Imaging has become an essential tool in medical diagnosis and treatment throughout the world. The development of digital radiology techniques has increased the scope and ease of use, making optimisation for radiological protection key to ensuring that the doses used and the levels of image quality are appropriate. The International Commission on Radiological Protection (ICRP) has recently prepared two reports to provide guidance on implementation of programmes of optimisation in digital radiology and providing guidance on taking the process forward.

ICRP Publication 154 [1] sets out general principles with sections on the optimisation process, analyses of patient doses and evaluation of image quality, as well as education and training needs. It highlights three fundamental requirements when setting up optimisation programmes.

(1)  

The need for close collaboration between professionals, with radiologists, radiographers and medical physicists. Each of these groups has unique key skills that compliment each other when working together in an optimisation team.

(2)  

Access to the appropriate methodology, technology and expertise. Staff must have the necessary instruments, tools and training to perform their roles and evaluate imaging performance for all types of equipment.

(3)  

Provision of an organisational structure with processes to ensure that equipment performance tests and patient dose surveys are carried out at appropriate time intervals and results feed into reviews of imaging protocols to improve radiological protection.

The first step after installation of x-ray equipment is commissioning to check that the equipment is ready for clinical use and to establish baseline performance values against which the results of routine tests can be compared in the future. This is the time when medical physicists first assess operation of the equipment and can evaluate different settings and options for clinical use. Clinical protocols should be set up for image acquisition at this stage, with advice from the vendor’s applications specialist and input from all members of the optimisation team.

ICRP Publication 160 [2] contains practical methodologies needed to carry optimisation forward. The general principles of optimisation apply to all digital radiology equipment, but the ways in which they are implemented depend on the modality and the features available to allow dose levels to be reduced while maintaining adequate image quality for the clinical task. The report contains sections for the different modalities: general radiography, fluoroscopy and computed tomography (CT), as well as sections on paediatric imaging and pregnant patients. It replaces content in earlier reports on digital radiography (ICRP Publication 93 [3]) and CT (Publications 87 [4] and 102 [5]) and will supplement material in Publications 85 [6], 117 [7] and 120 [8] on fluoroscopy and interventional procedures and Publication 121 on paediatric imaging [9]. These ICRP publications highlight the importance of surveys of patient dose, but emphasise that it is the next step, the optimisation of radiological protection, that is important. The improvements that can be made and actions taken to optimise radiological protection will vary significantly depending on the facilities available and the experience and expertise of the staff. The ICRP publications aim to address this challenge by providing detail on actions for facilities within broad categories related to the facilities available and the level of optimisation already achieved. They set out options for facilities covering four levels:

(D) Preliminary. Represents a basic infrastructural stage when a service is first set up. This is required as a prerequisite for initiation of the optimisation process, but before actions have been taken to start the proper process of optimisation.

(C) Basic. Represents the stage at which dose and image quality control (QC) tests are carried out, including documentation covering parts of the system and adoption of diagnostic reference levels (DRLs) under consideration [10].

(B) Intermediate. This marks the stage in which there is comprehensive QC testing relevant to the clinical applications and dose audits in the context of DRLs are carried out. Optimisation teams comprising radiographers, radiologists and medical physicists review and revise protocols using results from QC tests and audits.

(A) Advanced. At this stage, departments involve imaging professionals working together as a multi-professional team to continually review and improve optimisation. Software tools may be available ensuring regular testing, audits and review of protocols according to an established programme enabling a higher level of optimisation.

Surveys of patient doses should be carried out at an early stage, with audits performed to compare with national DRLs. In some cases, local DRLs (LDRLs) are set as they provide an indication of dose performance relative to other centres and so form an important part of the optimisation process. Dose surveys provide something researchers can ‘get their teeth into’, while the follow-up to a survey and the process of optimisation itself sometimes seem nebulous. As a result, dose surveys and audits against DRLs are often written up in papers without the most important optimisation aspect. A shift in emphasis is needed to encompass the whole optimisation process. When surveys are carried out, a question that often arises is how researchers should share results within the medical imaging community. ICRP Publication 135 [10], which is concerned with the establishment and use of DRLs as an aid to identifying x-ray procedures in need of optimisation, has been the most widely downloaded ICRP Publication on guidance in medicine.

Surveys of patient doses for the purposes of setting DRLs are within the scope of JRP and the journal has published a number of DRL-themed papers in recent years. As patient dose surveys have become routine around the world, the number of DRL-themed submissions has increased and it is simply not practical to publish them all. Submissions are typically regarded less favourably by reviewers and editors if they report the results of small-scale studies, lack novelty or breadth of interest, or lack clear descriptions of methodology. Those that are published often attract little interest, as indicated by the number of views and downloads. We do not wish to discourage performing dose surveys per se, rather advise careful consideration of whether results need to be published in a peer reviewed journal.

If an institution has carried out a dose survey and set DRLs, this will mainly be of relevance to the local imaging community or perhaps a national organisation collating data for the country. DRLs are an important tool in achieving optimisation, as the impact that exposure parameters have on patient dose may go undetected unless dose levels are monitored. But proposing DRLs based on a survey of one or more hospitals is not itself a reason to publish the results in a scientific journal. It is what comes out of the survey and analysis of the results that is important. It is hoped that ICRP Publications 154 [1] and 160 [2] will help to focus attention back onto optimisation as the main priority.

Before submitting a manuscript to be considered for publication, authors should consider whether their work helps radiation protection professionals and clinicians to do their jobs. Things that might be useful and provide value to such a study include:

(1)  

Novel methods for optimisation of radiological protection

(2)  

Application of new equipment, features and techniques

(3)  

Explanation of why doses for a procedure were high and changes made to address any issues

(4)  

Novel methods for collecting and updating patient dose data, including methods for handling missing data and overcoming data collection difficulties

(5)  

Approaches taken to check that image quality is adequate using new imaging methods.

Simply giving figures against which others can compare doses is unlikely to justify publication as a full paper, unless this is a large national survey collating data from many centres or procedures for which published results are limited. Moreover, if a study is written up, the methods used must be robust and the numbers of patients included must be sufficient to provide valid results. The minimum number of examinations for a patient dose study for publication would normally be 100–200, but ideally 300–400 are required to provide good accuracy [11]. Although ICRP 135 states that data should be collected for at least 20 patients, this is a minimum value for a local patient dose survey rather than a study for publication [10]. There should be an analysis identifying what has been learned and presentation of ideas that could potentially be useful to others. In this case, a small study might be suitable as a ‘practical matters’ article, but should still have a reasonable sample size and methodological clarity. Other suitable alternatives for dissemination include local journals, professional society magazines, local meetings and websites. To assist those planning a patient dose study, a survey gathering information on different approaches and methods used in patient dose studies across the UK was reported recently [12]. Below, we give a few pointers to things that authors preparing DRL-themed papers might consider.

Links between researchers and hospital staff

A problem that has diminished the value of patient dose surveys carried out by university researchers is the limited contact with those carrying out the procedures. This makes feeding results into the optimisation system for the radiology department difficult. The purpose of performing surveys and setting DRLs is to identify units and procedures requiring attention. The two recent ICRP publications discussed above [1, 2] highlight the need for close collaboration between professionals, with radiologists, radiographers and medical physicists working together in an optimisation team. If a survey is carried out by a group outside the hospital department, then close links need to be established with the team carrying out the examinations and action taken to address any deficiencies.

Study size

National surveys, reporting data from multiple centres, are preferred to single centre studies. In general, single centre or LDRL studies are only considered for publication if they report data on new procedures for which little or no published figures currently exist. This may include, for example, newly developed fluoroscopically guided interventions. Alternatively, single centre studies may be suitable for publication if they present a particularly novel optimisation methodology that could be applied more broadly.

Age and weight ranges

DRLs are typically quoted for standard sized patients and for particular age ranges [10], although this should not be regarded as being set in stone. The inclusion of, or sole focus on ‘non-standard’ sized patients, e.g. obese individuals, could be considered and could represent an interesting avenue for research. The important consideration is to ensure dosimetric data are representative of a clearly defined group of patients, even if this group is very inclusive, e.g. all adults.

While weight is a better indicator of patient size (from a dosimetric perspective) it is acknowledged that such data are often not recorded, or may be outdated or estimated. Patient age is an acceptable alternative, although limitations should be acknowledged. If studies are carried out with downloads of data for over 200 patients then the data can be taken as representative of an average patient although some indication of the average weight should be included. But for studies involving smaller patient numbers, data on size/weight is essential.

In order to help other researchers identify relevant research, the title of the paper should state whether the results represent children or adults. Calculation of DRLs for paediatric patients require special consideration as there is no single ‘standard size’. X-ray output must be raised with increasing patient size to maintain image quality, meaning DRL quantities tend to increase correspondingly with body size. Weight is preferred for DRL grouping [10]. The weight groups <5, [5–15), [15–30), [30–50) and [50–80) kg are recommended by European guidelines [13]. If weight data are not available, the age ranges <1 month, [1 month—4 years), [4–10 years), [10–14 years) and [14–18 years) may be used. The use of different weight or age ranges should be clearly justified (e.g. easier comparison to other data). Some groups have calculated dose per unit body mass (e.g. air kerma-area product [PKA] per kg). This requires caution, however, since such ratios are not necessarily independent of patient age or mass category, and so do not avoid the need for age or weight stratification.

Stratification by procedure type

Wherever possible, data for CT scans should be collected for specific clinical indications. Non-routine or specialised procedures should be kept separate. For example, brain perfusion imaging scans should not be included in the dataset for head CT scans. Radiotherapy planning CT scans should be identified and kept separate from diagnostic scan data. Usually, no distinction is made between diagnostic and screening examinations, although this could be worth exploring.

For interventional radiology, there is a large variation in complexity between procedure types and the grouping together of diverse procedure types into a single DRL is unlikely to be useful. If sample sizes are insufficient to allow stratification by specific procedure type, then consider extending the data collection period until sufficient data are available. In many cases, DRLs for interventional radiology procedures can be further stratified by complexity, for example, the number of stents placed during coronary angioplasty. Options that might be used for studies on interventional procedures are discussed in ICRP Publication 135 [10].

For general radiography, separate DRL quantities should be given for different projections, e.g. postero-anterior and lateral chest radiography. The combination of doses from multiple projections can make comparisons difficult and unusually high or low doses difficult to explain. If multiple projections are combined, this must be clearly stated and justified. Mammography DRL quantities should be reported with different projections, e.g. cranio-caudal and medio-lateral oblique.

Multi-phase CT

CT examinations frequently involve the use of iodinated contrast media (ICM). Scans may be performed prior to ICM administration or at different ICM phases, e.g. arterial, venous or portal. In addition, a single slice bolus tracking scan may be performed to monitor ICM concentration in a particular vessel, allowing arterial or venous phases to be triggered at the appropriate time. Each of these scan phases may have associated dosimetric data, including volume averaged CT dose index (CTDIvol). The failure to clearly describe procedures for identifying and accounting for multi-phase scans can make interpretation of data extremely difficult. We recommend that CTDIvol is presented as the mean of the individual scan phases (excluding bolus tracking scans) while dose length product (DLP) is reported as the sum for all phases. The number of scan phases should also be reported. Where multi-phase scans are routine for particular indications, then indication-specific DRLs should be considered. For example, head CT imaging for suspected stroke is typically single-phase, while scans for brain tumours are typically dual-phase. Different hospitals may have a different breakdown of scan indications (e.g. high proportion of dual-phase scans in a regional centre for brain tumours) meaning comparison of figures is challenging.

Units

Figures for national DRLs and LDRLs involving 10–20 centres should be based on 3rd quartile (75th percentile) of the medians of DRL quantities for individual centres. Local DRLs involving a smaller number of centres (<10) should be based on median figures [10]. Considering measurement uncertainties, overly precise figures should be avoided. We recommend that both CTDIvol and DLP are quoted to two or at most three significant figures. For CT studies, we recommend units of mGy for CTDIvol and size specific dose estimate, while mGy * cm should be used for DLP. The phantom size (e.g. 16 or 32 cm) should be clearly stated.

For general radiography and fluoroscopy, units of Gy * cm2 are preferred for reporting PKA, as opposed to cGy * cm2 or μGy * m2. Either way, it is important to mention procedures to calibrate measurements to national standards and whether the results include a calibration factor. The calibration factor is typically in the range ±25%, therefore comparison between different published figures is difficult if it is not known whether a calibration factor has been applied or not. It should also be mentioned whether PKA was based on measurements obtained using a large area ionisation chamber attached to the x-ray tube or estimated by the equipment based on exposure factors.

Image quality

Reductions in patient dose to the point where they affect interpretation of the clinical image are counter-productive. Thought should be given to ensuring that the imaging performance of the equipment is sufficient. Image quality is more difficult to quantify than dose and although there are a variety of objective parameters, these do not necessarily link readily to clinical imaging performance. Moreover, it is important to ensure that the clinicians interpreting the images find them suitable for the clinical task. Thought should be given as to whether objective measures are sufficient or more subjective ones should be used. This is another reason why including a radiologist or other medical radiological practitioner in the optimisation team is important.

Other data

For general radiography and fluoroscopy, PKA itself is a limited indicator of patient dose because the same value of PKA can result in very different patient doses depending on x-ray energy spectra. The energy spectrum for a given exposure is a function of various parameters but depends principally on tube potential and beam filtration. Presenting PKA in the absence of such data can make comparisons difficult.

Many fluoroscopy systems are biplane and have the capability to record dose quantities separately for each x-ray tube. Presenting separate PKA figures for each output, as well as combined, can allow greater insight. Conversely, the summing of cumulative air kerma (Ka,r) for both planes is of limited usefulness, especially in terms of evaluating risk of skin injuries. Additional useful information includes fluoroscopy time, the proportion of PKA from fluoroscopy, digital acquisitions and the number of acquisitions/digital subtraction angiography runs and whether an anti-scatter grid is typically used

Data governance

Data governance procedures should be clearly described. This includes describing how data are initially recorded (e.g. manually or automatically, on computer or on paper, radiological information systems or dedicated dose management systems), any processing steps (e.g. adjustment for calibration), how data are stored and how they are analysed. Data collection for the purposes of optimisation and DRL setting is often challenging and description of methods to overcome these challenges can be useful for readers. This includes methods for identifying and handling missing data and unusual figures. Various approaches to handling outliers have been identified, e.g. restricting to the 5%–95% or 10%–90% range, using median instead of mean etc. Whatever the approach, this should be justified and documented.

Analysis

One of the more frequent problems with DRL-themed papers is a lack of detailed analysis of results. Frequently, authors appear to be at a loss to explain why their doses are higher or lower than other published studies or are only able to speculate. Reasons for differences and similarities with previous publications should be thoroughly investigated before the work is submitted for publication. If reasons for higher or lower doses are not identified, then the paper is unlikely to be useful for readers. Importantly, results should be interpreted in the broader context of optimisation. This means, for example, that higher doses are not necessarily a bad thing. In some cases it can be shown that higher doses are justified in terms of meeting image quality requirements [14].

General considerations

A checklist of recommendations for DRL-themed papers is provided in table 1. Authors are advised to make best use of appendices and/or supplementary materials for large tables. JRP (along with many other similar journals) is now online-only and there is generally no limit to the size of supplementary datasets.

Table 1. Checklist for DRL papers.

TopicRecommendationTitleInclude the phrase diagnostic reference levels, study region and body part. Specify as adults, children or bothMethods Data governanceDescribe how data are collected and stored (e.g. paper or electronic, radiology information system or dose management systemData collection periodClearly stated, including overlap with other published studies.Patient age and weight rangesClearly stated for both data collection and DRL establishment. DRLs for paediatric examinations stratified by weight or age.Outliers and errorsDescribe procedure for identifying errors and outliers and how these are dealt withData analysisIndicate software used to analyse dataCalibration of equipmentClearly state whether dose measuring devices are calibrated to national standards and whether a calibration factor has been applied to dataSample sizeThe minimum number of examinations should be 100–200, but ideally 300–400Results DRL quantities and unitsCT: CTDIvol in mGy, DLP in mGy * cm (2 significant figures)General radiography/fluoroscopy: ESD in mGy, PKA in Gy * cm2 (usually 2 significant figures are sufficient), fluoroscopy time in minutes and secondsMammography: mean glandular dose (MGD) and ESD in mGyNuclear medicine: Activity in MBq (2 significant figures)Other data (where available)All: age and weightCT: pitch, number of scan phases, reconstruction protocolGeneral radiography/fluoroscopy: tube potential, filtration, detector typeMammography: projection angle, breast thickness, tube potential and filtrationNuclear medicine: activity per kg patient weight

If DRLs have already been published for a given region and procedure type, the reason for updating should be specified. This may include changes in equipment and adoption of new optimisation techniques, it is also helpful to mention if the data collection period in the updated study includes any previously published data.

What to do with your results

If researchers find that their work is not appropriate for publication in an international journal such as JRP, then it may still be suitable for presentation at local meetings, sharing the data with those to whom it is relevant. Effort in such projects is needed for progress. Putting effort into patient dose studies and preparing reports or papers is important for facilities at any level and should become a springboard for carrying optimisation projects forward. Optimisation is not a static process to be ignored and forgotten once a particular goal has been achieved but a continuing one. There should be frequent monitoring of performance, feedback of experience, and regular review of imaging protocols to refine procedures taking account of findings, building on the initial groundwork. It is here that provision of an organisational structure with processes to ensure that tasks, such as equipment performance tests, patient dose surveys and review of protocols, are carried out is important [1]. It is through these processes that programmes can be developed in hospitals throughout the world so that optimisation of radiological protection becomes part of routine practice in diagnostic radiology.

No new data were created or analysed in this study.