As opposed to conventional radiography, computed tomography provides 3D information of the anatomical structures and tissues of the patient, at a much better low-contrast performance. However, radiation doses associated with CT are up to 40 times higher compared to conventional X-ray examinations. In addition, the number of scans performed increases every year. Since a small but significant risk of radiation induced malignancies is associated with exposure to ionizing radiation, CT imaging must be optimized to assure the required diagnostic level of image quality (IQ) at the lowest possible dose.

The aim of this PhD work was to assess the performance of CT systems in terms of patient dose and IQ. To this end, Monte Carlo simulations were performed with voxelized phantoms, based on clinical CT data. IQ was evaluated by means of an automatic scoring algorithm based on local standard deviations. Using the proposed methods, the behaviour of tube current modulation (TCM) techniques was investigated. In addition, a simplified approach for patient-specific organ dose calculation and risk estimation was developed for use in clinical practice.

To reduce patient radiation dose and to optimize IQ in CT imaging, tube current modulation (TCM) systems were developed. By modulating the tube current along the length axis of the scan, a constant IQ level can be obtained throughout the patient. The tube current is adapted based on a projection radiograph, taken prior to the CT acquisition in either the anterior–posterior (AP), posterior–anterior (PA) or lateral (LAT) direction. As a result, exposure values will be lowered in less attenuating or smaller anatomical regions. In this work, the influence of the localizer type and the scan direction on the dose reducing efficacy of TCM systems was investigated. Chest CT scans based on AP, PA, LAT or dual AP/LAT localizers were acquired on systems from 3 different vendors. In addition the scan direction was changed for one particular system. Organ doses were calculated with Monte Carlo simulations and validated in an anthropomorphic phantom. Thyroid and lung doses increased with 60% for a PA- instead of dual AP/LAT-based scan, with significant differences in image noise. In addition, the thyroid dose halves by taking the scan in the caudocranial (feet first) direction. IQ was not significantly different when changing the scan direction. Our results demonstrate a strong need for detailed analysis of the TCM system performance during commissioning of CT scanners.

The traditional physical-technical image quality parameters are measured in dedicated technical phantoms which are not representative to the patient’s anatomy. Clinical IQ has to be determined with extensive human observer studies. An automatic algorithm based on local standard deviations in the image was evaluated for use in clinical practice. Chest CT’s of Thiel embalmed cadavers were made at different exposure levels. Soft and sharp datasets were acquired with filtered back projection and iterative reconstruction. A visual grading analysis study was set up to validate the outcome of the automated algorithm. A significant correlation was found between the observed clinical IQ and the proposed scoring method (ρ = 0.91, p < 0.001). The automatic scoring algorithm is a promising tool for the evaluation of thoracic CT scans in daily clinical practice.

The absorbed dose to the female breast in thoracic CT imaging is up to 10 times higher than in screening mammography. Since the breasts are rarely the object of interest in chest scans, concern is raised about the elevated population risk of breast cancer incidences from CT. We evaluated the performance and dose reducing potential of organ-based TCM (OBTCM), a technique developed to lower the exposure to anterior located organs. The position of the breasts with respect to the reduced tube current zone was determined. Monte Carlo simulations of standard and OBTCM based scans were performed with clinical CT data of 17 female patients. Individual organ doses and risks were compared between both scan techniques. The potential benefit of OBTCM to the female breast in chest CT is overestimated as the reduced tube-current zone is too limited. Despite a 9% reduction of the breast dose, posterior organs will absorb to 26% more dose, resulting in no additional benefit for reduction of radiation induced malignancies.

Performing Monte Carlo simulations is not possible in a clinical setting. Therefore, in this work a simplified method was developed to estimate patient-specific organ doses and lifetime attributable risks (LAR) resulting from CT torso examinations. Individualized voxel models were created from full body CT data of 10 paediatric patients. Patient-specific organ doses and LAR of cancer incidence and mortality were calculated by means of Monte Carlo simulations. Results were compared to the size-specific dose estimate (SSDE). The latter showed significant strong correlations with organ dose (r > 0.8) and LAR (r > 0.9). Consequently, this dose metric can be used to estimate patient-specific organ doses and risks by taking into account a linear correction factor. The SSDE method makes an on-the-spot dose and LAR estimation possible in routine clinical practice.

The correct use of tube current modulation systems can provide a significant dose reduction in CT imaging. However, positioning the patient in the isocenter of the gantry is extremely important. Miscentering causes an incorrect functioning of the modulation technique as the shadow on the scanned projection radiograph will be too small or too large. In addition, SSDE calculations will be over- or underestimated. Performing Monte Carlo simulations with voxelized phantoms based on clinical CT data of the patient provides an accurate estimate of the delivered dose to in-beam organs. However, the dose distribution outside the field of view is unknown so that conclusions about organs on the periphery of the model should be taken with care.

This PhD work emphasizes the need for patient-specific dosimetry and image quality assessment in computed tomography. The patient’s anatomy, including the relative positions of the different organs, needs to be taken into account when evaluating the performance of different systems. Preferably, individualized organ doses should be determined in a Monte Carlo environment using the patient’s CT dataset. However, as this is not possible in a clinical setting, the SSDE method established in this thesis provides a good approximation. In addition, the proposed IQ scoring algorithm has proven to be a valuable alternative for clinical image quality assessment. The method could allow for IQ monitoring over time without the need for human intervention.