Dual-energy CT for automatic organs-at-risk segmentation in brain-tumor patients using a multi-atlas and deep-learning approach

Brent Van Der Heyden,Patrick Wohlfahrt,Daniëlle B P Eekers,Frank Verhaegen,Karin Terhaag,Esther G C Troost,Christian Richter

doi:10.1038/s41598-019-40584-9

Abstract

In radiotherapy, computed tomography (CT) datasets are mostly used for radiation treatment planning to achieve a high-conformal tumor coverage while optimally sparing healthy tissue surrounding the tumor, referred to as organs-at-risk (OARs). Based on CT scan and/or magnetic resonance images, OARs have to be manually delineated by clinicians, which is one of the most time-consuming tasks in the clinical workflow. Recent multi-atlas (MA) or deep-learning (DL) based methods aim to improve the clinical routine by an automatic segmentation of OARs on a CT dataset. However, so far no studies investigated the performance of these MA or DL methods on dual-energy CT (DECT) datasets, which have been shown to improve the image quality compared to conventional 120 kVp single-energy CT. In this study, the performance of an in-house developed MA and a DL method (two-step three-dimensional U-net) was quantitatively and qualitatively evaluated on various DECT-derived pseudo-monoenergetic CT datasets ranging from 40 keV to 170 keV. At lower energies, the MA method resulted in more accurate OAR segmentations. Both the qualitative and quantitative metric analysis showed that the DL approach often performed better than the MA method.

Highlights

In the clinical radiotherapy workflow, the targeted tumor volume and surrounding organs-at-risk (OARs) are manually delineated on image datasets derived from computed tomography (CT), often in combination with magnetic resonance imaging (MRI)
To our knowledge, so far no studies explored the use of such methods on dual-energy CT (DECT) image datasets, which provide additional tissue information contributing to a reduction of the intra-observer variability of physicians[9]
As the pseudo-monoenergetic image (PMI)-70 dataset was originally used by the radiation oncologist for manual contouring, it was defined as reference PMI in the further analysis

Summary

Introduction

In the clinical radiotherapy workflow, the targeted tumor volume and surrounding organs-at-risk (OARs) are manually delineated on image datasets derived from computed tomography (CT), often in combination with magnetic resonance imaging (MRI). Multi-atlas (MA) or deep-learning (DL) methods have been investigated for automatic image segmentation[3] Both approaches use a set of labeled medical image datasets as input for model training and application. This study first aims to quantitatively evaluate pseudo-monoenergetic CT datasets of different photon energies ranging from 40 keV to 170 keV for OAR segmentation in primary brain-tumor patients using an in-house developed 3D MA and 3D DL based image segmentation method. Two experienced radiation oncologists and one experienced radiation technologist performed a qualitative scoring to assess the clinical relevance and accuracy of automatic OAR segmentation. For this evaluation, both methods were applied on two pseudo-monoenergetic CT datasets of different energy (40 keV and 70 keV)

Objectives

Methods

Results

Discussion

Conclusion