Image quality in CT: From physical measurements to model observers.

Evaluation of image quality (IQ) in Computed Tomography (CT) is important to ensure that diagnostic questions are correctly answered, whilst keeping radiation dose to the patient as low as is reasonably possible. The assessment of individual aspects of IQ is already a key component of routine quality control of medical x-ray devices. These values together with standard dose indicators can be used to give rise to 'figures of merit' (FOM) to characterise the dose efficiency of the CT scanners operating in certain modes. The demand for clinically relevant IQ characterisation has naturally increased with the development of CT technology (detectors efficiency, image reconstruction and processing), resulting in the adaptation and evolution of assessment methods. The purpose of this review is to present the spectrum of various methods that have been used to characterise image quality in CT: from objective measurements of physical parameters to clinically task-based approaches (i.e. model observer (MO) approach) including pure human observer approach. When combined together with a dose indicator, a generalised dose efficiency index can be explored in a framework of system and patient dose optimisation. We will focus on the IQ methodologies that are required for dealing with standard reconstruction, but also for iterative reconstruction algorithms. With this concept the previously used FOM will be presented with a proposal to update them in order to make them relevant and up to date with technological progress. The MO that objectively assesses IQ for clinically relevant tasks represents the most promising method in terms of radiologist sensitivity performance and therefore of most relevance in the clinical environment.

Assessment of radiation exposure in dental cone-beam computerized tomography with the use of metal-oxide semiconductor field-effect transistor (MOSFET) dosimeters and Monte Carlo simulations.

OBJECTIVES: The aims of this study were to assess the organ and effective dose (International Commission on Radiological Protection (ICRP) 103) resulting from dental cone-beam computerized tomography (CBCT) imaging using a novel metal-oxide semiconductor field-effect transistor (MOSFET) dosimeter device, and to assess the reliability of the MOSFET measurements by comparing the results with Monte Carlo PCXMC simulations. STUDY DESIGN: Organ dose measurements were performed using 20 MOSFET dosimeters that were embedded in the 8 most radiosensitive organs in the maxillofacial and neck area. The dose-area product (DAP) values attained from CBCT scans were used for PCXMC simulations. The acquired MOSFET doses were then compared with the Monte Carlo simulations. RESULTS: The effective dose measurements using MOSFET dosimeters yielded, using 0.5-cm steps, a value of 153 µSv and the PCXMC simulations resulted in a value of 136 µSv. CONCLUSIONS: The MOSFET dosimeters placed in a head phantom gave results similar to Monte Carlo simulations. Minor vertical changes in the positioning of the phantom had a substantial affect on the overall effective dose. Therefore, the MOSFET dosimeters constitute a feasible method for dose assessment of CBCT units in the maxillofacial region.

Calibration and features of air-kerma length product meters.

Pencil-type air-kerma length product meters are generally used for quality control and radiation exposure measurements in computed tomography. To ensure reliable results, these meters should be calibrated so that measurements are traceable to international standards. Suitable calibration procedures, together with the properties of these meters, were examined and compared with the international standards and recommendations. The calibration procedure and setup used in this study were slightly modified compared with international recommendations. The special collimator system was found to cause less scatter than similar setups in earlier studies. The energy dependence of the meter response was investigated for several types of meters with standard radiation qualities. With most tested meter types, the total variation due to energy dependence was <4 %, but some had strong energy dependence and the variation was up to 15 % or higher. This highlights the importance of a proper calibration. The response of one semiconductor meter type varied up to 8 % when rotating the meter around its axis; this should be taken into account when making calibrations with a static setup.

Weighting of secondary radiations in organ dose calculations.

The current system of dose quantities in radiological protection is based, in addition to the absorbed dose, on the concepts of equivalent dose and effective dose. This system has been developed mainly with uniform whole-body exposures in mind. Conceptual and practical problems arise when the system is applied to more general exposure situations where the radiation quality is altered within the human body. In this article these problems are discussed, using proton beam radiotherapy as a specific example, and a proposition is made that dose equivalent quantities should be used instead of equivalent doses when organ doses are of interest. The calculations of out-of-field organ doses in proton therapy show that the International Commission on Radiological Protection-prescribed use of the proton weighting factor generally leads to an underestimation of the stochastic risks, while the use of neutron weighting factors in the way as practised in the literature leads to a significant overestimation of these risks.

Effects of radiation quality on the calibration of kerma-area product meters in x-ray beams.

The calibration coefficients of kerma-area product meters significantly depend on the energy spectrum of the x-ray beam. This effect was examined by measuring the calibration coefficients for several radiation qualities in the range generally used in diagnostic x-ray imaging. The intention was to determine the calibration coefficients for other radiation qualities by interpolation between the measured values, relative to one or more suitable parameters. The x-ray tube voltage, total filtration and half-value thickness were examined as possible specifiers of the energy distribution. No single parameter provided an interpolation of calibration coefficients with the accuracy recommended by the ICRU and IAEA, except for a narrow range of radiation qualities. At least two of the parameters are needed to reliably specify the radiation quality for the interpolation of calibration coefficients.

Occupational radiation doses in interventional radiology: simulations.

In interventional radiology, occupational radiation doses can be high. Therefore, many authors have established conversion coefficients from the dose-area product data or from the personal dosemeter reading to the effective dose of the radiologist. These conversion coefficients are studied also in this work, with an emphasis on sensitivity of the results to changes in exposure conditions. Comparison to earlier works indicates that, for the exposure conditions examined in this work, all previous models discussed in this work overestimate the effective dose of the radiologist when a lead apron and a thyroid shield are used. Without the thyroid shield, underestimation may occur with some models.

Review of relationships between physical measurements and user evaluation of image quality.

Some of the findings of a review of the relationship between physical measurements and clinical image quality have been summarised. Mixed results were found: some studies had no relationship at presently typical dose levels, whereas others had a clear correlation between them. It is concluded that the various image quality evaluation tasks in an X-ray department are best done by different methods. Presently, exact physical measurements cannot supersede subjective evaluation in judging the acceptability of clinical images, whereas they are indispensable in specification and testing of technical performance.

Monte Carlo simulations of occupational radiation doses in interventional radiology.

Occupational radiation doses in interventional radiology can potentially be high. Therefore, reliable methods to assess the effective dose are needed. In the present work, the relationship between the personal dose equivalent, H(p)(10), the reading of a personal dosimeter and the effective dose of the radiologist were studied using Monte Carlo simulations. In particular, the protection provided by a lead apron was investigated. Emphasis was placed on sensitivity of the results to changes in irradiation conditions. In our simulations a 0.35 mm thick lead apron and thyroid shield reduced the effective dose, on average, by a factor of 27 (the range of these data was 15-41). Without the thyroid shield the average reduction factor was 15 (range 6-22). The reduction sensitively depended on the projection and the X-ray tube voltage. The dosimeter reading, when the dosimeter was worn above the apron and a thyroid shield was used, overestimated the effective dose on average by a factor of 130 (range 44-258) when the dosimeter was located on the breast closest to the primary X-ray beam. Without the thyroid shield the average overestimation was 69 (range 32-127). If the dosimeter was worn under the apron its reading generally underestimated the effective dose (on average by 20% with the thyroid shield). Our study indicates that, even though large variations are present, the often used conversion coefficient from the dosimeter reading above the apron to the effective dose, around 1/30, generally overestimates the effective dose by a factor of two or more.

Image quality measurements in radiology.

Image quality measurement methods are reviewed and difficulties in various approaches are highlighted. The main emphasis of the paper is on objective image quality measurements, however, subjective assessment methods are also discussed briefly.

A search for improved technique factors in paediatric fluoroscopy.

A Monte Carlo computational model of a fluoroscopic imaging chain was used for deriving optimal technique factors for paediatric fluoroscopy. The optimal technique was defined as the one that minimizes the absorbed dose (or dose rate) in the patient with a constraint of constant image quality. Image quality was assessed for the task of detecting a detail in the image of a patient-simulating phantom, and was expressed in terms of the ideal observer's signal-to-noise ratio (SNR) for static images and in terms of the accumulating rate of the square of SNR for dynamic imaging. The entrance air kerma (or air kerma rate) and the mean absorbed dose (or dose rate) in the phantom quantified radiation detriment. The calculations were made for homogeneous phantoms simulating newborn, 3-, 10- and 15-year-old patients, barium and iodine contrast material details, several x-ray spectra, and for imaging with or without an antiscatter grid. The image receptor was modelled as a CsI x-ray image intensifier (XRII). For the task of detecting low- or moderate-contrast iodine details, the optimal spectrum can be obtained by using an x-ray tube potential near 50 kV and filtering the x-ray beam heavily. The optimal tube potential is near 60 kV for low- or moderate-contrast barium details, and 80-100 kV for high-contrast details. The low-potential spectra above require a high tube load, but this should be acceptable in paediatric fluoroscopy. A reasonable choice of filtration is the use of an additional 0.25 mm Cu, or a suitable K-edge filter. No increase in the optimal tube potential was found as phantom thickness increased. With the constraint of constant low-contrast detail detectability, the mean absorbed doses obtained with the above spectra are approximately 50% lower than those obtained with the reference conditions of 70 kV and 2.7 mm Al filter. For the smallest patient and x-ray field size, not using a grid was slightly more dose-efficient than using a grid, but when the patient size and field size were increased a fibre interspaced grid resulted in lower doses than imaging without a grid. For a 15-year-old patient the mean absorbed doses were up to 40% lower with this grid than without the grid.

Efficiency of low-contrast detail detectability in fluoroscopic imaging.

The detectability of a static low-contrast detail in the dynamic fluoroscopic image of a homogeneous phantom was assessed by physical measurement of the signal-to-noise ratio (SNR) and by psychophysical measurement of the human observer detectability index d'. The two-alternative forced-choice method was used for human observer tests. The image data consisted of digitally recorded fluoroscopic image sequences which were displayed in a continuous loop of varying length (1-50 frames) at a rate of 25 frames/s. Human detection performance was seen to improve with the SNR in all cases studied: when the signal was made stronger, the image noise lower, or when the SNR in the image sequence was made higher by increasing the length of the image sequence. The results imply that the statistical efficiency of humans decreases slowly when the number of frames in the displayed loop is increased. This decrease of efficiency with loop length was not seen in all test series, however, and it is possible that the phenomenon is partly related to the high d' values found at the greatest loop lengths studied. When the display contrast was high, the statistical efficiency of the human observer was 30%-40% for both static and dynamic images. The efficiency was somewhat lower, 15%-25%, for images that were displayed with a display contrast gain setting more typical of fluoroscopy. The accumulation rate of SNR2 is a suitable quantity for the measurement of fluoroscopic image quality as related to a given static signal detection task. In contrast to this, visibility measurement by determination of the threshold contrast was seen to be unacceptably imprecise if the test is based on only one observer's opinion, as is often the case in practical quality assurance testing. The precision of the threshold contrast measurement could, however, be improved by using several observers and test objects with a smaller step between details than is usual.

Evaluation of image quality in fluoroscopy by measurements and Monte Carlo calculations.

We have studied image quality in fluoroscopy, as related to the detectability of low-contrast iodine or acrylic (PMMA) details added to a homogeneous 20 cm thick PMMA phantom, by experimental measurements of the signal-to-noise ratio (SNR) and by Monte Carlo calculation. The agreement between the measured and calculated SNR at equal absorbed dose in the phantom showed that the imaging performance of x-ray image intensifier (XRII) based fluoroscopic systems is well understood and can be mainly accounted for by x-ray attenuation in the phantom and the detail, and by the interaction statistics of primary and secondary (scattered) x-ray quanta in the input phosphor of the XRII. The electronic noise sources in the video chain had only a small effect on the detectability of the details studied here. The optimal x-ray tube potential was 50-60 kV for detecting the low-contrast iodine detail in the phantom, and 70-100 kV for detecting the thin PMMA detail. For the task of detecting the iodine detail the use of a fibre-interspaced antiscatter grid improved the dose-to-information conversion efficiency of the imaging system by a factor of 2.2 as compared to imaging without the grid, and additional filtering of the x-ray beam by 0.25 mm Cu increased the efficiency by a factor of 1.6. Monte Carlo results were further used to estimate the potential of increasing the dose-to-information conversion efficiency by imaging system design changes. For the detection task of a static, low-contrast, low-spatial-frequency iodine contrast material detail embedded in a 20 cm thick soft-tissue phantom, the greatest contributions for further improvement could be achieved by improved antiscatter devices, x-ray spectrum modification, and by decreasing the absorption in the material layers in front of the CsI phosphor of the XRII. Contrary to this, no significant efficiency increase could be obtained by increasing the CsI phosphor coating thickness from the present value of 180 mg cm-2, or by changes in the video chain characteristics. The maximum potential of efficiency improvement is a factor of 6.3 when compared to the reference fluoroscopy system operated at 60 kV with 2.7 mm Al primary beam filtration, and a factor of 3.9 when compared to the reference system at 50 kV with the primary beam filtration added by 0.25 mm Cu.

Detection performance of the ideal decision function and its McLaurin expansion: signal position unknown.

Although optimal decision functions for many simple detection/discrimination tasks can be cast in a form linear in the signal data, more complicated tasks require the addition of higher-order terms. This is typically the case when parameter uncertainty is allowed, in imaging for example, for the detection of a target of known size and shape but unknown position in a noise field. The simple task of detecting signals known exactly except for position, specifically detection of a "boxcar" shaped signal on a uniform data trace, has been studied in order to elucidate the relative importance of the first-, second-, or higher-order terms of the likelihood ratio decision rule. Analytical expressions have been developed to describe signal-to-noise ratios relevant for performance evaluation at low signal contrast levels, and computer simulations have been used to evaluate performance at higher contrast. It was found that for this task the first-order term (which corresponds to measuring the mean value of the data) dominates for low contrast signals but is superseded by higher-order terms (which is jth order correspond to the jth-order correlation of the data match filtered with the jth-order correlation of the signal) as contrast is increased. The quadratic term is found to be inferior to the linear term for small contrast and to the cubic for all values of signal contrast if the background is held constant. When the background level is allowed to vary, the performance of the odd-order terms decreases relative to that of the quadratic (and other even-order ones). Various measures of decision function efficiency are compared, demonstrating the severe limitations of using the simple signal-to-noise ratio (SNR) formalism for processes with non-Gaussian-distributed probability density functions. These results are valuable for guiding approaches to computational observers of signal data by showing the range of validity of suboptimal decision functions that are much easier to compute than the exact likelihood ratio solution.

A Monte Carlo study of grid performance in diagnostic radiology: task-dependent optimization for digital imaging.

A Monte Carlo computational model has been used to optimize grid design in digital radiography. The optimization strategy involved finding grid designs that, for a constant signal-to-noise ratio, resulted in the lowest mean absorbed dose in the patient. Different examinations were simulated to explore the dependence of the optimal scatter-rejection technique on the imaging situation. A large range of grid designs was studied, including grids with both aluminium and fibre interspaces and covers, and compared to a 20 cm air gap. The results show that the optimal tube potential in each examination does not depend strongly on the scatter-rejection technique. There is a significant dose reduction associated with the use of fibre-interspaced grids, particularly in paediatric radiography. The optimal grid ratio and strip width increase with increasing scattering volume. With increasing strip density, the optimal strip width decreases, and the optimal grid ratio increases. Optimal grid ratios are higher than those used today, particularly for grids with large strip density. It is, however, possible to identify grids of good performance for a range of strip densities and grid ratios provided the strip width is selected accordingly. The computational method has been validated by comparison with measurements with a caesium iodide image receptor.

SNR and noise measurements for medical imaging: I. A practical approach based on statistical decision theory.

A method of measuring the image quality of medical imaging equipment is considered within the framework of statistical decision theory. In this approach, images are regarded as random vectors and image quality is defined in the context of the image information available for performing a specified detection or discrimination task. The approach provides a means of measuring image quality, as related to the detection of an image detail of interest, without reference to the actual physical mechanisms involved in image formation and without separate measurements of signal transfer characteristics or image noise. The measurement does not, however, consider deterministic errors in the image; they need a separate evaluation for imaging modalities where they are of concern. The detectability of an image detail can be expressed in terms of the ideal observer's signal-to-noise ratio (SNR) at the decision level. Often a good approximation to this SNR can be obtained by employing sub-optimal observers, whose performance correlates well with the performance of human observers as well. In this paper the measurement of SNR is based on implementing algorithmic realizations of specified observers and analysing their responses while actually performing a specified detection task of interest. Three observers are considered: the ideal prewhitening matched filter, the non-prewhitening matched filter, and the DC-suppressing non-prewhitening matched filter. The construction of the ideal observer requires an impractical amount of data and computing, except for the most simple imaging situations. Therefore, the utilization of sub-optimal observers is advised and their performance in detecting a specified signal is discussed. Measurement of noise and SNR has been extended to include temporally varying images and dynamic imaging systems.

Radiation exposure during percutaneous nephrostomy.

Radiation doses of radiologists, assistants and patients during 21 percutaneous nephrostomies (PN) (including 11 unilateral and 5 bilateral procedures) were measured using an area-exposure meter and thermoluminescent dosimeters. The mean fluoroscopy time per PN was 12 min and the mean product of air kerma and the cross-sectional area of the fluoroscopic beam was 8.0 (range 0.41-24) Gycm2. Doses to the radiologists and assistants were generally modest, and the yearly dose limits of ICRP will not be exceeded in practice. The doses to the radiologist's fingers were found to be the most restrictive in this study. Regarding the mean dose to the radiologist's fingers (190 muGy), the yearly dose limit of 500 mSv would be exceeded after about 2600 PNs provided that his fingers are not otherwise exposed. With the maximal finger dose of 1100 muGy, this would occur after about 450 yearly PNs.

Ideal observer and absolute efficiency of detecting mirror symmetry in random images.

The problem of detecting symmetry has been studied by using digitally generated images with random pixel values. The statistical efficiency of humans and a computerized observer, the cross correlator of the image halves, has been evaluated. The efficiency of humans is approximately 100% when the image comprises only a few pixels and is notably better than that of the cross correlator. When the number of pixels in the image is increased, the detectability of symmetry gets better. For human observers detectability saturates, however, on a level corresponding to a modest number of pixels. Human efficiency in detecting symmetry is thus low when the image matrix size is large.

Dose distributions of X-rays in water: measurement with TL-dosimeters and comparison with Monte-Carlo calculations.

The dose distribution of x-rays in a water phantom was investigated with LiF (TLD-100) TL-dosimeters. The dose distribution was measured on the central axis and perpendicular to it at different depths. The x-ray tube voltage, field area, source to skin distance (SSD) and phantom thickness were varied between measurements. The experimental results were compared with Monte-Carlo calculations. Theoretical x-ray spectra were used in these calculations. The measured and calculated values agree well with each other; differences in doses were generally less than 10%. The variations in measured dose distributions due to variations in field area, SSD and phantom thickness agree well with the calculated values.