Breast cancer (BC) is the commonest cause of death in young women. Breast screening in women aged 35-45, at increased risk due to their family history, has been shown to improve survival. However, 80% of women who develop BC do not have a family history. Numerous studies have shown that high mammographic density (MD) is one of the strongest risk factors for BC development. Full field digital mammography (FFDM) can be used to assess MD, however it is not recommended for population BC screening in those <40 years of age due to the concerns about the use of ionising radiation. Safe and accurate high throughput methods to quantify MD in young women are thus required to improve risk prediction and reduce BC mortality. This study aims to develop a low dose mammogram, with quantification of density using artificial intelligence, to facilitate high throughput risk assessment in young women. 600 women aged 30-45, previously identified as being at increased risk of BC and attending for annual mammography at The Nightingale Centre will be recruited. Participants will undergo FFDM of the right breast as usual, however, following acquisition of the craniocaudal (CC) view, the breast will remain compressed and the mammogram dose reduced by 90% to deliver a LD mammogram. This process will be repeated for the right medio-lateral oblique (MLO) view. The left breast FFDM will proceed as normal. It is estimated that each extra exposure will take 1-2 minutes only. Deep machine learning methods will be used to define the relationship between standard FFDM views and their low dose counterparts and determine which view (CC vs MLO) provides the best correlation to be taken forward to the next stage of the research.

The more prolonged compression of the breast required to acquire the LD images may cause some additional breast discomfort. However, in a similar study in The Netherlands only 1% of women could not tolerate the procedure.

Participants will receive an extra dose of ionising radiation amounting to 20% of the radiation dose of a single FFDM view (of which 4 are usually taken, 2 on each breast). The risks associated with this are described in section B and entered into the PIS.

The research question for this study is whether an automated, low dose mammogram can be developed to provide an accurate assessment of mammographic density, and thus breast cancer risk, in women aged 30-45.

The two key objectives are:

1. To develop machine learning methods for automated quantification of mammographic density using low dose mammograms in a study of 600 women attending for annual Full Field Digital Mammography (FFDM).
2. To validate Automated Low Dose Risk Assessment Mammography (ALDRAM) against automated FFDM assessment that has been shown to predict BC risk.

Breast cancer is the commonest of all causes of death in young women. Currently the only factor triggering referral to risk assessment clinics is the presence of a family history. Although this is known to confer increased risk, only 20% of young women who develop BC actually have such a family history of the disease. Thus 80% of cases come 'out of the blue' and methods to assess risk in the general population are required if we are to improve survival (screening in those at increased risk aged 35-45 has been shown to reduce mortality). High mammographic density has been shown in multiple studies to confer significantly increased risk of BC, however population screening with full field digital mammograms is not recommended in those <40 primarily due to the reduced sensitivity and use of a moderate doses of ionising radiation. There is thus an urgent need for safe, high throughput, techniques to accurately define mammographic density in young women - the purpose of this study. Following consent and initial Case Report Form (CRF) completion, mammography will commence as standard. The right breast craniocaudal (CC) view will be performed first and the steps below followed:

1. The breast will be positioned as standard for the CC view.
2. The Automatic Optimisation of Parameters (AOP; also known as Automatic Exposure Control (AEC)) will be performed for target, filter, Peak KiloVoltage (kVp) and MilliAmpere second (mAs) selection for Full Field Digital Mammography (FFDM).
3. The full field CC view will be performed as standard.
4. The breast will remain compressed whilst the radiographer manually reduces the mAs by 90% or to the minimum value of 4 mAs (the machine does not deliver less than 4 mAs). No other parameters will be changed. A repeat CC exposure at this reduced dose will be performed.
5. The machine will be reset to automatic and the subsequent mediolateral oblique (MLO) view will be obtained and the low dose view performed again as above.
6. The left breast CC and MLO views will then be obtained as standard.
7. This concludes the woman's participation in the study. The images from the low dose and FFDMs will then be processed to determine whether mammographic density from low dose mammograms correlates with that from FFDM and what modifications are required to automated algorithms for future studies.

Inclusion Criteria:

• Females with moderate to high risk of breast cancer, identified through and attending for annual review and mammography at the family history clinic, Nightingale Centre, Manchester
• Current age 30-45 years
• Capable of providing informed consent to a patient information sheet written in English

Exclusion Criteria:

• Prior breast cancer
• Prior breast augmentation or reduction
• Participation in the TARA-Prev study which included an additional mammogram and thus radiation dose

Females with moderate to high risk of breast cancer, identified through and attending for annual review and mammography at the family history clinic, Nightingale Centre, Manchester

• Current age 30-45 years
• Capable of providing informed consent to a patient information sheet written in English

Share Protocol