SU-E-I-48: Noise Reduction with Over-Sampling for High Resolution Detectors Using a Spread Function Convolution Method

Abstract

Purpose: A method to reduce noise resulting from the use of higher resolution x‐ray detectors being developed to meet the demands of image guided vascular interventions is demonstrated. Methods: New direct detectors based on amorphous Se can have MTFs that remain high even at their Nyquist frequency. Since such detectors can be made with smaller pixels than may be required for even the high resolution requirements of many neurovascular applications, the resulting over‐sampled images can be convolved with various functions to lower the noise. The effect on resolution can then be compared with a simple pixel binning. The general method proposed by Cunningham et al as the apodized‐aperture pixel (AAP) design was compared with the Result for various simple 1D convolution spread functions with the effects on total noise and MTF compared to that of the binning case for a 25 μm aSe CMOS detector's MTF published by the University of Waterloo group lead by KS Karim. Results: Assuming a white noise image, various convolution kernels resulted in similar reductions of total standard deviation. Detailed comparisons were made with the simple the 2x binning case. While reducing noise, the over‐sampling convolution method using simple convolution low pass filters did not show advantage over 2x binning with regard to modifying the MTF; however, significant improvement was evidenced for the more complex sinc function used in the AAP design. Conclusion: As higher resolution detectors are being developed to meet the increasing demands for improved images to guide finer vascular interventions, use of super‐sampled aSe detectors where resolution is maintained with reduced noise may well fill this requirement. Partial support from NIH Grant R01EB002873 and Toshiba Medical Systems Equipment Grant

Comparison of high resolution x-ray detectors with conventional FPDs using experimental MTFs and apodized aperture pixel design for reduced aliasing

“…The Apodized Aperture Pixel design is a novel method to reduce signal aliasing and maintaining high spatial resolution up to the Nyquist frequency as shown here and previous works 3,7 . Compared to standard pixel binning kernels, the AAP design retains MTF values at higher frequencies without aliasing due to undersampling.…”

“…A single panel solution would be optimal for clinical practice and this detector should have smaller pixels for high resolution ROI needs. Extending this to a Flat Panel Detector using moving average pixel binning of smaller pixels, low pass/high pass filters or other image processing methods, would reduce the MTF near the Nyquist frequency 3 ; however, with resulting aliasing as a consequence. Apodized Aperture Pixel (AAP) design 1,2 provides an alternative means to maintain the high resolution over an ROI while expanding the FOV with larger effective pixels yet with minimized aliasing.…”

“…al, is an alternative to the conventional pixel design 1 . The advantages of AAP processing with a sinc filter in comparison with using other filters include non-degradation of MTF values and elimination of signal and noise aliasing, resulting in an increased performance at higher frequencies, approaching the Nyquist frequency 3 . If high resolution small field-of-view (FOV) detectors with small pixels used during critical stages of Endovascular Image Guided Interventions (EIGIs) could also be extended to cover a full field-of-view typical of flat panel detectors (FPDs) and made to have larger effective pixels, then methods must be used to preserve the MTF over the frequency range up to the Nyquist frequency of the FPD while minimizing aliasing.…”