Abstracts

Rationale: Stereoelectroencephalography (SEEG) involves the stereotactic placement of electrodes within targets in the brain. Ictal and interictal SEEG recordings from implanted electrodes help define the seizure onset zone and propagation pathways in focal drug resistant epilepsy when the non-invasive pre-surgical evaluation is discordant. The overall risk of haemorrhage during SEEG implantation is 1% per patient. Various forms of vascular imaging are performed to help identify critical vasculature that should be avoided during SEEG planning. Gadolinium enhanced MR venography has been used as the sole means of vascular segmentation and increased haemorrhage rates have not been reported. DSA is the gold-standard for intracranial vascular imaging but is invasive, requires a general anaesthetic in paediatric patients and involves additional radiation exposure. Once anatomical structures for SEEG sampling are identified the trajectory planning can be undertaken manually or using computer assisted planning (CAP). A safe trajectory should avoid critical vasculature, pial boundaries, conflict with other electrodes, minimise intracranial length and maximise grey matter sampling. Previous studies have shown that CAP optimises these parameters. Trajectories are clinically infeasible in 20-30% of cases when rated independently due to proximity to vessels that were not identified by segmentation. DSA identifies many more vessels and a dichotomy exists between increasing the number of segmented vessels for safety and having too many which unnecessarily restricts electrode placement. We therefore sought to estimate the size at which vessels become clinically significant for SEEG planning. Methods: Eleven consecutive patients who had DSA and underwent SEEG implantation were retrospectively selected. The intracranial position of the electrodes were segmented from post-implantation CT scans. Each electrode was manually reviewed and conflicts between the electrodes and vasculature were recorded. The diameter of vessels and position where the conflict occurred was noted. A collision detection algorithm based on DSA vessel segmentation was instituted to determine if vessel conflict detection could be automated. Results: In total 46 vessel conflicts were identified in 11/11 patients occurring 0-3 times within an individual electrode and 1-12 times per patient. Inaccuracy in the implantation method resulted in vessel conflict. No patients had a clinically significant haemorrhage. Average conflicting vessel size was 1.1 mm (range 0.6-2.4 mm). Automated collision detection had a false negative rate of 30% (11/46) and false positive rate of 19% (11/57). Conclusions: Vessel conflicts occurred on average 4 times per patient. We are unable to identify whether the vessels are venous or arterial in nature and whether vessels are displaced or transected by electrodes. As no bleeding occurred it was not possible to assess critical vessel size but no haemorrhages occurred in vessels < 2.4 mm. It may be possible to discount vessels < 1 mm from SEEG planning. Funding: N/A