gms | German Medical Science

Surgical goal for each kind of tumour is the complete or nearly complete resection to improve the outcome of the patient and to keep his quality of life [1], [2], [3]. Only an acceptable visualisation of tumour, remnants, surrounding normal tissue and oedema allows a more radical concept of tumour excision. Neuronavigation systems have been demonstrated to improve planning and performance of brain tumour surgery. But brain shift limits the accuracy in the millimetre range and increases the risk of morbidity [4], [5], [6]. Recent authors described intraoperative 2D-ultrasound advantages in resection control in comparison to surgery without intraoperative imaging [7], [8], [9]. Especially metastases and meningeomas can be visualised excellently with appropriately chosen US frequency due to their homogenous and higher echogenicity compared to the surrounding tissue [10]. In contrast to this fact the authors summarized problems in distinction between high grade gliomas and surrounding oedema comparing intraoperative iUS with CT/MRI images. Especially at the end of resection quality of images decreases due to oedema and small bleeding.

Gliomas demonstrated usually a wide variety in echoicity and the tumour border was not clearly delineated in intraoperative ultrasound [11], [12]. As a solution a combination of preoperative MRI with intraoperative 3D-ultrasound (3D-iUS) may enhance the convenience of neuronavigation by adding intraoperative information. 3D-ultrasound is an instrument for intraoperative imaging, orientation and brains shift control. 3D-ultrasound planes in orientation to the space and patient support the exact resection with different instruments in a “real time” hand-eye-coordination [12], [13].

Recently we showed our experiences with the new developed intraoperative 3D-iUS navigation system [14].

The use of this navigation tool offers update of a navigation dataset during tumour resection until the end of the operative procedure to detect tumour remnants.

Aim of the study was to investigate the suitability of an iterative concept of malignant glioma resection based on intraoperatively acquired 3D-ultrasound. First control parameters were aiming biopsy during resection in comparison to the postoperative MRI and clinical outcome.

Six patients were previously selected for our analysis of 3D-ultrasound resection control of high grade gliomas. Inclusion criteria were a brain tumour with MR signs typical for a malignant glioma, tumour localization that allows for complete resection in 4 cases and limited resection due to infiltration of eloquent brain areas (basal ganglia) in 2 cases, and the accordance of the patient supported by the local ethic commission. For navigation a 3D ultrasound workstation was used consisting of a standard personal computer containing a video grabber card in combination with an optical tracking system (NDI Polaris) and a standard ultrasound device (Siemens Omnia) with a 7.5 MHz probe. No other probe was shared to compare operative results of different neurosurgeons [15].

Preoperative 3D-MRI-Dicom data were acquired with an 1.5 Tesla magnet unit (Symphony, Siemens, Erlangen, Germany) and transferred to the navigation workstation (Localite Navigator). A total of 250 T1w 3-D-MR images with a thickness of 1 mm with contrast agent were obtained in each case (FOV: 25 cm, matrix: 256 x 256, repetition time: 11.4 msec) for data processing. The area of contrast enhancement was defined as tumour and determined as volume in the summery of each slice. Registration was performed with skin fiducials [14].

All patients underwent early postoperative 3D-MRI including contrast agent within 24 hours after surgery.

The data acquisition was carried out in T1w 2D spin echo technique within only few minutes. In accordance to the preoperative procedure the postoperative imaging was performed in T1w with and without contrast agent and subtraction technique to identify tumour remnants. The tumour extent of regarding complete and incomplete tumour removal was assessed by radiologist and neurosurgeon independently and recorded in the protocol [11].

The ultrasound probe was tracked with an active tracker and could be processed directly in different depth (6-12 cm).

3D-ultrasound datasets were acquired after craniotomy (transdural), at different subsequent times of the resection and at the end of the operation on an average of four times. The tracked microscope (NC 4, Carl Zeiss Oberkochen, Germany) was used as a pointer to define tumour remnants in different scans and to find fixed targets of tumour tissue [16], [17], [18].

Several ultrasound display methods were used. Especially the “real time panel” shows real time 3D-iUS in 3 orthogonal slices matched with preoperative MRI and other preoperative images (PET, SPECT, fMRI) in “real time” direction and movement of different tracked instruments (pointer, ultrasound probe, microscope; Figure 1 [Fig. 1]). Additionally the compare panel shows the scan of the ultrasound probe in the 3D ultrasound dataset in comparison to 2D-iUS, MRI and spatial components (Figure 2 [Fig. 2]).

A standard protocol was prepared to describe brain shift, results of the biopsy, postoperative tumour remnants in the MR image as well as the clinical outcome [19].

Unclear tumour remnants were investigated by biopsy. At least two biopsies were codified in the protocol. Definition of unclear tumour remnant was given by echogenicity signal in the real time 3D-iUS dataset, but not visible in the microscope and the preoperative MRI dataset as well.

The biopsies were separately characterised as numbers in the histopathological report without special information to the neuropathologist [20].

In the 6 patients with supratentorial tumours brain shifting ranged at most from 2-8 mm with a mean of 4 millimetres. The size of the tumour ranged from 2-6 cm in diameter. All patients suffered from a glioblastoma multiforme (WHO IV; Table 1 [Tab. 1]).

Tumour resection was complete according microscopically criteria, intraoperative 3D-ultrasound as well as postoperative MRI in four cases. In each of these cases case two biopsies were taken from the resection cavity at the end of the tumour removal. In all patients the biopsy of washy borders between tumour and brain tissue showed the transition of normal to pathological brain. In the remaining two patients with tumour infiltration in eloquent areas removal was planned incomplete. In both patients 3D-iUS and postoperative MRI showed tumour remnants with a small distinction of 10% in residual size and location. Histopathological characterisation of 3 biopsies in both cases showed tumour remnants or infiltration zone but no normal brain tissue. In both cases 3D-IUS allowed for a more extended resection of the tumour.

A good correlation between 3D-IUS and postoperative MRI was found in all cases. There was no surgical morbidity observed after surgery.

The navigation started with the registration of preoperative 3D MRI to the patient. Five self-adhesive fiducials were therefore preoperative attached to the patients head. The accuracy of the fiducial based error (FFE) was at most 2 mm ranging from 0.5-2 mm. After acquisition of the first transdural 3D-ultrasound dataset typical tumour structures as well as eloquent areas like speech or central motor were identified.

Anatomical landmarks (falx, ventricle), necessary for image mapping, were compared in preoperative MRI/ fMRI and in 3D ultrasound dataset and signed by targets. These different targets were detected using the pointer, the tracked ultrasound probe or the tracked microscope as a pointing device during the operation. The mean accuracy was 0.8 mm. Stepwise resection could be visualised by real time 3D-iUS dataset with an expenditure of time up to 7 minutes. After incomplete resection all known targets could be found at iterative 3D-ultrasound datasets. At the end of the operation a 3D-iUS dataset was acquired to detect possible tumour remnants and hyperechoic lesion. In the case of unscheduled incomplete resection tracked instruments guided the neurosurgeon to the target to complete the removal in an iterative process (Figure 3 [Fig. 3]). Last 3D-IUS was done to compare the result postoperative MRI that is still the gold standard. Radiological findings were described by neuroradiologist independently.

Different image modalities demonstrate a wide range of physical and biological visualisation for the same area or lesion [1], [3], [21]. In ultrasound images , the echoicity is related to mass density and ultrasound propagation in diverse tissue [8]. Le Roux [20] showed an overestimation of tumour volume by intraoperative ultrasound in glioblastoma. Unsgaard et al. [12] reported 80% right positive biopsies in the histopathology. In our study visualisation of glioblastoma, surrounding tissue as well as areas of infiltration was possible in the same manner. Intraoperative 3D ultrasound was superior to preoperative MRI regarding intraoperative identification of tumour remnants and getting selective biopsies [14]. All biopsies from areas suspicious for tumour remnants showed tumour cells or the infiltrating zone. In all patients up to three selective biopsies could be taken without any problem. Neither tumour 3D-IUS nor MRI were suitable to detect infiltrating tumour cells and complete tumour removal in a biological sense cannot be achieved.

The small number of patients in these series inhibits a statistical analysis, but 3D-IUS was shown to offer a safe way to get selective biopsies tumour. All biopsies were done by US experienced neurosurgeons. The precision of tumour removal using 3D-ultrasound navigation depends on the experience in the use of intraoperative ultrasound and the number of operations as well [22]. Postoperative MRI demonstrated in all patients the same findings regarding tumour remnants compared to intraoperative 3D ultrasound. The size of the tumour remnants was different between MRI and 3D-IUS in two cases <10% possibly to different tissue representation and tissue visualization in both methods. This fact supports the high value of the ultrasound navigation in comparison to the gold standard. Le Roux et al. [20] demonstrated a tumour overestimation in ultrasound images for recurrent and irradiated glioblastoma. The presence of gliosis may be the reason for increasing volume [10], [12], [13].

Previous authors demonstrated that 3D-IUS offers advantages in brain shift analysis in comparison to preoperative imaging in a online measurement [12], [23], [24]. The accuracy of 0.8 mm for the tumour 3D-ultrasund used in these series fulfills the criteria for precise navigation even close to risk structures [17], [25], [26], [27].

In accordance to the results of Unsgaard et al. [12] the present study demonstrates no loss of image quality up to the end of the operation, when the operative approach is prepared to get best ultrasound quality. Preparation includes a blood free operating field, elimination of cotton swaps and a head position for ideal vertical ultrasound scans without air bubbles.

A clear distinction between infiltration zone and tumour mass was possible in all but two cases. In these patients the size and localization of the glioblastoma in combination with a huge oedema permit only an incomplete resection. Additional it is known that infiltrating cells can be found outside and far from the tumour border [20], [28].

The introduction of 3D-ultrasound has increased the value of neuronavigation substantially, making it possible to update ultrasound scans several times during surgery and minimize the problem of brain shift. Tumour resection control was possible in all cases of high grade gliomas in the present series using an iterative system of targets and tracked instruments in connection with mapped intraoperative 3D-ultasound and preoperative MRI. An important advantage to B-Mode ultrasound is the real time resection with tracked instruments under target control. 3D-ultrasound images can use during the tumour resection process with high accuracy. Clinical results and postoperative images support this new resection concept of malignant gliomas and offer a link between preoperative image and intraoperative dynamic changes.

The author received advancement by BMBF and this study is now part of the BMBF project ICCAS in Leipzig. No benefits in any form have been received from a commercial party related directly or indirectly to the subject of the manuscript.