You are here


3D MRI in Non-Invasive Proton Therapy Treatment Planning of Uveal Melanomas

Uveal melanoma (UM) is a rare subset of melanoma, but is the most common primary intraocular tumor in adults. In 85% of cases, melanocytes in the choroid are affected. UM can occasionally arise from the ciliary body or iris. With an incidence of about 6 cases per million people per year in Europe and the USA, UM is the second most common type of melanoma after cutaneous melanoma. Early diagnosis and local therapies are crucial, as survival correlates with the size and location of the tumor. Currently, about 50% of patients will develop systemic metastasis after treatment of the primary site.

Uveal Melanomas & Proton Beam Therapy

The management of localized UM is grouped into either globe-conserving treatment or enucleation treatment, i.e. eye removal surgery. The three well-accepted globe-conserving modalities are plaque brachytherapy, stereotactic radiosurgery (SRS), and proton beam therapy (PBT). Ruthenium plaque brachytherapy spares most of the healthy tissue and preserves visual function, but is only suitable for small to medium-sized UM. For large melanomas and those touching or close to the optic nerve, SRS or PBT are the current acceptable treatment options.

PBT uses a higher radioactive dose distribution than SRS, allowing for a highly specific applications to the lesion while sparing surrounding healthy tissues. As a consequence, PBT offers better clinical outcomes in terms of achieving tumor control and preserving visual acuity. Clinical reviews and 5-year survival rate studies indicated proton therapy to be successful in sparing the eye in 89% of the patients.

Proton therapy centers have become increasingly available worldwide since the late 1980s; however, the accessibility still remains limited. This is largely due to the high treatment costs and limited insurers covering PBT. Additionally, high set-up cost has limited the number of PBT centers, often causing patients to travel long distances to receive treatment. As a result, photon therapy is the most expensive overall treatment modality.

Background of the study

In the past few decades, the clinical workflow for PBT, including simulation (planning sessions to determine radiation dose and angles), surgical placement of clips onto the eye, and post-therapy routines have been described by numerous studies. To aim the photon beam at a correct gazing angle, surgical clips are sewn onto the sclera. For simulation, patients undergo CT scans. Radiation oncologists use the acquired images to define tumor delineation and to determine the optimal gazing angle of the eye. Dose parameters are also determined manually by a trade-off between target coverage and OAR (Organs-at-Risks) sparing.

Automated dynamic gaze mapping is highly recommended for PBT but is limited by the lack of 3D data for target and OARs. In order to find a type of eye model that might provide a more accurate representation of tumor volumes and organs at risk, the German Cancer Research Center (Deutsches Krebsforschungszentrum, DKFZ) in collaboration with Helmholtz-Zentrum have developed OCTOPUS (Ocular Tumour Planning UtilitieS). With this technology, the eye model is designed to interactively adapt to Computed Tomography (CT) or Magnetic Resonance Imaging (MRI) data of melanoma patients. This type of eye model might provide a more accurate 3D-representation of the tumor and organs at risk. This is essential as 3D information is highly useful and currently has restricted availability.

MRI In Ophthalmological Treatment Planning

Due to recent developments, ophthalmologists consider MRI a promising modality in ophthalmological treatment planning. Recent studies have demonstrated the MRI-compatibility of tantalum markers. Additionally, ultra-high field 3- and 7-tesla MRI sequences have been developed to improve UM and OARs delineation.

The specialists at the Paul Scherrer Institute (PSI) are using MRI to investigate how to best integrate tumor images into a virtual eye and tumor models. An optical tracking system is used to capture eye position in the OCTOPUS device and this helps in localization.

Some studies have demonstrated that DW-MRI (diffusion-weighted magnetic resonance imaging) can be useful in the evaluation of intraocular tumors such as choroidal melanoma. As more ophthalmologists and radiation specialists recognize the urgent need for MRI for a non-invasive ocular proton beam delivery, E. Fleury et al. conducted this study to assess the feasibility of adopting a full 3D MRI-based treatment planning approach for proton therapy in UM.

Materials and Methods

Patient Population

The researchers retrospectively analyzed the data from UM patients diagnosed in 2018 at the Department of Ophthalmology at Leiden University Medical Center. A total of 8 patients were included in the study population, 5 with right eye melanomas, and 3 with left eye melanomas. According to the TNM staging system, 3 patients were at T1 melanoma, 2 at T2 stage, and 3 at T3 stage. The uveal melanomas involved the equatorial region of the eye in 37.5% of the patients and were located posteriorly i.e., at a distance of 3mm from the optic nerve in 67.5% of the cohort. The median tumor thickness and basal diameter in the patient population were 6 and 12mm, respectively. The closest mean border-to-border 3D surface-mesh dish from the lesion to the optic nerve was 2.6mm and to the macula was 3.9mm. For 3 patients, the researchers determined tumor dimensions using high-frequency ultrasound. They employed fundus photography to inspect and adjust the contour of the tumor base and determine the macula’s actual position.

Methods In Detail

In-house algorithms were developed for semi-automatic segmentation of the eye. Fast-radial symmetry was used to estimate the center of the eye on T2-weighted images. This eye center was used as a reference to build a three-dimensional triangulated-surface mesh to detect the inner and outer borders of the sclera and the melanoma boundary.

Gross tumor volume (GTV) and organs-at-risk (OARs) were determined from the T1- and T2-weighted 7 Tesla high-resolution MRI scans. The contrast-enhanced MRI images were used to reconstruct the MR-eye. Extended eye contour was defined with a 2.5 mm isotropic margin derived from the GTV. A broad beam algorithm was used to determine the relative absorbed dose of a proton beam by the OARs. Clinically favorable gazing angles were defined from the simulated 3D rotational movement of the eye.


The dosimetric results for the uveal melanoma cohort were as follows:

  • In total, 441 simulated gazing angles were obtained per patient. And, for all 441 gazing angles, the target coverage was achieved (V95% > 95%), including margins.
  • In T1 tumors near the macula and optic nerve (patients 1 to 3), only well-defined extreme gazing angles (peripheral field of vision views) resulted in acceptable weighted-sum objective-function values.
  • In patients with T2 equatorial tumors (patients 4,6 and 7), there were more choices to select optimal gazing angles. This is because of the localization of the tumors and further distance from OARs. In patient 4, objective-function values were 0.46, 0.90, and 0.08 for iso-weighted, optic nerve, and macula prioritization, respectively. Downward or straight ahead gazing would provide the most macula sparing for this patient.
  • For patient number 6 with T3 equatorial tumor, the distance from the tumor to the optic nerve and macula were 7mm and 9mm, respectively. This allowed for a full sparing of both optic nerve and macula. Researchers demonstrated that sparing the optic nerve would result in selecting one gazing angle at the peripheral field of view. Gazing straight would enable a full sparing of the macular region.
  • For patient 5 with a posterior T2 tumor, looking straight would spare the macula. However, if the priority is to spare the optic nerve, the straight gazing should be avoided.
  • In the case of patient 8, the T3 nasal tumor almost covered the optic nerve fully and the tumor edge was more than 3mm away from the macular edge. For this patient, the top quadrant of the gazing angle corresponding to eyeball elevation would be optimal. As the objective-function value is higher – about 0.96 for both optic nerve and macula – straight gazing angle should be avoided.


High-resolution 3D MRI of uveal melanoma (UM) can be important not only for the diagnosis but also for non-invasive proton therapy planning.

The first and foremost particularity of this treatment planning proposed by E. Fleury et al. is the use of high-resolution 3D MRI for the segmentation of GTV and OARs. This approach is said to reduce the errors and uncertainties in tumor volume and shape definition that exists with currently available non-patient-specific geometrical models. The researchers demonstrated that the dosimetric results obtained in the study are accurate. However, an intensive study involving a larger cohort of patients is suggested due to the differences between this study and the current clinical practices. This will ensure the full potential of the MRI in uveal melanoma proton therapy as well as in generic tumor control. New approaches for defining organs-at-risk and tumor margins are also likely to be needed.

Geometrical modeling based on ellipsoidal eye shape is the current treatment planning standard. Patient-specific eye parameters such as naso-temporal, sagittal, anteroposterior, and limbus diameters are used to scale the geometrical models. Fundus photography, CT/MRI, or ocular biometry are used to determine the eye parameters. In a review, Slopsema et al. demonstrated the applicability of 3D CT to accurately model the position of the clips in melanoma patients.

Over the past three decades, many studies have already reported the effectiveness of MRI for the diagnosis, choice of treatment, and radiotherapy planning of uveal melanomas. However, the clinical outcome was questionable due to the low-resolution images compared to the current 3D high-resolution MRI protocols. This study employed 7 Tesla MRI with multi-element receive coils for better image quality. The coil integration reduced the scanning time as well. The proposed 3D MRI-based treatment planning for ocular proton beam therapy is field-strength independent.

When it comes to delineation of the OARs, the major limitation was macular segmentation. This is because the macula is not visible on MRI sequences and therefore, only position and shape assessment was done. Similarly, improving ciliary body delineation was also the subject of studies. That said, none of the current techniques have an effective means to calculate dose distribution within the ocular structures. πDose, a dose calculation engine, was developed exclusively to calculate radiation dose distribution for every clinically possible gazing angle. The researchers displayed the results as a grid map for the global weighted-sum objective function values. In this study, global weighted-sum objective function calculation remained manual; bringing automation into the system would be the next step in optimizing gazing angle calculation.

Usage of 2.5 mm margins in ocular proton therapy planning may cause uncertainties in,

  • Stopping-power ratio, range, and penumbra zone
  • The axial length of eye and tumor volume measurements
  • Patient eye positioning and rotation
  • Clip positioning

A novel 3D MRI-based treatment modality can help mitigate uncertainties during treatment planning, as the uveal melanoma and ocular regions can be more accurately segmented.


The shifting toward MRI for ocular proton therapy will become a valuable tool to compliment traditional approaches to the treatment planning of uveal melanoma patients. It  is also an important step towards a clip-less, painless and most importantly, accurate ocular tumor treatment.



About Author

Similar Articles

Leave a Reply