the function of white matter tracts by comparing postoperative DTI-ft with the position of subcortical mapping sites.
Surgical approach and resection strategies There is still debate about the optimal size of the craniotomy; some authors prefer a “minimally invasive” approach and advocate a small craniotomy that exposes the least cortex. Although this latter approach could be understandable in non-EATs, some authors demonstrated that even in these tumors, it is possible and convenient to expose and remove more cortex and, thanks to the CSES, to achieve larger resection (so-called supra-total resection) with an increased impact on the natural history of the disease [43]. However, for EATs, direct mapping requires that the presumed eloquent cortex be accessible. In fact, it is important to expose as much cortex as needed to obtain responses following stimulation and to expose the specific motor or sensory cortex where the current has to be set; in this way, false negative mapping can be reduced [44–46]. Moreover, tumors that infiltrate large areas of subcortical brain tissue sometimes require a larger cortical exposure to provide sufficient space to perform subcortical stimulation. Actually, for small tumors near the central sensory-motor area, we also use a circumscribed craniotomy that encompasses the gyri that require stimulation. Once the dura is opened and reflected, the first step is to accurately examine the cortex and try to recognize all the gyral, sulcal, and vascular anatomy. In this early stage, navigation can help to accurately locate cortical and vascular anatomy. Although gliomas are infiltrating tumors, it is very useful to understand which sulcus or gyrus surrounds the tumor [81]. This strategy is intended to avoid intratumoral debulking that could sometimes leave tumor behind and to achieve, when possible, the resection of the whole infiltrated gyri. This concept is particularly relevant to LGGs that typically do not trespass the pia of the sulcus and hence this pia can be used as a topographical limit. In such cases, tumor resection starts with subpial dissection (see below). Conversely, HGGs can sometimes extend past sulci and obviously distort the anatomy. Moreover, HGGs can present with central necrosis and thick peripheral walls. If these conditions are present and the tumor is large, it is sometimes preferable to first perform an intratumoral debulking then continue the resection until apparently healthy brain is encountered. However, HGGs do not have distinct margins between the tumor mass and the surrounding brain, so gross total resection is challenging. Numerous surgical technologies have been developed to detect residual tumor tissue. Recently, imageguided resection (intraoperative MRI, ultrasound, and CT) has been surpassed by fluorescence-guided surgery with 5-
Neurosurg Rev (2015) 38:59–70
63
aminolevulinic acid (5-ALA). 5-ALA is a precursor in the hemoglobin synthesis pathway that leads to the accumulation of fluorescent porphyrins in HGG cells. Oral administration of 5-ALA several hours before surgery and a modified neurosurgical microscope are required. In our experience, decisionmaking based on 5-ALA fluorescence positivity had a sensitivity of 91.1 % and a specificity of 89.4 % [47]. The impact of fluorescence in surgery for gliomas was clearly shown by a large, multicenter, phase III randomized controlled trial in which 5-ALA-guided resection resulted in a significantly higher resection rate that translated into a longer progression-free interval, as well as a longer median survival [37]. The major venous system and terminal arteries, especially in delicate areas such as perirolandic/perisylvian region and parasagittal region, must always be respected because violation of these vessels is a primary cause of postoperative infarction and unsatisfying results (Figs. 2 and 3). In fact, if large caliber arteries are safely handled, both with or without microsurgical technique, venous trunks and terminal arteries are more prone to rupture or thrombose. Recently, a study reported that postoperative diffusion-weighted MRI indicated that ischemic lesions could account for 30 % of postoperative deficits after surgery for primary gliomas and up to 80 % of deficits in recurrent gliomas [48]. For tumors in the parasagittal area (infiltrating SMA or prerolandic and postrolandic area), the dura must be opened with great care because large bridging veins very often enter the sagittal sinus few millimeters away from the midline, and hence, there is always a risk of cutting them inadvertently. In this situation, it is preferable to preserve the dura along the vein and continue the opening anteriorly and posteriorly. Subpial dissection Subpial dissection is a very familiar technique, performed in medio-temporal lobe surgery (epilepsy, tumors) [49] where the dissection and resection of the uncus to the parahippocampal gyrus can be safely performed by respecting the pial layer that separates these structures from the neurovascular contents of the ambient and crural cistern. The first steps are the following: coagulate the pia just in front of the sulcus delimiting the tumor, open it with scissors, and detach the brain from the internal surface of the pia with a dissector or bipolar tips (Figs. 1 and 4). Following this dissection plan, it is possible to rapidly reach the bottom of the sulcus from its cortical layer (Figs. 2 and 4). This maneuver can be performed either microscopically or macroscopically and avoids any violation of the intrasulcal vascularization. This strategy is particularly advantageous, because intrasulcal dissection could damage vascularization of the closest eloquent gyrus and should therefore be avoided when manipulating infiltrated brain just in front of an eloquent gyrus. By
Fig. 2 a, b Preoperative MRI in T1 postgadolinium and T2. The tumor completely invades the left frontal operculum and is in close contact with the sylvian vasculature. Although the margins are not well-defined, a sulcus with a draining vein delimitates posteriorly the tumor (green arrow). c Intraoperative picture. Language areas were found only on the gyrus posteriorly (namely the foot of the frontal ascending gyrus), while the tumor area was unresponsive. In the depth of the cavity, all the sylvian arteries were dissected and separated from the tumor. Resection stopped in the postero-medial aspect of the tumor as stimulation disrupted articulatory disturbances (green arrow). d, e Postoperative gadolinium enhanced and T2 MRI demonstrating gross total resection
following these pial layers, it is possible to quickly delimitate the relevant cortical surface of the tumor and, when in noneloquent areas, to continue dissecting subcortically until noninfiltrated brain is apparent. Conversely, in those cases where the tumor infiltrates subcortical functional pathways (i.e., corticospinal tract), our strategy is to stop the subpial dissection at the bottom of the sulcus where subcortical connections start (u-fibers). Due to the absence of any anatomical reference in the white matter, resection is alternated with stimulation to
64
Neurosurg Rev (2015) 38:59–70
Fig 3 A right-handed, 37-yearold male presenting with seizures. a, b T2-FLAIR MRI demonstrated a large frontotemporo-insular tumor. c Intraoperative picture. Cortical mapping demonstrated the absence of language sites in the anterior frontal and temporal opercula and motor area of the mouth posteriorly (numbered tags 2, 3, and 4). The tumor was resected through a trans-opercular approach. Green arrow represents the sylvian fissure. d–f A near total resection is displayed on postoperative T2 MRI
define functional limits which will represent the boundaries of the surgical cavity (Figs. 3, 4, and 5). This part of the resection is complex because there is no uniform surface, such as cortex, Fig. 4 The artist’s drawing depicts the fundamental stages of the subpial dissection. See text for explanation
to stimulate and to delineate resection margins. Very recently, Bello et al. demonstrated how the choice of the subcortical stimulation protocol (monopolar or bipolar) adapted to the
Neurosurg Rev (2015) 38:59–70
patient’s clinical history, and tumor characteristics can improve resection of gliomas and impact directly on the extent of resection and patient functional integrity [50]. To define functional limits, we both request patients to perform specific tasks or to spontaneously speak or move to anticipate the approaching to critical fascicles. This “active” mapping provides continuous feedback on the effect of the resection and additional insights into brain functioning. Regarding motor function, many methods (evoked potential and direct mapping in an asleep patient) can readily detect primary motor cortex, but spontaneous movement in an awake subject provides more information about coordination and other aspects of movement than does simple muscle contraction. Many authors have long postulated the need to maintain a safe distance, at least 1 cm, from a functional site [45, 46, 51]. More recently, this recommendation has changed because accumulated experiences have clearly demonstrated that continuous cortical and subcortical stimulations enable the surgeon to identify and preserve eloquent cortex and white matter bundles (Fig. 5). Because replacing “safe margins” with
65
functional boundaries increases the EOR, this recent strategy is believed to affect the natural history of the tumor. This more aggressive strategy is related to a higher frequency of transient postoperative neurological deficits, but it has also led to very satisfying long-term neurological outcomes [52]. When the tumor is deep-seated and completely hidden in the white matter, surgeons must find the route to the tumor that is both most direct and that sacrifices the least normal brain. Classically, the most frequently used routes were trans-sulcal, trans-fissural, or trans-gyral depending on the individual anatomy [53–56]. In those brain areas presumptively defined as “non-eloquent”, navigation devices can precisely localize a purely subcortical lesion, and the most direct path is then exploited. Conversely, when the overlying cortical area and white matter encapsulating the mass are eloquent, simple localization systems or microsurgical techniques cannot guarantee a safe route and hence low morbidity. Stimulation is applied over the gyrus identified as the best entry point to reach the tumor. If mapping detects no responsive cortical sites in that gyrus, the corridor is started exactly there. If some eloquent area is encountered along the predetermined route, the surgeon must modify the route based on the mapping results. If a trans-sulcal approach is chosen, stimulation of deep cortex is still performed to exclude the presence of functional cortex. Similarly, if a trans-gyral approach is chosen, after cortectomy, subcortical stimulation is again performed. In doing so, the dissection through the brain is completed, leaving apparently functional white matter untouched.
CSES: why, when, and what to expect Why?
Fig. 5 a Preoperative T2 MRI of a LGG presenting sharp margins. b Intraoperative pictures. Numbered tags refer to responsive sites during stimulation (1, 2, and 5, motor and sensory of left hand; 3 and 4, sensory area of the mouth; 6, dysarthria). Blue arrows represent the arachnoid plan delimiting the sulci and the veins passing over the sulci. This plane together with pia contouring the sulcus is used for the subpial dissection. c At the end of the resection, the pia of the sulcus (white arrow) posteriorly is untouched as well as the veins (green arrow). White tags refer to subcortical pathways that represent the deep boundaries of the resection (movements of the left hemiface and left conjugated eyes deviation). d Postoperative MR showing complete resection of the tumor
Although the direct electrical stimulation of the brain has been used since the first decades of the past century [57–61], the refinement of technical equipment and availability of ultrashort acting anesthetics and new analgesics are compelling reasons to revive this technique. The goals of CSES are to detect the areas of the brain that are required for a given function and to continuously check the integrity of both the cortical and the subcortical circuitry. Different from functional imaging, CSES enables the surgeon to map, point-by-point, the entire cortico-subcortical area of interest with excellent reproducibility, reliability, and safety. In EATs, it was demonstrated that postoperative sequelae were more frequent (19 vs 4 %) in case series without CSES than in series that performed intraoperative mapping [62]. A study comparing LGGs that had been resected with or without CSES in the same institution demonstrated that with CSES, 62 % of gliomas selected for surgery were located within an eloquent area. In contrast, without CSES, only 35 % were located within an eloquent
66
Neurosurg Rev (2015) 38:59–70
area. The proportion of resections that were subtotal and total were 37 and 6 % (with no signal abnormality), respectively, without CSES, whereas these proportions were greater, 50.8 and 25.4 %, respectively, with CSES (P
