Surg-01. Ultra-Fast Deep-Learned CNS Tumor Classification Using Sturgeon; The Impact on Strategy for the Pediatric Neurosurgeon

INTRODUCTION Central nervous system tumors still represent one of the most challenging fields to create curation in pediatric oncology. The primary treatment always starts with a neurosurgical intervention to obtain tumor tissue for diagnosis. In most cases this intervention is a maximal safe resection of the tumor. The oncological prognostic value of extent of resection is determined by the precise molecular diagnosis and per-operative ultra-fast molecular classification of central nervous tissue provides the neurosurgeon relevant information for choosing the right strategy. This per-operative molecular diagnosis determines the balance between the need of radicality versus the prevention of neurological morbidity in the individual patient. We shall report on our neurosurgical experiences with the per-operative use of Sturgeon in our national practice of pediatric neuro-oncology in the Netherlands. METHODS The ultra-fast AI-based nanopore sequencing method, Sturgeon, has been implemented as per-operative diagnostic in most neurosurgical tumor resections in our center (1). An overview of our nanopore sequencing pipe- and timeline will be presented and its impact on neurosurgical strategies will be shared. RESULTS Reliable and accurate diagnoses can be obtained within 40 minutes after starting sequencing in 45 out of 50 retrospectively sequenced samples (while abstaining from diagnosis of the other 5 samples). The implementation in real time surgery results in a turnaround time between taking out sample material to diagnosis in less than 90 minutes. The most important result is that there is a high sensitivity and virtually no false positives since it is refraining from diagnosis when there is too much uncertainty. Several illustrative cases with impact on the neurosurgical strategy have been collected. CONCLUSION Ultra-fast deep-learning CNS tumor classification using Sturgeon is feasible. Results of per-operative implementation illustrate the impact on neurosurgical strategy. 1. Ultra-fast deep-learned CNS tumour classification during surgery. Vermeulen C et al. Nature. 2023 Oct;622(7984):842-849.