Quality control in diagnostic imaging using images of patient studies

Abstract

Introduction Weekly lean quality checks (LQC) of imaging devices using simple phantoms are commonly performed in many hospitals, aiming at early detection of changes in image quality or radiation dose. Purpose We propose to monitor the constancy of imaging devices by deriving quality parameters from images of routine clinical studies. Materials and methods Chest images of patients were collected from a DR system (Oldelft). As an indicator of image quality, an estimation of noise was derived by application of a spatial high-pass filter to patient images and normalization to the local signal intensity in the original images. As an indicator of dose constancy, the DAP was extracted from DICOM meta-information (LQC) and used to calculate the tube output [mGy/mAs] (clinical images). Results Preliminary results using LQC and clinical images collected during 10 months are given in relative peak-to-peak (PtP) and standard deviation (stdev). For noise levels, LQC resulted in PtP 22% and stdev 5.2%, clinical images resulted in PtP 36% and stdev 6.6%. For dose, LQC resulted in PtP 33% and stdev 8.5%, clinical images resulted in PtP 30% and stdev 7.5%. Conclusion The similar results for LQC and clinical images show it is feasible to perform regular quality checks of diagnostic imaging devices using patient images. To validate the potential of this approach to detect changes, low quality images will be analysed with the same algorithms. Additionally, the methodology will be extended to other anatomies and imaging modalities. This will enable performance monitoring of all diagnostic imaging devices, without the need for weekly phantom constancy tests. Disclosure Nothing to disclose.