Fiducial markers and daily kV imaging improve patient setup during proton RT of esophageal cancer

Purpose or Objectives
The success of light ion radiotherapy is crucially sensitive to daily setup variation and anatomic changes. Irradiation of the esophagus, in particular, suffers from significant target mobility [1].

Since its commissioning, the proton facility at our institute has been using daily kV imaging for patient setup based on bony anatomy, supplemented with in-room dual-energy CT scans as needed.

We assessed target positioning with the aid of kV-visible fiducial markers placed around esophageal tumors by tracking their displacement from the planned location throughout treatment.

Materials and Methods
Five patients with esophageal cancer scheduled for neoadjuvant (4) or definite (1) radiochemotherapy using protons (all male; age 52-64 years; tumor of the middle or lower third of the esophagus; mean GTV: 24 ml) underwent endoscopic ultrasound-guided trans-esophageal placement of gold fiducial markers (VisiCoilTM, IBA Dosimetry, Belgium; diameter 0.35 mm, lengths 5 or 10 mm) demarcating the proximal and distal tumor borders.
Pairs of orthogonal kV images taken in treatment position were retrieved for 104/110 fractions. Markers were manually located in the images and their position in 3D determined by triangulation [2]. Vertebral landmarks were similarly reconstructed and used to register kV imaging to the planning CTs. Triangulation errors were estimated per marker from the geometrical compatibility of image point pairs with single 3D points. All calculations were carried out in custom software based on SciPy [3].
 
Results
Marker implantation proceeded without complications and placement of the distal marker was hindered by an obstructing tumor in one case only. All but one marker remained stable throughout treatment, and their visibility in kV images was generally limited, but sufficient. Marker-induced dose perturbations were not clinically relevant.
Figure 1 shows example triangulation results for one patient. Overall the standard deviation [inter-quartile range] of the distributions of marker excursions was 2.6 [2.1], 1.8 [2.0], and 3.1 [3.6] mm in the lateral, sagittal, and longitudinal directions, respectively. A Friedman test failed to reveal significant differences between the directions (p>0.7). Inter-marker distances varied by a few mm during treatment with no discernible trend. Triangulation errors ranged from 0.9 to 7.1 mm.
One notable instance of marker-indicated target displacement by 26 mm was observed (Fig. 2). This prompted the acquisition of a control CT in treatment position, which revealed an overlap of the planned and actual GTVs of merely 4% (Jaccard index 0.02). The dosimetric impact would have been a reduction of the mean and near-min (D98%) doses by 6% and 47%, respectively.

Conclusion
Exclusive reliance on bony anatomy for positioning of esophageal cancer patients cannot prevent occasional target misalignments. Gold fiducials (e.g., VisiCoilTM) and daily kV imaging are valuable tools for target-centric setup verification.