TH‐E‐BRA‐05: Improving the Contrast of Proton and Carbon Radiography by Using CT Prior Knowledge

Purpose: To enhance the quality of proton/carbon radiography for real time tumor tracking using CT prior knowledge. Methods: A method for contrast enhancement was applied to particle radiographies (230 MeV proton, 330 MeV proton, 500MeV/n Carbon) virtually generated by means of Monte Carlo simulations. CT data volumes from 6 lung patients, with different lesion size, were processed and analyzed. The tumor region (generally the PTV) and low density voxels were segmented out to generate a new processed CT. A Digital Reconstructed Radiography (DRR) was computed from this processed CT volume. After equalizing the histogram between 0 and 1 of both particle radiography and DRR, the two images were subtracted thus obtaining a high signal around the tumor region. Contrast to Noise Ratio (CNR) was used as a metric to measure the enhancement. A normalized cross‐correlation based algorithm was implemented to automatically detect the GTV area. The estimated center of mass (CM) of the GTV was compared to projection contours drawn by physicians. Results: CNR figures showed improvement up to 6 times when comparing the enhanced contrast image vs. the original particle radiography. The 2D distance between the real and the automatically estimated CM of the GTV was 2,12±0,62 mm (median±quartile). Conclusions: The advantage of using proton or carbon radiography to detect soft tissue during patient set up and radiation delivery can be further on improved by using prior knowledge derived from the planning CT. The method we propose is able to significantly enhance the contrast of the tumor region with acceptable computational time for real time applications. Further analysis is required to study the benefit of such a methodology to track the lesion over time during treatment. The authors declare that no conflicts of interest exist. © 2012, American Association of Physicists in Medicine. All rights reserved.