Feasibility of a user-independent systematic method for medical image quality assessment
Commonly used image quality assessment methods are influenced by the user manual selection, or the algorithm automatic selection, of a Region of Interest (ROI). An ROI is used in measurements of the signal mean, the signal standard deviation, and a signal representing the system response such as an Edge Spread Function (ESF). Noise, artefacts and the user manual selection of a ROI size affect the interpretation of a system’s response. Challenges in image quality assessment are more prominent when dealing with poor SNR imaging systems. The aim of this thesis is to develop and evaluate the feasibility of a medical image quality assessment. The method simultaneously estimates image degradation factors, applies them to the known object in the image, and compares the output with the test image. The comparison is executed in the histogram space, removing the spatial dependence present in ROI-based methods. The proposed method was tested using Monte Carlo simulated planar images of a simple disk phantom. The limitations of the proposed method are discussed. The best test results were achieved at SNR values of at least 11.0 (+0.3, -0.2) with an average error in signal and noise measurements of no more than 0.1 (+0.1, -0.1) % and an average error in the measurement of the frequency at 10% Modulation Transfer Function (MTF) of 0.1 (+0.2, -0.1) cycle/mm. At SNR values less than ~ 10, conventional methods of image quality assessments are expected to be superior to the proposed method.
Image quality, Modulation transfer function (MTF), Signal to noise ratio (SNR), Resolution, Contrast, Digital radiography, Cylinder edge