A comparative study evaluating the performance of diagnostic radiography units and protocols for paediatric and adult chest radiography examinations

Purpose: Little is known about the variations in image quality (IQ) and radiation dose for paediatric and adult chest radiography (CXR), between and within hospitals. Large variations in IQ could influence the diagnostic accuracy, and variations in radiation dose could affect the risk to patients. This thesis aims to develop, validate and then use a novel method for comparing IQ and radiation dose for paediatric and adult CXR imaging examinations and report variation between a series of public hospitals. Method: A Figure of Merit (FOM) concept was used for the purposes of comparing IQ and radiation dose, between and within hospitals. Low contrast detail (LCD) detectability, using the CDRAD 2.0 phantom, was utilised as the main method for IQ evaluation. The validity of utilising LCD detectability, using CDRAD 2.0 phantom, for evaluating visual IQ, simulated lesion visibility (LV) and CXR optimisation studies, was investigated. This was done by determining the correlation between the LCD detectability and visual measures of IQ and LV for two lesions with different locations and visibility in the Lungman chest phantom.The CDRAD 2.0 phantom and two anthropomorphic phantoms (adult Lungman and the neonatal Gammex phantom) were used to simulate the chest region. Radiographic acquisitions were conducted on 17 X-ray units located in eight United Kingdom (UK) public hospitals within the North-west of England using their existing CXR protocols. The CDRAD 2.0 phantom was combined with different thicknesses of Polymethyl methacrylate (PMMA) slabs to simulate the chest regions of 5 different age groups: neonate, 1, 5, 10 years and adults. A Lungman phantom, with and without the fat jacket, was used to simulate average and larger sized patients. IQ was evaluated using a number of methods, including: 1) physically, by calculating LCD detectability as represented by an image quality figure inverse (IQFinv) using the CDRAD analyser software; 2) using images acquired from the anthropomorphic phantoms – for this, a relative visual grading analysis (VGA) method was used. Additionally, signal to noise ratios (SNR), contrast to noise ratios (CNR) and conspicuity indices (CI) were calculated for all phantom image data in this study. Incident air karma (IAK) was measured using a solid-state dosimeter. Results: Regarding the validation of utilising LCD detectability for evaluating visual IQ and LV, and CXR optimisation studies, a strong positive correlation (r = 0.91; p < 0.001) was observed between IQFinv and the visual IQ scores from the Lungman phantom. A good correlation was observed between IQFinv and visual LV from the Lungman phantom for both lesions (lesion 1 (with low visibility) (r = 0.79; p < 0.001); lesion 2 (with high visibility) (r =0.68; p < 0.001), respectively). Considerable variation in standard imaging protocols/techniques, radiation dose, IQ and FOM were observed between the hospitals, while within hospital variation was lower. A weak correlation between IQ and radiation dose was observed across most of the age groups studied. Conclusion: A novel method has been established to evaluate and compare IQ and radiation dose between and within hospitals based on an FOM concept. This combines IQ and radiation dose into a single factor and is the first of its kind to reported within the field of medical imaging. It can be confirmed that LCD detectability using the CDRAD 2.0 phantom is valid for evaluating visual IQ and LV and can be of use within routine quality assurance and optimisation studies in digital radiography. Further radiation dose optimisation for the paediatric age groups and adult group, especially in hospitals /X-ray machines with low IQ and high IAK, are required