VASARI-auto
This is the codebase for automated VASARI characterisation of glioma, as detailed in our article.
Table of Contents
What is this repository for?
The VASARI MRI feature set is a quantitative system designed to standardise glioma imaging descriptions.
Though effective, deriving VASARI is time-consuming to derive manually.
To resolve this, we release VASARI-auto, an automated labelling software applied to open-source lesion masks.
VASARI-auto is a highly efficient and equitable automated labelling system, a favourable economic profile if used as a decision support tool, and offers non-inferior survival prediction.
Usage
VASARI-auto requires only a tumour segmentation file only, which allows users to apply code efficiently and effectively on anonymised lesion masks, for example in using the output of our tumour segmentation model (paper | codebase).
For segmentation files, this code assumes that lesion components are labelled within a NIFTI file as follows:
- Perilesional signal change = 2
- Enhancing tumour = 3
- Nonenhancing tumour = 1
Relying on tumour segmentation masks and geometry only, VASARI-auto is deterministic, with no variability between inference, in comparison to when cases are reviewed by different neuroradiologists.
The time for neuroradiologists to derive VASARI is substantially higher than VASARI-auto (mean time per case 317 vs. 3 s).
We identify that the best-performing survival model utilised VASARI-auto features instead of those derived by neuroradiologists.
VASARI-auto is demonstrably equitable across a diverse patient cohort (panels B and C).
The Medical Research Council; Wellcome Trust; UCLH NIHR Biomedical Research Centre; Guarantors of Brain; National Brain Appeal; British Society of Neuroradiology.