Premium
SU‐E‐T‐565: Multiple Channel Radiochromic Film Calibration with Spatial Dose Gradient Regularization
Author(s) -
Li Darrell T H,
Lee Louis K Y,
Chan Anthony T C
Publication year - 2015
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.4924927
Subject(s) - regularization (linguistics) , mathematics , algorithm , calibration curve , calibration , optics , mathematical optimization , physics , computer science , statistics , artificial intelligence , detection limit
Purpose: To develop a new algorithm for multiple channel radiochromic film calibration and to evaluate the accuracy of the algorithm with virtual simulated film images. Methods: The proposed calibration method was based on least‐square optimization of discrepancies between measured and estimated optical density of the three colour channels, with the inclusion of the weighted dose gradient as penalty to this minimization problem. The regularization parameter λ was selected using the L‐curve method to reduce the L2‐norm of the dose gradient while preserving the data consistency. The situation of λ = 0 represented no regularization and the least‐square solution would be obtained. The algorithm was tested with 4 virtual film images which were produced from dose planes exported from the planning system. Zero‐mean Gaussian noise were added to simulate the real situation. The resultant dose maps calculated by both least‐square method and regularization method were then compared with the original dose planes. Results: Comprehensive study of the 4 simulated cases with the L‐curve method yielded an optimal regularization parameters of 2×10 −8 . The dose planes generated by least‐square method showed significant noise retained when compared with that from the regularization method. In performing Gamma Index analysis with tolerance of 1% and 1 mm, it was observed that in the least‐square solution, failed points exist everywhere on the dose plane, while only a small amount in the sharp dose gradient region of the regularized solution. It was also noted that the number of failed point in the least‐square solution can be 20 to 150 times more than that in the regularized solution. Conclusion: The calibration algorithm proposed in this study successfully recovered the original dose plane from the noise contaminated film image with the regularization method. Regularization should be a better option to the least‐square one in solving the radiochromic film calibration problem.