In the present meta-analysis, we report 15 studies, encompassing 867 patients, with various types of LE and NLE. Most common surgical indications in LE were dysembrioplastic neuroepithelial tumor (DNET) and gangliogliomas included together or separately, cavernomas, HH, or primary brain tumors and in NLE were sclerosis, dysplasia, ischemia and gliosis. With the use of iMRI, the rate of complete or adequate microsurgical resection significantly increased from 74.8% (67.3-82.3%) to 97.6% (95.8-99.3%). This was translated in an overall clinical benefit of Engel class I at last follow-up of 77.1% (71-83.2%). While this meta-analysis included various types of epilepsy, the favorable clinical outcome is fairly comparable to what one would expect in temporal lobe epilepsy (TLE), which is universally considered to have the best postoperative seizure control.
The surgical aim of resective epilepsy surgery is either complete excision or disconnection of the epileptic network, with preservation of eloquent cortex(32). In order to achieve this, presurgical correct diagnosis is mandatory. Modern epileptologists can use multimodal diagnostic tools, including semiology analysis, electrophysiological recordings, functional testing and complex neuroimaging techniques(32). All these remain complementary and should define the location and boundaries of the epileptic zone. In challenging cases, stereo-electroencephalography (SEEG) by depth electrodes is considered of major help for evaluation and tailored navigated resection, even in patients with non-lesional or extratemporal focal epilepsy(33). In fact, it has been already acknowledged that presence of a lesion does not necessarily correlate with an epileptogenic network solely driven by that particular lesion. Moreover, a lesion itself can recruit and further drive other epileptogenic zones in the brain, making diagnosis even more challenging(34, 35). The clinical aim of resective epilepsy surgery remains the best possible seizure control, while experiencing the least possible neurologic or neuropsychological impairment. This is achieved by a maximized resection, which is mandatory for good seizure outcome.
Surgically accessible pathologies associated with medically refractory epilepsy often include hippocampal sclerosis, long-term epilepsy-associated tumors, cavernous malformations or focal cortical dysplasia(36, 37). Presence of a visible lesion on MRI correlates with better postsurgical outcomes, as does complete surgical excision of such lesions(38, 39). In contrast, intrinsic epilepsy-associated tumors might have diffuse borders and less accessible anatomical locations, such as adjacent to motor or speech areas(22, 25).
Optimal resection goals are mostly attained in temporal locations, translating into good surgical outcomes as opposed with those in extratemporal ones(27). In a recent meta-analysis(40) on long-term postsurgical outcomes seizure freedom was reported in 64% of TLE and only 34% in extra TLE, particularly for cortical malformations. Of note, one of the important reasons was incomplete removal, due to proximity to eloquent motor areas and white matter tracts. In this context, iMRI can be a valuable therapeutic adjunct, together with neuronavigation. Its utility should be seen in the context of the specific pathology related to medically intractable seizures. In TLE, iMRI can document the extent of mesial and neocortical resections as alternative to classical tailored anterior or posterior temporal lobe resections rather than standard lobectomy in attempt to maximize preservation of adjacent parenchyma(18).
In extra TLE, a variety of histopathological findings can be identified. Not all of them are necessarily well defined on neuroimaging (in contrast to hippocampal sclerosis for example). Some are further located near functional cortex or fiber tracts and thus are not necessarily amenable to complete resection without neurological impairment(25). Combined with BOLD and tractography, iMRI can offer a higher extent of resection. It can further compensate the brain shift and can adjust inaccuracy of coregistration, improve functional mapping and fiber tracking.
In cortical dysplasia, the concept of seizure outcome as related to extent of resection is also applicable. iMRI can be a valuable tool, as it is often more subtle and difficult to visually differentiate cortical dysplasia from normal brain parenchyma, in terms of intraoperative macroscopic aspects(1), causing even in the experienced neurosurgeon a feeling of false complete resections(21).
One of the limitations is the retrospective nature of the included studies, the small sample size of each of them, as well as the lack of randomized controlled trials. A second limitation is the use of multiple surgical procedures, depending on the main pathology, inside the same report. A third limitation is that not all studies included the complete spectrum of pathologies. A fourth limitation is that we cannot comment on the surgeon’s personal insight, leading to the decision-making. A fifth limitation is related to a potential selection bias of choosing the microsurgical approach, while including iMRI in the surgical armamentarium. Moreover, there was a progressive increase in the iMRI field during time; as such, our results have to be balanced with the constant evolution of intraoperative technologies. Only one study included in the present meta-analysis used 3 T iMRI. An open question is what should be the minimum postoperative follow-up in these cases (e.g. 12 months?). A technical limitation can be related to errors during calculation of functional data, target registration of the navigation systems, registrations between pre- and postoperative MRI etc. As an alternative for iMRI and neuronavigation, in selected cases, awake craniotomy represents a valuable option in experienced centers(47). A further limitation relates to specific terminology of what some authors consider as LE, while other as NLE. A last, yet important limitation, is the lack of randomized controlled trials, which would increase the level of evidence.