Fully automated brain resection cavity delineation for radiation target volume definition in glioblastoma patients using deep learning

Ekin Ermiş,Urspeter Knecht,Daniel M Aebersold,Evelyn Herrmann,Mauricio Reyes,Robert Poel,Peter Manser,Michael K Fix,Raphael Meier,Marcela Blatti-Moreno,Alain Jungo

doi:10.1186/s13014-020-01553-z

Ekin Ermiş, Urspeter Knecht + Show 9 more

Open Access

https://doi.org/10.1186/s13014-020-01553-z

Copy DOI

Abstract

BackgroundAutomated brain tumor segmentation methods are computational algorithms that yield tumor delineation from, in this case, multimodal magnetic resonance imaging (MRI). We present an automated segmentation method and its results for resection cavity (RC) in glioblastoma multiforme (GBM) patients using deep learning (DL) technologies.MethodsPost-operative, T1w with and without contrast, T2w and fluid attenuated inversion recovery MRI studies of 30 GBM patients were included. Three radiation oncologists manually delineated the RC to obtain a reference segmentation. We developed a DL cavity segmentation method, which utilizes all four MRI sequences and the reference segmentation to learn to perform RC delineations. We evaluated the segmentation method in terms of Dice coefficient (DC) and estimated volume measurements.ResultsMedian DC of the three radiation oncologist were 0.85 (interquartile range [IQR]: 0.08), 0.84 (IQR: 0.07), and 0.86 (IQR: 0.07). The results of the automatic segmentation compared to the three different raters were 0.83 (IQR: 0.14), 0.81 (IQR: 0.12), and 0.81 (IQR: 0.13) which was significantly lower compared to the DC among raters (chi-square = 11.63, p = 0.04). We did not detect a statistically significant difference of the measured RC volumes for the different raters and the automated method (Kruskal-Wallis test: chi-square = 1.46, p = 0.69). The main sources of error were due to signal inhomogeneity and similar intensity patterns between cavity and brain tissues.ConclusionsThe proposed DL approach yields promising results for automated RC segmentation in this proof of concept study. Compared to human experts, the DC are still subpar.

Highlights

Automated brain tumor segmentation methods are computational algorithms that yield tumor delineation from, in this case, multimodal magnetic resonance imaging (MRI)
The reference segmentation generated by fusion, were all accepted by the three raters upon review
In terms of relative volume error, we found a median error of − 13.17%, (IQR: 24.17%) between automatic and reference segmentations, which indicates the deep learning (DL) method underestimated the resection cavity (RC) with respect to the raters

Summary

Introduction

Automated brain tumor segmentation methods are computational algorithms that yield tumor delineation from, in this case, multimodal magnetic resonance imaging (MRI). One of the most tedious and time-consuming tasks in radiotherapy planning is target and organ at risk (OAR) contouring This is still done manually in a slice by slice fashion, using multiple magnetic resonance imaging (MRI) sequences [2]. For GBM in particular, post-op target definition shows substantial inter-rater variability even amongst advanced experts [8]. In this regard, automated contouring methods would be very useful for RT target volume definition. Automatic segmentation, where no interaction of the user is required, has the potential to substantially limit the time for target volume and OAR definition. It can introduce a more consistent and reproducible standard for volume definition leading to a better agreement among institutes and possibilities for global implementation

Methods

Results

Discussion

Conclusion