Medical physics
Target delineation in post-operative radiotherapy of brain gliomas: Interobserver variability and impact of image registration of MR(pre-operative) images on treatment planning CT scans

https://doi.org/10.1016/j.radonc.2005.03.012Get rights and content

Abstract

Background and purpose

To investigate the interobserver variability of intracranial tumour delineation on computed tomography (CT) scans using pre-operative MR hardcopies (CT+MR(conv)) or CT-MR (pre-operative) registered images (CT+MR(matched)).

Patients and methods

Five physicians outlined the ‘initial’ clinical tumour volume (CTV0) of seven patients affected by HGG and candidates for radiotherapy (RT) after radical resection. The observers performed on screen-tumour delineation using post-operative CT images of the patients in the treatment position and pre-operative MR radiographs (CT+MR(conv)); they also outlined CTV0 with both CT and corresponding MR axial image on screen (CT+MR(matched)). The accuracy of the image fusion was quantitatively assessed. An analysis was conducted to assess the variability among the five observers in CT+MR(conv) and CT+MR(matched) modality.

Results

The registration accuracy in 3D space is always less than 3.7 mm. The concordance index was significantly better in CT+MR(matched) (47.4±12.4%) than in CT+MR(conv) (14.1±12.7%) modality (P<0.02). The intersecting volumes represent 67±15 and 24±18% of the patient mean volume for CT+MR(matched) and CT+MR(conv), respectively (P<0.02).

Conclusions

The use of CT and MR registered imaging reduces interobserver variability in target volume delineation for post-operative irradiation of HGG; smaller margins around target volume could be adopted in defining irradiation technique.

Section snippets

Patient population and image acquisition

This is a retrospective study on patients who underwent diagnostic pre-operative MR, surgical resection and radiotherapy at our Institution. Between November 1999 and December 2001, seven patients affected by glioblastoma multiforme (n=5) or anaplastic oligodendroglioma treated with macroscopically total resection for whom preoperative MR scans were available, were selected for this study. Five were males, median age was 51.5 years (range: 42–69) and median Karnofsky performance status was 90

Results

In the qualitative evaluation of CT/MR registration, the physician found good spatial correspondence in each image pair. Quantitative registration accuracy values are summarised in Table 1: for all five patients the maximum error in 3D space is less than or close to the slice thickness.

The maximum dimensions of volumes measured on BEV plots along the three main axes are similar for CT+MR(conv) and CT+MR(matched) procedures: the mean values are 37.8±13.7 vs 38.6±12.9 mm along cranio–caudal

Discussion

In this work, registration was carried out by a surface-matching algorithm [30]; this registration approach has been widely investigated by many authors, reporting, in the case of brain application, an accuracy of within a few millimetres [42]. The analysis of registration accuracy, carried out in our study on five patients, confirms previous data, showing an average residual misalignment error of 2 mm or less, indicating the adequacy of the technique in correcting not only for spatial

Conclusions

There exists quite a great degree of inter-observer variability in target volume delineation of high-grade gliomas submitted to total resection.

The use of CT and MR registered images greatly reduces uncertainties in the spatial location of target volume within the skull. This is more evident for cases in which MR and CT images are acquired with the patient head in different positions.

References (43)

Cited by (51)

  • Anatomical and functional MR imaging to define tumoral boundaries and characterize lesions in neuro-oncology

    2020, Cancer/Radiotherapie
    Citation Excerpt :

    Even if 2D FLAIR is currently the basis for initial diagnosis recommendations, this may change in the years to come [31]. Three-dimensional sequences are of special interest when planning conformal radiation therapy, as registration of 3D sequences onto CT is easier than with 2D images, and the quality of this registration can affect the target delineation and the resulting treatment plan [32]. Identifying more effective brain tumour therapies was several years ago partially limited by the lack of reliable criteria to determine tumour response and progression.

  • Modern Radiation Therapy Planning and Delivery

    2019, Hematology/Oncology Clinics of North America
  • Inter-rater agreement in glioma segmentations on longitudinal MRI

    2019, NeuroImage: Clinical
    Citation Excerpt :

    Manual segmentation by experts is still considered to be the gold standard and therefore required for quantitative interpretation of MR images and for the validation of automated segmentation algorithms. Reproducibility of manual segmentations has been investigated previously by others (Ben Abdallah et al., 2018a, 2016; Bø et al., 2017; Cattaneo et al., 2005; Gutman et al., 2013; Huber et al., 2015; Kleesiek et al., 2016; Kubben et al., 2010; Provenzale et al., 2009; Provenzale and Mancini, 2012; Sorensen et al., 2001; Weltens et al., 2001). Most of this work was focused on manual segmentation of preoperative MRI in glioblastoma, although a few of these studies consider longitudinal data (Huber et al., 2015; Kleesiek et al., 2016; Kubben et al., 2010; Meier et al., 2016).

View all citing articles on Scopus
View full text