OCTOPUS – OCular TumOur Planning UtilitieS


OCTOPUS is special extension of our virtual radiotherapy simulator and planning system VIRTUOS for proton therapy of ocular tumours. OCTOPUS was developed in a national research project together with the Hahn-Meitner-Institute in Berlin. At the Hahn-Meitner-Institute in Berlin the first proton therapy facility in Germany was established in August 1998.

Proton therapy of ocular tumours

Radiotherapy of eye tumours with proton beams is a successful treatment method established in more than twenty medical proton facilities throughout the world. So far the therapy planning is done with the program EYEPLAN [1] that now has been used for more than 15 years. Some restrictions of Eyeplan (e.g.: a simple spherical eye model, no support of modern three-dimensional diagnostic input) lead to planning uncertainties which must be compensated by larger security margins. This may happen in about 20% of the treated cases as experiences of radiotherapists at the Hahn-Meitner-Institute (HMI) in Berlin has shown. This may increase side effects as radiosensitive structures near the target volume suffer from high doses. and can lead to severe impairment or even loss of sight.

Treatment facility at the HMI in Berlin
© dkfz.de

OCTOPUS (OCular Tumour Planning UitilitieS) was designed to overcome these uncertainties. Because of the exceptional anatomy of the eye some extra difficulties have to be considered in contrast to the standard planning process for conventional radiotherapy:

  • Distances between target volume and structures at risk in the eye are in the order of 0.1 mm rather than mm. A higher precision in treatment planning, positioning and verification is required.
  • The eyeball can move freely. An active or passive fixation method is necessary to reposition the patient respectively his eye exactly before applying the single fractions. Normally a passive fixation is preferred. The patient is trained to fix a light during therapy. To control patient setup, tantal clips were sutured on the eye before radiation therapy. The clips can be clearly distinguished in X-ray verification images, which are acquired before each fraction. By comparing the current marker positions with that one calculated during the planning procedure it is possible to reposition the patient very precisely.
  • Additional diagnostic input is available, as the tumour can be seen through the lens non invasively (Fundus Photos).

A detailed description of the standard treatment procedure can be found in [2].

schematic sketch of the treatment process.
© dkfz.de

To support the usual planning process the program has to provide the maximum depth and width of the SOBP (spread out bragg peak), the form of the collimator, the position of the fixation light, parameters for the wedges used and two artificial orthogonal X-ray images with the clip positions for iterative positioning of the tumour (see figure 1). The typical planning process consists of five independent steps:

  • Modelling the patients eye
  • Choose of treatment parameters: (i) Position of fixation light = gazing angle (ii) Lateral, distal and proximal safety margins (iii) Wedges (orientation, inclination, distance to isocenter) (iv) Estimation of patient contour and eyelids
  • Dose calculation with precise algorithm
  • Evaluation of plan and iterative adaption à step 2.

Modelling of the eye

Figure 2: Top: Adaptation of eye model to CT images (transversal, sagittal and frontal view) by rotation, translation and scaling. Centre: Cross check of measured clip positions with positions of clip artefacts in CT (left, cnetre) and positions marked in the fundus view photo. Bottom: Input of tumour shape based on fundus photo information. White clouds are projections of clip artefacts in the CT data set.
© dkfz.de

Most eyes can be modelled by a simple geometrical model without problems. To make the eye model as flexible as possible the modelling of the patient’s eye geometry is done in three steps:

  • Creation of an elliptical eye model based on different diagnostic measurements (eye length and width, cornea radius, distance lens-cornea etc.).
  • Interactive fit of the model to CT images or other three-dimensional image information (MRI, US). In this procedure the whole eye or only some selected sub-structures can be transformed by translation, rotation and scaling.
  • At last all structures may be adapted individually to fit even very deformed eyes that differ from elliptic geometry. If no CT cube is available an artificial CT cube is created filled with published mean densities of each eye structure.

After the adaptation of the eye model clips and tumour are added and a consistency check with diagnostic images is performed.

Treatment planning

Choosing gazing angle in the Beam’s Eye View.
© dkfz.de

Even if OCTOPUS support the use of three-dimensional image sequences, the way of considering that information is quite different compared to conventional radiotherapy. In conventional radiotherapy imaging is done carefully in the estimated treatment position, to be able to transfer planning decisions made on the acquired image information correctly into the treatment situation.

In OCTOPUS CT information is used to model the anatomy correctly but the position and orientation of the patient, respectively of his eye, during imaging is quite different to that one during therapy. There are two major reasons for that difference. The beam line is fixed, that means it cannot be rotated around the patient. Even if the patient could be rotated relatively to the beam line, this is done only in a very restricted range, since it should be avoided to irradiate through bony structures. Due to their high density they would restrict the maximal depth of the SOBP, and, depending on the available beam energy, it might be impossible to be able to cover the target volume completely with the necessary dose.

As result the irradiation direction is determined by selecting a gazing angle which will assure that the target volume can be enclosed completely by the fixed beam. During therapy the patient is asked to gaze at a fixation light to realize the correct direction. That means the eye might be in a completely different position and orientation (relatively to the rest of the patient’s anatomy) compared to that one during image acquisition.

There is often another modification of the treatment situation compared to image acquisition. Eye lids and lashes normally show a very short-term response to irradiation with displeasing side effects. Therefore they are often removed out of the beam by applying some retractors. In summary this geometrical deviations doesn’t allow a precise dose calculation based on the pre-therapeutically acquired CT images. To consider these deviations the shape of the lid and the orientation of the patient’s surface relatively to the eye in treatment position must be estimated when a plan is created. As basis for the precise pencil beam algorithm we create therefore an artificial CT cube, where the various eye structures are filled with mean density values retrieved from several publications. This way we can overcome some other uncertainties introduced into the CT images by clip artefacts, too. Of course, while our elliptical eye model these estimations still limit the overall precision

Aim of the treatment planning process is to find an optimal set of treatment parameters: gazing angle (i.e. irradiation direction), safety margins and wedges. The resulting treatment parameters can be visualized in the Observer’s View (figure 3).

More efficient planning with real time dose calculation

Checking treatment parameters in the Observer’s View
© dkfz.de

To make the planning process more efficient and ergonomic treatment planning steps (2) and (4) were fused by means of a real time dose calculation that allows a direct evaluation of the current plan while changing its parameters. This dose distribution is an approximate estimation of the deposed dose in the eye for the given treatment parameters and allows immediate rejection of sub-optimal parameter sets. The dose calculation is fully integrated into the graphical user interface of OCTOPUS. While the user selects therapy parameters as e.g. gazing angles in the Beam’s Eye View (fig. 3, left) the dose distribution is calculated and visualized in the eye model displayed in the Observer’s View (fig. 4, left) or on CT slices. To speed up the calculation, several simplifications are made:

  • The density in the eye was assumed to be homogeneous. This is true within a few percent.
  • Only one beam is considered. Complex geometrical beam configurations need not to be calculated. This is no limitation for proton therapy of ocular tumours, as usually only one beam is used.
  • Calculation is performed on the eye model (»104 triangles) and not on the CT cube (» 106 voxels).
  • The dose is not represented in a dose cube but by triangulated surfaces of isodose volumes. This restriction to a few fixed dose intervals limits the amount of data and is sufficient for a first guess.

checking treatment parameters in the Observer’s View
© dkfz.de

With an error of less than ±7% this algorithm shows all topological information of the dose distribution. It must be emphasised, that the intention is not to give the correct answer, but a first impression of the qualitative dose distribution in relation to the eye structures.

Due to the described acceptable simplifications OCTOPUS is the only one 3D planning system in radiotherapy which offers a real-time dose response to parameter changes.

Once the dose calculation has modelled the isodose surfaces, an estimation of the dose for any point in space can be given by determining that isodose volumes which enclose this point. This way the dose distribution can be displayed on an artificial Fundus View or on the surface of any structures in the three-dimensional eye model (fig. 4).

To achieve an accurate dose distribution and thus evaluate the selected parameters as precise as possible a pencil beam algorithm was developed at the HMI Berlin that considers the properties of the proton facility in detail [4]. It takes into account the inhomogeneous density in the eye and describes scattering accurately. Planning should always be finished by using that precise algorithm.

In result mode all features available in the standard VIRTUOS mode for qualitative and quantitative evaluation of dose distributions are available, too. Of course, this applies also to all features for comparing dose distributions of concurrent plans.


The project was funded by the DFG (Deutsche Forschungsgemeinschaft) project no. Schl 249/4-1 – Schl 249/4-4 and FU 50/3-1 – FU 50/3-4.

Currently OCTOPUS is tested and evaluated against EYEPLAN at our partner sites in Berlin. Based on these experiences the overall handling is continuously modified and adjusted to the needs of routine treatment planning. If all tests are finished, it is planned at HMI Berlin to replace EYEPLAN completely by OCTOPUS.

We hope we can continue the fruitful joint developments in the future Programme-oriented Funding of the Helmholtz Association. A challenging task for future work would be for example the extension of the real-time dose calculation to multiple beams and to extend the concept of real-time calculations to other radiation types.


[1] Goitein M, Miller T (1983) Planning proton therapy of the eye. Med. Phys. 10: 275 - 283

[2] Alberti WE, Sagermann RH (eds.) (1993) Radiotherapy of Intraocular and orbital tumours. Springer Verlag

[3] Chauvel P et al. (1997) Clinical and Technical Require­ments for Proton Traetment Planning of Ocular Diseases. in T. Wiegel (Hrsg) Radiotherapy of Ocular Disease; Front. Radiat. Ther. Oncol. 30: 133

[4] Paganetti H (1998) Monte Carlo method to study the proton fluence for treatment planning, Med. Phys. 25: 2370

[5] Pfeiffer K (1999) Implementierung eines Echtzeitdosisberechnungsalgorithmus für die Protonentherapieplanung bei Augentumouren. PhD Thesis, University Heidelberg

[6] Dobler B (2000) Präzise Modellierung von Risikoorganen und Zielvolumina für die Protonentherapie intraocularer Tumoren. PhD Thesis, University Heidelberg

[7] Pfeiffer K, Bendl R (2001) Real-time dose calculation and visualization for the proton therapy of ocular tumors. Phys. Med. Biol. 46: 671 - 686

[8] Dobler B, Bendl R (2002) Precise modelling of the eye for proton therapy of intra-ocular tumours. Phys. Med. Biol. 47: 593 - 613

to top