Introduction

Intraparenchymal brain lesion biopsies are indicated for the diagnosis of various neurological conditions, including brain tumors, infections, and autoimmune diseases [15, 30]. In the past decades, biopsy procedures were commonly performed under stereotactic CT scan localization using a Leksell frame [7]. Two new technical approaches are gradually implemented in daily neurosurgical practice: frameless robot-assisted stereotactic biopsy [27, 32] and image-guided biopsy [21, 29]. Both techniques require trajectory planning. Frameless robot-assisted stereotactic biopsy is based on the use of a robotic arm to guide the biopsy cannula to the brain target, following the trajectory planned on preoperative MRI which will be fused with intraoperative CT scan controls [9, 10]. Image-guided biopsies use a navigation system including a computer-based image processing module, a reference frame, and a pointer tracked by an optical or electromagnetic detector, allowing for real-time navigation of the biopsy needle to the brain target [21]. Both techniques have been shown to be safe and effective in sampling intracranial lesions [8, 18, 21, 27, 32, 33].

Compared to historical framed-based stereotactic biopsies, several studies have shown both of these techniques to be equivalent in terms of efficacy and safety [5, 25, 28]. Nevertheless, the comparative effectiveness of frameless robot-assisted stereotactic intraparenchymal brain lesion biopsies versus image-guided biopsies has not been well established in clinical practice. To date, only one study compared the accuracy of both techniques to reach a predefined target in a small subset of patients but did not report clinical outcomes [20]. It could be challenging to design a prospective study comparing two surgical techniques due to each surgeon’s team habits, various costly surgical tools required, and the learning curve necessary to be able to use the new method efficiently. However, it is possible to analyze two different daily practices from two centers with the same preoperative and postoperative protocols, by determining control variables for population comparison in terms of age, sex, and major comorbidity [12].

In the present study, we compared the safety, efficacy, and diagnostic yield of frameless robot-assisted stereotactic versus image-guided intraparenchymal brain lesion biopsies in two different neurosurgical centers.

Methods

Study population

All consecutive adult patients > 18 years who had undergone biopsy for a newly diagnosed brain lesion between September 2019 and September 2022, at two French tertiary neurosurgical centers, were included in the present retrospective cohort study. Exclusion criteria were (1) extra-parenchymal intracranial lesion, (2) open brain lesion biopsy, and (3) previous surgery or oncological treatment. The inclusion flowchart is presented in Fig. 1.

Fig. 1
figure 1

Patient inclusion flowchart

Full size image

Variables and data sources

The following data were systematically collected at the time of surgery using a dedicated table: age, sex, clinical symptoms, Karnofsky Performance Status (KPS) score, previous history of cancer, preoperative treatment (corticosteroid, antiplatelet, and/or anticoagulant therapies), tumor location, contrast enhancement, spontaneous hemorrhage within the lesion, cystic component, surgical duration (from head clamp positioning to end of closure), and type of neurosurgical procedure (frameless robot-assisted stereotactic biopsy or image-guided biopsy), number of biopsy trajectories, number of biopsy samples, occurrence of intraoperative bleeding from biopsy cannula, histomolecular diagnosis according to the World Health Organization (WHO) classification of central nervous system (CNS) tumors 2021, immediate postoperative KPS score, and postoperative clinical (new neurological deficit, infection, and death) and imaging (postoperative hemorrhage, postoperative ischemia, increased mass effect) complications. The size of the postoperative intracerebral hematoma was quantified on the postoperative CT scan (performed the day after the surgery) by measuring the major axis of the hematoma on the axial plane.

Surgical procedures

Figures 2 and 3 illustrate the frameless robot-assisted stereotactic intraparenchymal brain lesion biopsy and the image-guided biopsy procedures, respectively (in each case, the patient consented to the publication of their image). Indications for intraparenchymal brain lesion biopsy were similar among both centers and defined according to French guidelines from the Association des Neuro-oncologues d’Expression Française [16]: (1) lesion ineligible for surgical resection, (2) tumor-like lesion confirmed on preoperative MRI, and (3) KPS score ≥ 50. Both centers used the French guidelines from the Société Française d’Anesthésie-Réanimation concerning the intraoperative management of anticoagulant and antiplatelet therapy [11]. In both centers, the biopsy trajectory was planned to avoid sulci and vessels. Center A performed frameless robot-assisted stereotactic biopsies with a NeuroMate Renishaw® minimally invasive robot, using a methodology previously detailed [33]. The biopsy trajectory was planned before surgery using the iPlan® software (Brainlab®, Munich, Germany). Under general anesthesia, the head was fixed using a Talairach head clamp. Preoperative MRI was merged with intraoperative 3D imaging acquired with the O-Arm, using the Neuroinspire software of the NeuroMate robot. This was done to align the biopsy cannula with the planned trajectory and to ensure its exact position at the target during the procedure. The cranial opening was made with a drill bit of 2.5 mm. Center B performed image-guided biopsies using the VarioGuide neuronavigation system (Brainlab®, Munich, Germany). The preoperative MRI was imported into an optical neuronavigation system (Curve Navigation, Brainlab®, Munich, Germany) for 3D reconstruction. The biopsy trajectory was performed directly on the curve module. Under general anesthesia, the head was fixed using a Mayfield skull clamp, and the reference array was connected to the head skull clamp. A 3D volume rendering of the patient’s head was created through the infrared detector according to the interpolation of the reflective markers’ position. The VarioGuide arm supporting the biopsy needle (Disposable Biopsy Needle, Brainlab®) was manually aligned to the planned trajectory by locking the six joints successively, and the biopsy needle was inserted to the target under optical control. The cranial opening was made with a self-stopping craniotomy drill bit of 14 mm.

Fig. 2
figure 2

Illustration of the frameless robot-assisted stereotactic brain biopsy. A Overview of the patient in prone position, his head in the Talairach head clamp fixed to the base of the robot, with the cone beam CT allowing frameless registration. In the top left corner, details of the Talairach head clamp. B Description of the five degrees of freedom of the robotic arm (NeuroMate, Renishaw). C The drill hole is performed with a 2.5-mm diameter drill bit. D The 10-mm window side-cutting biopsy cannula is inserted, and its movements are controlled by the robot

Full size image
Fig. 3
figure 3

Illustration of the image-guided brain biopsy. A Overview of the patient in prone position, his head in the Mayfield skull clamp with the reference star, allowing to calibrate the optical neuronavigation. In the top left corner, details of the Mayfield 3-pin skull clamp. B Description of the six degrees of freedom of the neuronavigation arm (VarioGuide, BrainLab). C The drill hole is performed with a 14-mm diameter drill bit, and then, the joints of the neuronavigation arm are locked accordingly to the optical neuronavigation. D The 10-mm window side-cutting biopsy cannula is inserted, and its movements are controlled on the neuronavigation screen

Full size image

Statistical analyses

Univariate analyses were carried out using Fisher’s exact or chi-squared tests to compare categorical variables and using the Mann–Whitney rank-sum test or unpaired t-test for continuous variables, appropriately. Significance in univariate analysis was set at a p-value of < 0.05. Analyses were performed by using the JMP software (Version 16.2.0; SAS Institute Inc, Cary, North Carolina, USA).

Standard protocol approvals, registrations, and patient consents

The authors have full access to the data and information presented for publication. The authors declare that they have no conflicts of interest. This study received the required authorizations (IRB#1:2023/10) from the institutional review board (IRB00011687). According to French legislation, the requirement for informed consent was waived for this observational retrospective study.

This manuscript was prepared in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) checklist.

Results

Study population

During the inclusion period, 134 patients were treated in center A, and 121 patients were treated in center B. Three patients who underwent an image-guided biopsy in center A were excluded, and two patients aged < 18 years were excluded from center B. A total of 250 patients (60.8% of men, mean age 65.3 ± 13.5 years) were included in this study (Fig. 1): 131 patients in center A and 119 patients in center B.

Clinical, radiological, and histomolecular findings

Clinical, radiological, and histomolecular findings and group comparability (in terms of age, sex, and major comorbidity) are detailed in Table 1. The median preoperative KPS score was 80 (IQ, 70–90). The main presenting symptoms were neurological impairment (72.8%), followed by epileptic seizures (19.2%), signs of raised intracranial pressure (5.2%), and incidental discovery (2.8%). An anticoagulant and/or antiplatelet therapy, which was stopped 5 days preoperatively according to the guidelines of the French National Health Agency [6], was preoperatively administered to 24.4% of patients (antiplatelet therapy only in 16%, both in 8.4%). Preoperative corticosteroid treatment was administered without discontinuation until surgery (mean duration 6.7 ± 14.9 days, range 0–150; mean dose 28.9 ± 38.7 mg/day, range 0–200) to 42.8% of patients. Thirty-eight patients (15.2%) had a medical history of cancer. The main lesion location was supratentorial and lobar in 79.2% of cases, deep-seated (i.e., brainstem, pineal gland, and basal ganglia) in 19.2% of cases, and cerebellar in 1.6% of cases. The lesion presented contrast enhancement in 83.2% of cases, was cystic in 6.8% of cases, and was spontaneously hemorrhagic in 6.4% of cases. Clinical and radiological findings did not significantly differ between the two centers (p > 0.05), except for the anticoagulant and/or antiplatelet therapy administration that was more frequent in patients from center A (30.5%) than from center B (16%, p = 0.018).

Table 1 Clinical, radiological, histomolecular findings, and group comparability
Full size table

A conclusive histomolecular diagnosis was obtained in 96.0% of cases, and the diagnosis of glioblastoma IDH-wildtype CNS WHO grade 4 was the most common diagnosis (66.8%). The diagnostic yield did not significantly differ between centers (p = 0.877), even in the subgroup of patients with a deep-seated lesion (p = 0.116).

Intraoperative findings

Intraoperative findings and comparisons between techniques are detailed in Table 2. The mean duration of the surgical procedure (55.0 ± 21.3 min, range 23–150) was longer in center A (61.9 ± 25.3 min, range 23–150) than in center B (47.4 ± 11.8 min, range 25–81, p < 0.001). No procedure aborted. In all but two patients, one biopsy trajectory was performed (99.2%). In one patient from center A, two trajectories were performed: the first to sample the lesion and the second to drain a tumoral cyst. In one patient from center B, a second trajectory was performed because the first samples did not appear abnormal on visual inspection. The median number of biopsy samples (8, IQ 6–8) was higher in center B (8, IQ 8–8) than in center A (6, IQ 4–8, p < 0.001). Intraoperative bleeding from biopsy cannula occurred in 4.4% of cases: 3.8% in center A and 5% in center B.

Table 2 Intraoperative findings and comparison between techniques
Full size table

Postoperative outcomes

Postoperative outcomes and comparisons between techniques are detailed in Table 3. The median preoperative KPS score was 80 (IQ, 70–90), similar between center A (80, IQ 70–90) and center B (80, IQ 60–90, p = 0.21).

Table 3 Postoperative outcomes and comparison between techniques
Full size table

Early postoperative CT scan was performed at a mean of 1.5 ± 1.3 days post-operatively (range 0–11) in all but eight patients: eight patients of center B did not have early postoperative imaging due to good clinical condition and so were excluded from statistical analysis regarding postoperative hematoma. A postoperative hemorrhage > 10 mm in diameter was observed in 10% of cases and was observed in 15.3% of cases in the subgroup of patients with an anticoagulant and/or antiplatelet therapy administration preoperatively. In the subgroup of patients with an anticoagulant and/or antiplatelet therapy preoperatively, a postoperative hemorrhage > 10 mm in diameter was observed more frequently in center B (36.8%) than in center A (5.0%, p = 0.001). The mean duration of hospital stays (4.0 ± 4.9 days, range 0–34) was longer in center B (4.9 ± 5.5 days, range 0–34) than in center A (3.1 ± 4.1 days, range 1–34, p < 0.001).

Thirty-four patients (13.6%) presented with postoperative complications: 9.2% presented with a worsened neurological deficit that remained permanent in 4% of cases, 3.6% presented with signs of increased intracranial pressure requiring osmotherapy, 3.2% presented with postoperative epileptic seizures that were not present preoperatively, 2.0% presented with general complications (two thromboembolic event, two pulmonary infections, one urinary infection), 0.8% presented a postoperative intracerebral hematoma requiring surgical evacuation, and 0.8% presented a postoperative surgical site infection. Overall, 3.2% of patients died during the first postoperative month. The overall rate of postoperative complications did not differ between groups (p = 0.76) nor did the rates of any specific complication (Table 3).

Discussion

Key results

In this bicentric retrospective study which included 250 patients who underwent a brain biopsy for a brain lesion by comparing frameless robot-assisted stereotactic biopsies (center A, 131 patients) versus image-guided biopsies (center B, 119 patients), it was demonstrated that (1) both neurosurgical procedures were safe with low and similar complications rates, (2) both neurosurgical procedures were efficient with a similar diagnostic yield of 96%, (3) the duration of the procedure was shorter using image-guided intraparenchymal brain lesion biopsies than using frameless robot-assisted stereotactic biopsies, and (4) the rate of postoperative hemorrhage > 10 mm in diameter was higher using image-guided biopsies than using frameless robot-assisted stereotactic biopsies, in the subgroup of patients with an anticoagulant and/or antiplatelet therapy administration.

Interpretation

Image-guided and robot-assisted biopsies have gradually replaced frame-based stereotactic biopsies for intraparenchymal brain lesions. Image-guided biopsies use classical and user-friendly tools, i.e., optical neuronavigation, at the expense of a loss of accuracy [20]. This lower accuracy may be due to registration errors, distortions in MRI, or lack of rigidity of the skull clamp [24, 26]. By contrast, the more expensive (in comparison with optical neuronavigation) surgical robot used by center A provided better trajectory accuracy but required a longer learning curve [4]. The surgical accuracy of the rigid robotic arm using bone registration has been measured ranging from 0.59–0.7 mm, which is superior to that of frame-based biopsies [8, 9, 19]. In comparison, the surgical accuracy of a neuronavigation system such as the one used by center B has been measured ranging from 1.8–2.2 mm [10]. Robot-assisted stereotactic biopsy procedures frequently require intraoperative imaging acquisitions, allowing for intraoperative check of the accuracy of the actual biopsy trajectory [33]. This explains the longer duration of the surgical procedure we observed in the frameless robot-assisted stereotactic biopsy center [8, 33]. The surgical accuracy may impact the efficacy of the surgical procedure, but we reported similar diagnostic yields between the two techniques. The high diagnostic yield (96%) is consistent with previous series ranging from 98–99% [1, 21, 23] for frameless robot-assisted stereotactic intraparenchymal brain lesion biopsies and ranging from 96.6–100% [3, 17, 18, 27] for image-guided biopsies. Of particular interest, the diagnostic yield of the two procedures remained similar in the subgroup of patients harboring a deep-seated lesion, recalling a previous series of pediatric brainstem biopsies using a neuronavigation system [13]. This suggests that if the neurosurgical procedure is performed properly, whichever the technology of biopsy guidance, the diagnostic yield is excellent. Persisting causes of failure remained poor planning or a “pathological” cause, including the preoperative administration of corticosteroids in patients harboring a CNS lymphoma [2, 14].

We reported a 13% rate of post-biopsy complications in the present series encompassing general, major, and minor postoperative events. We reported a 1% rate of postoperative intracerebral hemorrhage requiring surgical evacuation, which is consistent with previous series [5, 18, 21, 27]. We screened systematically for intracerebral hematoma, independently from their clinical relevance, on early postoperative imaging, and observed a lower prevalence of postoperative intracerebral hematoma > 10 mm in diameter, after frameless robot-assisted stereotactic biopsies compared to those after image-guided biopsies, in those patients who received a preoperative administration of an anticoagulant and/or antiplatelet therapy. This cannot simply be related to the different number of biopsy samples between centers A and B: according to previous studies, there is no correlation between the number of biopsy samples and postoperative hemorrhage [32]. This can reflect a significant difference in trajectory planning, which is a key step in stereotactic biopsy: while the neurosurgical team in center A performs a two-step neurosurgical planning, the neurosurgical team in center B performs the planning in the operating room, shortly before surgery [31]. This difference in terms of postoperative intracerebral hematoma suggests using preferentially the frameless robot-assisted stereotactic biopsy procedure in the population at risk of postoperative intracerebral hemorrhage. The 1-month 3% mortality rate we reported is similar to that reported in other intraparenchymal brain lesion biopsy series whatever the neurosurgical technique and is related to the evolution of the disease rather than a surgically induced event [21, 22, 27].

Generalizability

The present study selected a homogeneous population of adult patients harboring a newly diagnosed brain lesion operated on with two different neurosurgical procedures. This is the first study comparing the efficacy and the safety of frameless robot-assisted stereotactic biopsies versus image-guided biopsies using a retrospective bicentric design. We analyzed data provided from a daily practice: our study is a result of a naturally occurring experiment of different management strategies in two tertiary French centers. The French healthcare system ensured comparable study populations. The present results support the use of these two neurosurgical procedures to perform a biopsy of brain lesions in adults.

Limitations

These findings should be interpreted with caution, given the retrospective design and the exploratory design of the statistical analyses limiting the generalizability of the results. The original design (comparison between two neurosurgical centers favoring a particular biopsy procedure) limited selection biases (the practitioner was not tempted to favor one technique or another based on experience, preference, or tumor characteristics). Nevertheless, this may have led to a lack of comparability between groups, especially considering the use of antithrombotic/anticoagulant treatment, possibly leading to statistical biases. Further confirmatory analyses are required to reproduce the present results.

Conclusions

In this original bicentric retrospective study, we have shown that frameless robot-assisted stereotactic and image-guided intraparenchymal brain lesion biopsies are both safe and efficient techniques. The two techniques are comparable, except with regards to operating time (shorter with image-guided biopsies) and the presence of a hematoma > 10 mm in patients on early postoperative CT scan in the subgroup of patients with anticoagulant and/or antiplatelet therapy (higher with image-guided biopsies).