Purpose: As EPIDs are increasingly used for IMRT QA and real‐time treatment verification, comprehensive quality assurance (QA) of EPIDs becomes critical. Current QA with phantoms such as the Las Vegas and PIPSpro™ can fail in the early detection of EPID artifacts. Beyond image quality assessment, we propose a quantitative methodology using local noise power spectrum (NPS) to characterize image noise and wavelet transform to identify bad pixels and inter‐subpanel flat‐fielding artifacts. Methods: A total of 93 image sets including bar‐pattern images and open exposure images were collected from four iViewGT a‐Si EPID systems over three years. Quantitative metrics such as modulation transform function (MTF), NPS and detective quantum efficiency (DQE) were computed for each image set. Local 2D NPS was calculated for each subpanel. A 1D NPS was obtained by radial averaging the 2D NPS and fitted to a power‐law function. R‐square and slope of the linear regression analysis were used for panel performance assessment. Haar wavelet transformation was employed to identify pixel defects and non‐uniform gain correction across subpanels. Results: Overall image quality was assessed with DQE based on empirically derived area under curve (AUC) thresholds. Using linear regression analysis of 1D NPS, panels with acceptable flat fielding were indicated by r‐square between 0.8 and 1, and slopes of −0.4 to −0.7. However, for panels requiring flat fielding recalibration, r‐square values less than 0.8 and slopes from +0.2 to −0.4 were observed. The wavelet transform successfully identified pixel defects and inter‐subpanel flat fielding artifacts. Standard QA with the Las Vegas and PIPSpro phantoms failed to detect these artifacts. Conclusion: The proposed QA methodology is promising for the early detection of imaging and dosimetric artifacts of EPIDs. Local NPS can accurately characterize the noise level within each subpanel, while the wavelet transforms can detect bad pixels and inter‐subpanel flat fielding artifacts.

No abstract

The key of wavelet image threshold de-noising is the choice of the threshold function and the threshold value. To overcome the shortcomings of constant deviation existing between estimated wavelet coefficients and decomposition coefficients in the soft threshold function and discontinuity of the hard threshold function, a new threshold function based on wavelet shrinkage in image de-noising is presented in this paper. Threshold values of images with different edges and texture degrees are fine-tuned when the threshold value is set. Furthermore, a self-adaption optimal threshold which is fit to all scale levels is designed based on features of multiscale and multiresolution of wavelet transform. Simulation results show that the proposed methods are efficient to reduce the noise while preserving the detail information of the image.