Guidance

Breast screening: guidance for image reading

Updated 11 March 2024

Applies to England

This guidance is to support all staff undertaking image reporting and interpretation within the NHS Breast Screening Programme (NHS BSP). It outlines the expectations of the NHS BSP for clinical and working practices for staff involved and in relation to cancer detection, screening specificity and sensitivity.

It should be read in conjunction with other breast screening programme guidance, which includes specific guidance and information on:

• NHS BSP pathway requirements specification
• best practice for leading a breast screening service
• radiology and advanced radiographic practice in the NHS BSP
• clinical guidelines for breast cancer screening assessment
• use of tomosynthesis in assessment within the NHS BSP
• reporting, classification and monitoring of interval cancers and cancers following previous assessment
• interval cancers and duty of candour toolkit
• how to record vacuum-assisted excisions
• organising higher risk screening
• surveillance protocols for higher risk women
• higher risk technical magnetic resonance imaging (MRI) guidance
• implementing remote image reading
• reporting breast images from home

Mammography readers should, by repeated self-assessment, audit of practice and continuing education, strive to maintain and improve their skills to make sure that all women attending for mammographic screening receive an excellent service with minimal adverse effects.

1. Screening image reading

This guidance refers to individuals who interpret screening mammograms who will be referred to as mammography readers.

Mammography readers need to maximise screening sensitivity and specificity by detecting early stage cancers and limiting the referral and biopsy of women who do not have breast disease to minimise the adverse physical, psychological and or clinical impact of unnecessary investigations.

It is the responsibility of a mammography reader involved in breast screening to:

• report a minimum of 5,000 screening and or symptomatic cases per year, of which at least 4,000 should be screening and 1,500 should be as a first reader – this should be monitored locally on at least a quarterly basis to assess numbers being read
• adhere to local right result procedures so that the result of image reading is recorded accurately and conveyed to the woman in a timely manner (within 2 weeks of screening attendance)
• participate in external quality assurance activities (EQA schemes) the results of which will be used to compare with real life image reading quality assurance (QA) data annually
• participate in screening assessment, interval cancer reviews and multidisciplinary team (MDT) meetings

Mammography readers should also monitor their own individual and service quality data by reviewing the formal annual audit of mammography reading data (see section 4 below). This requires mammography readers to compare their data with their peers, and to demonstrate a willingness to alter their practice if indicated by the outcomes. In a few cases outcomes may suggest that retraining is required, for example by secondment to a Royal College of Radiologists (RCR)-accredited training centre. This audit should include the regular review of NHS BSP objectives, which are to:

• minimise the number of women unnecessarily recalled for assessment
• maximise the number of invasive cancers detected which have the potential to cause harm in the screened population
• maximise the number of small invasive cancers detected
• minimise the number of interval cancers, particularly false negative cases
• encourage symptomatic services to request recent mammograms when a cancer is detected symptomatically to minimise the number of unclassifiable interval cases

1.1 Protocols for image reading

1.1.1 Double reading of mammograms

Double reading of mammograms by 2 readers is mandatory within the NHS BSP. There are no restrictions as to which mammography readers can double report cases, although ideally inexperienced readers should be paired with experienced readers. Two advanced radiographic practitioners can double read mammograms but single reader arbitration must be undertaken by a third reader who is consultant level with appropriate experience and levels of screening sensitivity and specificity (see section 1.1.7 below).

Individual level image reading data is readily accessible via the Breast Screening Information System (BSIS) (see section 4.1). There is a guide on BSIS which aids analysis of these data and suggests ways to improve reader sensitivity and specificity. This guide can be found within the film reading module; view reports; under the ‘For further information click here’ banner.

The national breast screening system (NBSS) allows services to select ‘blind reading’ as an option when reporting. This allows the second reader to report images independently and blind of the previous outcome. Blind reading is considered by some to be the optimal methodology when reporting breast images. Evidence from Klompenhouwer and others (2015) suggests an increase in sensitivity at the expense of specificity. However, the previous readers outcome may be known due to supporting paperwork available as current IT solutions do not fully support paperless reporting without work arounds in place.

1.1.2 Trainee mammography readers

Any trainee mammography readers must be recorded as such within the NBSS. A trainee reader can enter results even if a case already has an overall result or reporting is finalised. The automatic arbitration algorithms ignore the opinions of trainee readers when determining the overall outcome. As such, trainee readers should not be included when specifying the number of readers on NBSS. Image reading data for trainee readers are available within the film reader quality assurance (FRQA) report in NBSS, but are not included in BSIS.

1.1.3 Direct NBSS entry

All mammography readers should use NBSS to identify the cases requiring reporting and to record their findings. This should be carried out in real time during reporting. Whether paperwork requires annotation will depend on local protocols. Where paperwork is used, all diagrams and handwriting should be clear and legible. The documentation of final outcome of reading, including arbitration or consensus, should also be clear and consistent across all systems (NBSS, screening form and clinic control sheet) used by the service.

NBSS desktop integration should be used to synchronise the picture archiving and communication system (PACS) and NBSS. Readers must cross check the paperwork, NBSS and PACS workstations are all showing the same woman. There is an NBSS user guide accompanying this guidance, which documents how to use NBSS when reporting images.

If a reader recommends that assessment is required, a reason for recall must be entered in NBSS. The reason for assessment information is shared by all mammography readers. If a reader enters reason for assessment information for a particular side of the breast, another reader or an arbitration reader can amend the data but only if they also enter an opinion of ‘Abnormal’ or ‘Clinical’ for that side. The reason for assessment is displayed on the imaging assessment record on NBSS and on the imaging assessment form.

1.1.4 Previous mammograms

Previous mammograms held within the screening service should be available to readers at the time of image reading. The mammography reader is responsible for deciding whether to obtain previous mammograms held at another service. To make sure the time from screen to result is minimised, there should be a local policy to detail the maximum waiting time for image exchange.

Use the image exchange portal (IEP) to request and transfer images. Within the NHS no charge should be made for the transfer of patient information from one service to another. Services should establish reciprocal links between the NHS BSP and the private sector to encourage the free flow of relevant radiological data and ensure that patient care is not compromised.

1.1.5 Signs and symptoms

Each breast screening service must have an agreed written policy for documenting clinical signs and symptoms, developed collaboratively between the screening office, mammographers and mammography readers. All staff must be aware of the agreed process within the service.

All relevant clinical signs or symptoms noted by the mammographer or reported by the woman while attending for screening should be documented on the client screening form and entered in NBSS.

Prior to image reading, the administration team must complete an audit to make sure that all information recorded on client forms relating to signs and symptoms is also recorded on NBSS to ensure that the mammography reader is alerted automatically.

It is the responsibility of the mammography reader to determine the significance of these breast symptoms or signs and make sure that appropriate further assessment of the woman takes place if indicated.

1.1.6 MRI for women at very high risk

Guidance on MRI for women at very high risk is available. This guidance includes:

• requirements for those reporting MRI images taken as part of the NHS BSP
• the reporting categories to be used
• NBSS data entry

There should be a written and agreed service policy for very high risk women attending for magnetic resonance imaging (MRI) screening. This information must also be flagged and highlighted to the MRI reporter. There are no equivalent NBSS flags for MRI screening. There is a standard MRI screening request form available from NBSS (a copy is also in guidance on using MRI with higher risk women). The form can be used to record clinical signs and symptoms at the time of MRI attendance. Printed copies of this form can be provided by the screening service administration team to the MRI department for completion in every case, and then sent back to the breast screening service ahead of reporting. The forms should be scanned into the hospital radiology information system (RIS) and reviewed in every case by the reader.

1.1.7 Arbitration

Services need to undertake arbitration to make a definitive decision whether a woman is recalled to an assessment clinic or returned to routine screening. Each breast screening service should determine their local reading policy. This may involve arbitrating in certain circumstances such as when:

• the first and second readers disagree
• both readers report an abnormality (for example where a service seeks to reduce a high recall rate)
• women have reported a symptom during their screening appointment

Arbitration maximises potential sensitivity for the mammography reader and allows subsequent arbitration of all potentially abnormal cases to maximise specificity of recall. Where recall rates are outside programme standards at a screening service level, arbitration may be a useful method to improve quality where both readers identify an abnormality. For some services this will increase the number of cases requiring arbitration, but for those services with a high recall rate, there is the opportunity to increase specificity of recall. Whilst arbitration in this case may be regarded as increased workload, it is likely to reduce the number of women requiring assessment.

Breast screening services undertake arbitration in different ways. Services may:

• convene a small group of mammography readers (consensus panel arbitration)
• use a third mammography reader

Where new or inexperienced mammography readers are part of a consensus arbitration panel, it can be helpful to ensure an experienced individual co-ordinates and leads the group. They can use the session as an opportunity to share image reading learning.

The arbitration process requires highly discriminative skills of screening specificity and sensitivity. These skills come with experience, continuous feedback from clinical involvement and decision-making in the assessment clinic – along with participation in audit, continuing professional development (CPD) and case review such as interval cancers. Clearly, the arbitrator cannot increase the sensitivity of the screen reading, but they can increase specificity and reduce the unnecessary recalls. These skills are not necessarily related to the profession of the arbitrator.

Single reader arbitrators, or those leading consensus panel arbitration should:

• participate fully in assessment clinics including decision-making (working at radiographic consultant practitioner or responsible assessor level)
• be a fully qualified mammography reader meeting the appropriate standards including suitable training, reporting at least 5,000 cases per year including 4,000 screening mammograms of which 1,500 are first reads
• be an experienced mammography reader with ideally at least 2 years in breast screening; new consultant radiologists must have completed appropriate training including a minimum of 5,000 cases read as a trainee, and ideally have additional experience (for example, completed a breast fellowship post)
• attend regularly, and participate in, MDT meetings
• regularly audit and review personal and team results, with evidence of reflective learning: this should include review of interval cancers, previously assessed intervals and screen detected cancers and participation in external quality assurance activities (EQA schemes)
• participate in ongoing professional development and annual appraisal

The director of breast screening should agree that an individual is suitable for the role of arbitrator and document this locally. The results of the individual and the screening service should be reviewed annually by the director of breast screening as part of local audit, clinical governance and the appraisal process.

1.1.8 Remote reading

Where remote reading is in place or is being considered, it is the responsibility of the host organisation, screening service and commissioners to make sure that appropriate IT links and reporting equipment is fit for purpose and has suitable governance in place.

Guidance on supporting breast services to report images remotely and report breast images from home is available.

1.1.9 Right results

Once a batch of images has been reported, there should be clearly documented local protocols for ensuring that each woman gets the correct screening result. Cases should be separated according to their specific outcome prior to processing within the screening office. This responsibility for separation lies with the second reader or readers following arbitration, and is part of the right results checklist. The suggested result grouping categories are:

• routine recall
• recall to assessment
• technical recall
• arbitration
• awaiting previous mammograms

1.2 Retirees

In the case of retirees who may want to read images but not participate in screening assessment, this is at the discretion of the director of breast screening. This arrangement could be considered if the individual:

• has worked within breast screening for at least the past 5 years in a substantive post
• has met or exceeded the median levels of image reading data, or higher, for screening positive predictive value (PPV) of referral and has discrepant rates that are no worse than the national average (as documented within BSIS)
• undertakes all routine audit activity as stated above in section 1, including interval cancer review – regular attendance at MDT meetings is also mandatory

If there has been an employment gap, the host organisation is responsible for this appointment but should make sure that the individual is subject to local audit and participates in external quality assurance (EQA) activities.

Retirees may participate in screening assessment and not image reading, where service needs dictate. This could be considered if the individual has worked within breast screening for at least the past 5 years in a substantive post and have met the median levels of image reading data or higher for screening specificity and sensitivity. Regular attendance at MDT meetings is mandatory.

1.3 Locums and staff from abroad

Guidance is available outlining processes and considerations for appointing locums and international recruitment.

2. Facilities and equipment

There is quality control and testing guidance for the testing of equipment used within the NHS BSP.

2.1 Image reading

Image reading must take place in a suitable room with appropriate lighting, and allow for uninterrupted reporting of images. Health and safety recommendations specify air conditioning with a minimum level temperature to ensure equipment functions adequately. There is no upper temperature level specified, but dedicated facilities should be a comfortable and suitable environment for staff.

Appropriate high-resolution dedicated mammography reporting workstations must be provided for primary diagnosis. Each monitor within the reporting pair must be a minimum of 5 megapixels (MP) with a maximum luminance no lower than 450 cd/m2. All mammography reporting workstations must be regularly calibrated to remain within 10% of the Digital Imaging and Communications in Medicine (DICOM) Grayscale Standard Display Function. Reporting monitors should be regularly checked in accordance with guidance on full-field digital mammography quality control tests.

Workstations must have full desktop integration with NBSS and PACS and be able to access all required systems as necessary.

2.2 Magnetic resonance imaging (MRI)

Guidance on MRI screening for very high risk women is available. This covers:

• MRI equipment requirements
• DICOM compliance
• reporting workstation requirements

Appropriate software, including region of interest (ROI) analysis, should be loaded and utilised onto reporting stations to support accurate reporting.

3. Quality standards for radiology

The NHS BSP standards are outlined in the consolidated standards document. Details about standards specific to radiology quality are outlined in guidelines for radiology and advanced radiographic practice.

3.1 Improving image reading

Services and individuals should access BSIS to analyse image reading data. Full details to support the interpretation of outcomes are available within BSIS.

Data for breast mammography readers is produced annually on BSIS and demonstrates individual level image reading data against national averages for comparison. The Screening Quality Assurance Service (SQAS) has produced a resource which is available on the FRQA section of BSIS to assist with:

• interpreting these data
• supporting individuals to compare their data with peers
• producing data for directors of breast screening and lead radiology advisors to enhance quality within a screening service

Further information is in section 4 below.

It is the responsibility of individual readers to be aware of and understand the FRQA data. This should be discussed and reviewed on an annual basis with the director of breast screening or lead radiology advisor, who will have overview of the reading data for the service as a whole. FRQA data supports both personal development and team quality review.

It is the responsibility of all medical and non-medical practitioners providing radiology services to monitor their team’s data as well as their own, and report any problems through their host organisation’s clinical governance process.

PPV (particularly when used in a PPV recall diagram) is a powerful audit tool for demonstrating the relationships between sensitivity and specificity. Services who need to improve their PPV of recall should consider incorporating the following into practice – services with:

• prevalent and/or incident recall rates higher than the acceptable standard are strongly advised to carry out arbitration (by an individual or panel) of all prevalent and/or incident recalls (by one or both readers) until recall rates reach achieve programme standards (see Blanks and others, 1998)
• overall recall rates of less than 2% at all screens are more likely to be associated with lower cancer detection rates and services may want to examine interval cancer rates to make sure that cancer detection at screening is maximised

Monitoring service level data can mask variation between individuals. To further improve image reporting specificity, the following audits are recommended. If:

• recall rates are too high, readers should also review their false positive recalls to ascertain if there is systematic recall of radiological features where recall is not required
• first reader recall rates are satisfactory, individuals should review all the images they did not recall where women were subsequently proven to have cancer

As a minimum, all mammography readers should formally audit their reading data and compare their results with those of their peers.

4. Breast screening information system (BSIS)

4.1 Accessing BSIS

Individuals need to request a login to access BSIS, which needs to be verified by their local caldicott guardian. An online data certification form is available. A knowledge article regarding access to the BSIS is also available on the NBSS website – see within the training section, under ‘BS Select articles’.

BSIS is accessed via Open Exeter.

4.2 Film reader QA

A guide to interpreting and using the FRQA data on BSIS is available within BSIS. It is found within the film reading module, within view reports (click on the blue text ‘For further information click here’).

It is important to remember that the data in BSIS is updated annually after completion of KC62 data. Please review the KC62 data before reviewing the film reader report to get an idea of how the service is performing and understand the service strengths and weaknesses.

Each mammography reader will have their own unique code within BSIS, which is available to that reader and the director of breast screening or lead radiology advisor.

Total number of reads as a first reader is important. Mammography readers must report a minimum of 1,000 first reads annually and therefore 3 year data should be a minimum of 3,000 reads. This is important because the data is more accurate and reflective of real life practice when the numbers are larger.

The recall rate is also more accurate the greater the numbers. Each individual will know their own recall rate and also the recall rate of their peers.

Readers should then look at the quadrant chart to see where they sit. These quadrants are a visual representation of sensitivity and specificity (see Figure 1 below). Services will expect to see mammography readers across the different quadrants, showing the different skills of the individuals. Not all mammography readers are expected to fall within quadrant 1.

Quadrant 1 (top left): indicates a cancer detection rate higher than the national average with a recall rate below the national average,

Quadrant 2 (top right): indicates a cancer detection and recall rate, both above the national average,

Quadrant 3 (bottom left): indicates a cancer detection and recall rate, both below the national average,

Quadrant 4 (bottom right): indicates a higher than national average recall rate and a cancer detection rate lower than the national average.

It is important for readers to understand their own data and to consider ways to improve their practice if it falls within quadrants 2, 3 or 4. It is important to discuss any potential actions with the director of breast screening or lead radiology advisor.

Figure 1: quadrant chart showing reader practice relating to sensitivity and specificity.

If a reader in is in:

• quadrant 1 (top left): no action is is required
• quadrant 2 (top right): review all false positive recalls – if you are in quadrant 2 then your recall rate is high which will lead to a lower PPV (although you are identifying cancers you are also recalling cases that are not cancers); it would also be useful at this stage is to look at the biopsy rate (KC62) for the service, and also useful to audit your own practice by reviewing the false positive recalls
• quadrant 3 (bottom left): consider whether it is possible to increase recall rate – if you are in quadrant 3 you need to review all the missed cancers and consider what may be your area of weakness; there is a need to increase your recall rate but also a need to ensure you are not recalling benign lesions
• quadrant 4 (bottom right): review all discrepant cancers and false positive recalls – if you are in quadrant 4 then you need to review all your false positive recalls and discrepant cancers; it may be worth sitting in with mammography readers that lie within quadrant 1 to gain insight in their image reading practice, and a training package should be developed for you with the director of breast screening or lead radiology advisor

In BSIS it is possible to view data by first reader only, second reader only or by first and second reader combined. The data should be viewed for the first reader only.

Mammography readers need to identify their own unique code and then look at the:

• total number of cases read as a first reader
• number recommended for recalled to assessment against their recall rate
• number recommended for assessment and diagnosed with cancer
• positive predictive value (PPV)
• numbers of cancers not recommended for recall (the discrepant rate)

Readers can compare themselves with peers working within the same breast screening service, and against the national average.

If the number of first reads is less than 1,000, interpret the data with caution as it is statistically weak.

The FRQA report on NBSS provides the screening identification (Sx) number of the discrepant cases and false positive recalls. This information is not available on BSIS.