Background: The authors have conducted a prospective study to evaluate the amount and course of brain shift during microsurgical removal of supratentorial cerebral lesions, and to assess factors which potentially influence these shifts.
Method: In 61 patients the displacement of 2-3 cortical landmarks on the cerebral surface was dynamically quantified during surgery, i.e. during dissection of the tumour at the estimated half-time of surgery, and at the end of microsurgical removal of the cerebral lesion using the neuronavigation system EasyGuide Neuro. In 14 of these patients the displacement of a subcortical landmark was additionally analysed. Age of the patients, preoperative midline shift, location of the lesion, lesion volume, depth of the lesion below the cortical surface, presence or absence of oedema, and size of the craniotomy were analysed for potential influence on the amount of brain shift. Correlations were analysed for all patients together and for the subgroups of vault meningiomas (n=10), gliomas (n=30), and nonglial intra-axial lesions (n=21).
Findings: The mean displacement of the cortical landmarks ranged between 0.8 and 14.3 mm (mean: 6.1 mm, standard deviation: 3.4 mm) during surgery (10-210 minutes [mean: 50.7 minutes, standard deviation: 34.5 minutes] after dura opening) and between 2.4 and 15.2 mm (mean: 6.6 mm, standard deviation: 3.2 mm) at the end of microsurgical removal of the tumourous cerebral lesions (20-375 minutes [mean: 107.2 minutes, standard deviation: 65.6 minutes] after dura opening). Significant correlations (p<0.01) for the entire patient group were found between brain shift and tumour volume, midline shift, and size of the craniotomy, respectively. For the subgroup of vault meningiomas a significant correlation (p<0.01) between brain shift and patient age was found. For the subgroup of gliomas a significant correlation (p<0.01) between brain shift and tumour volume, midline shift and size of the craniotomy, respectively, was found. For the subgroup of nonglial intra-axial lesions a significant correlation (p<0.01) between brain shift and midline shift and between brain shift and size of the craniotomy was found. The quantity of shared common variance ranged between 10-50%. Performing a discriminant analysis, lesion volume was the only certain factor influencing brain shift intra-operatively as well as at the end of lesion removal. 58.5% of the extent of brain shift could be correctly classified by the tumour volume as the only discriminating variable during dissection of the tumour and at the end of surgery. Comparing superficial with subcortical brain shift over the same time period, a mean superficial shift of 4.6 mm (1.6-10.8 mm, standard deviation: 2.8 mm) and a mean subcortical shift of 3.5 mm (1.0-7.7 mm, standard deviation: 2.3 mm) was found. A highly significant Spearman correlation (Rho:.97, p<0.001) between superficial and subcortical brain shift emerged. Shifting of superficial landmarks exceeded shifting of subcortical structures in all patients.
Conclusions: The data demonstrate the dynamics of brain shift and the limits of conventional neuronavigation and add additional support for the unavoidable inaccuracy of contemporary neuronavigational systems once the cranium is opened. Brain shift leads to a significant loss of reliability of neuronavigation systems during microsurgical removal of intracranial lesions and there are differences of the course and the amount of brain shift in relation to special subgroups of supratentorial cerebral lesions. However, because of the heterogeneous nature of lesions neurosurgeons have to remove, the modest quantity of shared common variance, and the differences between superficial and subcortical brain shift, it seems unlikely that the amount and course of brain shift become exactly predictable pre-operatively. Only an intra-operative update of image data should have the capacity to overcome this fundamental problem of modern neuronavigation.