Rapid Identification of Brain Tumours with Differential Mobility Spectrometry

Brain tumours are a major source of disease burden worldwide. Currently, the only established method for rapid brain tumour diagnostics during surgery is the histopathological analysis of frozen sections. However, the method has several disadvantages: the performance of this is difficult and often subjective, and the increasingly important genetic alterations remain undetected. This directly calls for novel solutions. Differential ion mobility spectrometry (DMS) is a modality that characterises substances based on the mobility differences of their gaseous ion swarms in an alternating electric field. Previously, it has shown potential for identifying biological tissues. To our knowledge, this is the first time DMS has been studied for brain tumour identification. The purpose of this study was to explore the performance and usability of DMS in rapid (near real-time) brain tumour diagnostics ex vivo.

Freshly frozen brain tumour or control samples were prepared from 65 patients (adult and paediatric) altogether covering the most common histologies and including up-to-date genetic information. In addition, tumour samples from 12 patients were cultured and analysed as living glioma organoids. Both the frozen and living samples were multiplied into hundreds of smaller specimens that were vaporised by either electrocoagulation or laser, and the resulting smoke was analysed using DMS. The analysis provided substance- specific dispersion spectra, which were consequently identified and classified using computational data-algorithms.

In a seven-class classification of several histologically different tumour types, a classification accuracy (CA) of 50% was achieved, which improved up to 83% after the exclusion of a confounding factor (samples immersed in the Tissue- Tek® conservation medium). In selected head-to-head comparisons, the CA was enhanced to up to 94% (GBM vs. LGG). When classifying malignant glioma samples based on their IDH mutation status, the CA was 86% with fresh-frozen and 90% with actively proliferating samples in Study IV. The CA for detecting 1p/19q codeletion among living IDH mutated glioma samples was formidable (98%) and for CDKN2A/B deletion 86%, but the sample sizes were imbalanced. In the differentiation of the most common paediatric brain tumour samples (medulloblastomas, ependymomas and pilocytic astrocytomas), the classification accuracy was 65% and increased up to 75% when pilocytic astrocytomas and ependymomas were pooled together.

DMS can identify and classify the most common brain tumour samples ex vivo based on both different histology and genetic profiling with considerable accuracy. The CA improves when actively proliferating, cultured tumour organoids are used, which sets high expectations for future in vivo study.