ESTRO 38 Abstract book

S1162 ESTRO 38

environment of the ExacTrac v3.5 system (Brainlab, München, Gernmany). The density of the lesion was used as a surrogate for tracking. The accuracy of the target identification for a reference image pair, which is provided to the markerless tracking algorithm, was compared against the marker-based implants detection. The result gives an estimation of the manual target identification error. Additionally, the user-variability and the patient-variability of the manual target identification error were analyzed. Furthermore, 15 patients treated with internal target volume (ITV) were simulated for markerless tracking. Results Root mean square error between the markerless tracking and the marker-based tracking for all patients in LR, AP and SI directions were [01.1, 0.3] mm, [0.8, 0.9] mm and [1.2, 0.9] mm, respectively. This markerless tracking algorithm was able to track 87 % of all images. In the case of simulating real time patients treated with an ITV approach, 80% were eligible for markerless tracking. Conclusion In this study, we successfully implemented a markerless environment for dynamic tracking leading to a patient specific dose delivery. EP-2102 Accurate software detection of light markers coincidence using a computed radiography system M.A. Benito Bejarano 1 , A. Del Castillo Belmonte 2 1 Hospital Provincial de Zamora, Department of Radiation Physics, Zamora, Spain ; 2 Hospital Clínico Universitario, Department of Radiation Physics, Valladolid, Spain Purpose or Objective The current recommendations on quality assurance (QA) of medical accelerators set a monthly frequency for testing the coincidence of the light field crosshair projection and patient positioning lasers. Nowadays, the standard way of performing this test is by naked eye examination of the projection of those markers over graph paper. However, the accuracy of this method is compromised by interobserver variability. We propose a new method that allows a precision gain by means of a Computed Radiography (CR) system and an in-house developed analysis software. Material and Methods A Konica Regius model 140 CR System is used. The system has been previously tested to determine and correct geometric distortions. The CR cassette is irradiated with 16 MU in a 40x40 cm2 6MV field at isocenter distance to obtain a latent uniform background image. In our procedure, we use the sensitivity of the CR plate to visible light to acquire an impression of the optical indicators. The uncovered phosphor is directly exposed to the laser light for 20 s. The collimator is set to several angles, switching the light field on for 20 s in each one. The CR plate is put back inside the cassette and is read by the CR system. The result consists of a superposition of the image from the laser cross plus the image star from the different projections at each collimator position. An in-house software routine has been developed to analyse the images. It imports the image in RAW format and finds the laser crossing point. The lines of the star-shot pattern of the light crosshair are detected, and the minimum circle encompassing all the lines is determined. The software returns the radius as well as the distance of the circle centre to the laser cross point. The accuracy of this procedure has been tested by running it against synthetic images with intentional known shifts applied. Results Geometric distortions were found negligible, no image correction was needed. The coincidence of lasers with cross-hairs could be determined with an accuracy better than 0.2 mm with a 95% confidence interval. The image acquisition process takes less than 15 minutes, image processing less than 2 minutes.

Figure1: Input image with laser marks in white, and crosshair star pattern in black

Figure2: Analysis results. Zoom of the area near the detected laser cross mark, showing the minimum encompassing circle and its centre. Conclusion This method allows the measurement of light crosshair and laser lines coincidence. The accuracy is improved, as it eliminates the inter-observer dependence. The resulting digital images can be archived after processing, for future reference. The materials involved are readily available in a modern hospital setting. The test is inexpensive as no consumables like film are used. This method has been proven useful when performing the light field crosshair to lasers coincidence test, mainly after a field service intervention in the linac, i.e. mirror or crosshair substitution.