Volume of interest identification and imaging visualization
The definition of the anatomical region, which was selected as the putative location of DL, was based on landmark cytological, histological, imaging works and atlases, defining OR positioning at the level of the TPJ (Burgel 1999; Mori 2005; Burgel 2006), following these steps:
- visualization of the OR volumes, downloaded from the on-line Juelich Institute of Neuroscience and Medicine fiber tracts repository. These volumes have been produced by means of myelin-stained histological studies (Burgel 1999; Burgel 2006), and hereby visualized in a dedicated MNI-152 space (ICBM 2009c nonlinear symmetric);
- given the need to define the volume of interest (VOI) related to DL in the context of the OR, the posterior thalamic radiation (PTR) label obtained from the JHU-ICBM-DTI-81 WM atlas (Mori 2005) was applied to the OR volumes, so that anterior, posterior, lateral and mesial limits of the DL volume were defined;
- the inferior limit of the volume was chosen as the plane passing through the most anterior point of the calcarine fissure, parallel to that identified by the PTR label (Mori 2005), based on the segregation of OR fibers directed to the supra and infracalcarine cortices;
- the superior limit identified by the histological studies was preserved (Burgel 1999; Burget 2006), to avoid the risk of undersampling the eventual most superior fibers of the DL (Figure 1).
Once the VOI was defined (Figure 1), the useful brain-MR volumetric sequences of each patient were registered to the dedicated MNI space. Each tractography and the anatomical reconstruction of the electrode contacts were visualized throughout the multimodal imaging integration workflow available in our Centre (Cardinale 2012).
Ex-vivo anatomical study
Cadaver dissections were authorized from the Ethical Committee of the APSS of Trento. Two human cerebral hemispheres (one left and one right) were prepared according to the modified Klingler’s technique previously detailed (Sarubbo 2015). Microdissection was performed in April 2019 by a neurosurgeon (MR), with the supervision of two WM dissection experienced neurosurgeons (SS, ADB). Dissection started with the removal of both sulcal gray matter of the lateral aspect of the brain and U-fibers. Once the three components of the superior longitudinal fascicle have been detected, the posterior portion of the inferior fronto-occipital fascicle (IFOF), the inferior longitudinal fascicle, the OR and the tapetum were identified in a stepwise manner, from outside to inside (De Benedictis 2014). Grey matter at the tip of cuneus and lingual cortex was preserved according to a cortex-sparing technique (Martino 2011), to obtain the correct definition of the bundle and its terminations territories. Moreover, precuneus cortical structures were removed, helping the identification of fascicles connected only to the occipital cortex, helping the exclusion to those terminating also in the parietal cortex (i.e. IFOF). All the WM bundles covering the OR were removed, layer by layer, until the unambiguous identification of the OR, following a stepwise technique (Sarubbo 2015).
In vivo studies
Clinical data were obtained on our prospectively maintained database. All patients or their guardians gave their informed consent. The local Ethical Committee approved this study in 2020 (ID 348-24062020).
We selected those cases on which DL was supposed to be investigated by anatomical in-vivo studies (DTI-based tractography) or neurophysiological explorations (VEPs extracted from SEEG electrodes [Microdeep Intracerebral Electrodes-D08, Dixi Medical] following flash administration), excluding those patients with lesions (observed at the presurgical brain-MR) undermining the OR and/or those with clinical abnormalities involving the visual systems.
In particular, we considered all the patients who underwent SEEG at the “C. Munari” Epilepsy Surgery Centre, in Milan, between February 2017 and March 2021. SEEG is a methodology providing relevant information about seizure onset zone (SOZ) when the non-invasive analysis of anatomo-electro-clinical correlations does not allow a clear definition of it. SEEG is generally followed by SOZ intraprocedural radiofrequency thermocoagulation and/or resective surgery (Cardinale 2019).
The multimodal imaging, including tractography and the anatomical reconstruction of the electrodes contacts, was visualized throughout the imaging workflow integrated in our Centre (Cardinale 2012), once brain-MR volumetric sequences and post-processed datasets of each patient were registered to the dedicated MNI space.
All SEEG implanted patients underwent a preoperative work-out including dedicated brain-MR sequences (including DTI). MR datasets were acquired with a 1.5T Philips Achieva scanner, using a receive coil head SENSE 8 with 8 channels (Supplementary File). After DICOM to nifty image conversion, brain volumes were extracted from DTI and T1 images. The DTI volumes were then co-registered to the first volume acquired without gradients applied and corrected for noise reduction and eddy current algorithms using the specific tools of the FMRIB Software Library (FSL; http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/). Fractional Anisotropy images and color maps indicating the principal diffusion directions (i.e. fiber orientations) were calculated from the DTI. DTI calculations and fiber tracking were performed with FSL-based FDT Diffusion toolkit, using probabilistic methods based on multiple regions-of-interest approach (ProbtrackX). OR tractography started with the definition of the seed-region of interest (ROI), drawn at the level of the lateral geniculate nucleus. Way-point ROIs were localized at the level of the green fibers (antero-posteriorly oriented) in the occipital cortex (Kamali 2014). Additional ROIs were considered, in case of local anatomical distortion, or fibers which are tracked outside the occipital lobe. The 3D structure of the OR was defined as the volume encompassed by the 10% of maximum value of output path distribution after being co-registered with the 3D reference sequence (i.e. volumetric T1). Each patient output path distribution was co-registered, through affine transformation (12 degree of freedom), to the dedicated MNI space, which was considered as the reference common space. A neurosurgeon with expertise in the field of advanced neuroimaging (MR) analyzed the position of the OR with as compared to the VOI volume. Tractography analysis was considered to be “positive" when at least a portion of the OR tractography lied in the VOI; on the other hand, if tractography did not lie at this level, it was considered as “negative”.
For each hemisphere, a probability map was finally calculated summing up the binarized 3D OR structures of all the patients and dividing by the number of reconstructed tracts, in the dedicated MNI space.
SEEG DATA RECORDING AND PROCESSING
We selected patients whose implantations explored the VOI (Figure 1). For these patients, VEPs were collected during SEEG monitoring.
For each electrode targeting the VOI, we first identified all the contacts in the WM, thus excluding those directly recording from the cortex. This identification first relied on neuroimaging, i.e. the co-registration of post-implantation cone-beam computed tomography scan with the pre-implantation MR. Since co-registered imaging datasets suffer from accuracy limitation, the identification of the contacts localized in the WM was assisted by the analysis of the signal recorded in each contact (Greene et al., 2020)
Visual stimulation was performed routinely to identify the explored regions involved in the visual processing. Patients wear goggles, and receive 100 bilateral visual stimulations (i.e., flashes) at 1 Hz, with an intensity of 3 cd/m2. Visual evoked potentials are extracted for each channel as the average of the first 50 trials and of the last 50 trials in the time window (0, 200 ms), following the stimulus delivery. Two electrophysiologists (IS, FMZ) blindly evaluated the traces for each electrode, and identified the contact presenting the earliest and most pronounced deflection.
To quantitatively assess the prevalence of phase-locking components in the identified channels relative to the neighbor territories, we further computed the inter-trial coherence (ITC) (Delorme and Makeig, 2004) for each of the identified channels, and compared statistically this variable against the channels located before and after the selected one. In other words, we evaluated the specificity of the identified features against the same signals recorded at 3.5 mm of distance along the electrode direction. ITC was computed in the interval (0, 200 ms), and for frequencies ranging from 0 and 500 Hz. Statistical comparison contrasting distribution of identified contacts against the adjacent one was performed via t-test (p < 0.01, 1000 permutation).
Once neurophysiological and tractography datasets have been separately and blindly analyzed, data were matched. In particular, each recording SEEG contact was labelled as intersecting, tangential or negative, according to its position relative to the reconstructed OR fibers. Intersecting were defined those contacts which lied, at least partially, within the reconstructed OR. Contacts were defined tangential if their minimum distance from OR was lower than 1.5 mm. In the case of a greater distance, they were labeled as negative. To evaluate comparatively the performance of tractographic and electrophysiological assessments, we evaluated the convergence between the two sets of results, and in particular how many times the “relevant contacts” as identified by VEPs coincided with intersecting contacts as revealed by fiber tracking.