Session Item

Ocular MRI has become a valuable tool in the diagnosis and therapy selection of uveal melanoma (UM) patients as it provides excellent soft tissue contrast and a 3D representation of the tumour and organs at risk. Currently, gross target volume (GTV) definition in proton therapy planning for uveal melanoma is based on a generic model of the eye and tumour, constructed using marker positions and 2D imaging. Several efforts are being undertaken to enable a 3D treatment planning for ocular PBT that includes a MRI based GTV. Before introducing 3D MR-based tumour models into the clinical workflow, it is important to know the inter-observer variability of the gross target volume (GTV) delineation on MRI for reliable treatment planning. Therefore, the aim of this study was to assess the inter-observer variation in GTV delineation of UM on MRI.

Six observers (two radiation oncologists, two radiologists and two ophthalmologists) delineated the GTV in ten different patients. Patients were scanned on a 3T MR scanner with a 4.7cm surface coil according to a previously described protocol (Ferreira 2019) and tumours were delineated on 3D T1gd and 3D T2-weighted scans (acquistion voxel size (0.8mm)³, resolution after reconstruction with zero filling 0.4x0.4x0.4 mm³ (T1gd) and 0.4x0.4x0.3 mm³ (T2)). For delineation, Big Brother training contouring software (Steenbakkers 2005) was used.

A median surface was formed based on the delineated GTVs of all observers. The interobserver variation was expressed as the median of the local standard deviation from the median surface. On each median surface, points adjacent to the sclera, vitreous, retinal detachment, or sclera and vitreous (edge) were labelled(Fig 1C). 

The average delineated tumour volume was significantly higher when delineated on T1gd (0.57 cm³) compared to the delineations on T2 (0.51 cm³, p = 0.01). The average interobserver variation appeared slightly higher on T1gd (0.41 mm) compared to T2 (0.35 mm), although the difference was not significant (p=0.12).

Analysing the separate regions, significant higher variations were found at the edge of the tumour compared to sclera and vitreous. Based on the individual delineations, the source of these higher local SDs at the edge was identified as variance in whether choroidal enhancement was included in the GTV or not. We also found that the contrast between retinal detachment and tumour on T2 was low in some patients, which underscores the importance of side-by-side evaluation of all available sequences.

The interobserver variation on T1gd (0.41 mm) and T2 (0.35 mm) are low with respect to the voxel size. Higher inter-observer variations were found at the edge of the tumour. However, these might be resolved with clear guidelines and after histopathological validation of the underlying reason for choroidal enhancement at the tumour edge. We recommend to delineate based on the T1gd-weighted scans, as parts of the tumour might be underestimated on T2.