Mapping Whole Exome Sequencing to In Vivo Imaging with Stereotactic Localization and Deep Learning

Objective: Intra-tumoral heterogeneity complicates the diagnosis and treatment of glioma. Multiparametric imaging enhances heterogeneity characterization, but limitations exist in assessing cellular and molecular properties across space due to a lack of easily accessible, co-located pathology and genomic data. This study presents a multi-faceted approach combining stereotactic biopsy with standard clinical open-craniotomy for sample collection, voxel-wise analysis of MR images, regression-based generalized additive model (GAM), and whole-exome sequencing. This work aims to demonstrate the potential of machine learning algorithms to predict variations in cellular and molecular tumor characteristics.  

Methods: This retrospective study enrolled ten treatment-naïve patients with radiologically confirmed glioma. Each underwent a multiparametric MR scan (T1_W, T1_W-CE, T2_W, T2_W-FLAIR, DWI) prior to surgery. During standard craniotomy, at least 1 stereotactic biopsy was collected from each patient, with screenshots of the sample locations saved for spatial registration to pre-surgical MR data. Whole-exome sequencing was performed on flash-frozen tumor samples, prioritizing the signatures of five glioma-related genes: IDH1, TP53, EGFR, PIK3CA, and NF1. Regression was implemented with a GAM using a univariate shape function for each predictor. Standard receiver operating characteristic analyses were used to evaluate detection, with AUC (area under curve) calculated for each gene target and MR contrast combination.  

Results: Mean AUC for five gene targets and 31 MR contrast combinations was 0.75±0.11; individual AUCs were as high as 0.96 for both IDH1 and TP53 with T2_W-FLAIR and ADC and 0.99 for EGFR with T2_W & ADC. An average AUC of 0.85 across the five mutations was achieved using T1_W, T2_W-FLAIR, and ADC combined.  

Conclusion: These results suggest the possibility of predicting exome-wide mutation events from non-invasive, in vivo imaging by combining stereotactic localization of glioma samples and a semi-parametric deep learning method. This approach holds potential for refining targeted therapy by better addressing the genomic heterogeneity of glioma tumors.