Silvia Pani

Silvia completed her undergraduate and postgraduate studies at the University of Trieste (Italy), where she worked on the development of novel imaging techniques with synchrotron radiation.

After a Marie-Curie Intra European Fellowship at UCL and a post-doctoral position at Queen Mary, University of London she joined the University of Surrey, where she is now a senior lecturer in Applied Radiation Physics and is the programme director for the MSc in Medical Physics.

Breast imaging has been a common theme in her 25-year career, starting from the highly-specialised environment of a synchrotron facility, to move to more accessible conventional sources coupled with advanced detector technology.

She has over 80 peer-reviewed publications.

Abstract:

Towards personalised breast screening intervals: using of the HEXITEC pixellated spectroscopic technology for breast density calculation

The current protocol for breast screening in the UK involves a mammogram every 3 years for women between 50 and 75 years old, which may not be needed for low-risk individuals and be too long a delay for at-risk individuals. A personalised screening interval based on individual risk factors would allow directing the healthcare resources where they are most needed, increasing the interval for low-risk individuals and reducing it for high-risk individuals.

Whilst it is widely accepted that breast density, i.e., the fraction of glandular tissue in a breast, is a risk factor for breast cancer, it is not routinely calculated or reported in the UK during screening, due to a lack of consistent and reliable approaches.

This talk will present an approach to calculating breast density based on the use of the HEXITEC pixellated spectroscopic technology and AI.

After reviewing the characteristics of HEXITEC in relation to clinical applications, the talk will present current work to determine breast density in breast-equivalent test objects from both experimental and Monte Carlo-simulated data using a modified convolutional neural network. When used on Monte Carlo generated data, the model predicted the breast density of dual-material test objects with good accuracy, with a mean square error of 0.013 ± 0.002 %.

Finally, prospectives and limitations for the application of HEXITEC to real-world scenarios will be discussed.