Breast cancer screening backlog solution? Faster, more accurate mammogram analysis

Breast cancer screening backlog solution? Faster, more accurate mammogram analysis

Routine cancer screening has been suspended during lockdown and nearly 1 million women in the UK have missed their breast cancer check in according to charity Breast Cancer Now.

Zegami, an Oxford based medical image analysis platform, has developed a system to analyse large numbers of mammograms and identify abnormalities, which is the first stage of breast cancer screening. The system uses the recently announced Medical Imaging Server for DICOM from Microsoft.

The system also allows scientists to develop Machine Learning models to automate this analysis, making it faster and more accurate.

Initial mammogram data for the system has been sourced from The Cancer Imaging Archive (TCIA)  and consists of 3486 DICOM (Digital Imaging and Communications in Medicine) images, which while anonymised, includes pathology data, allowing this to be factored into the analysis:

https://zegami.com/demo/dicom-mammograms/

One example of the analysis is a map of the "Calcification" breast cancer abnormalities - the system highlights regions distinctly containing examples with little light areas, which are typically the benign (without a callback) instances.

“We’re really proud to build on the work we have done in identifying Covid-19 versus other forms of pneumonia from lung X-rays, with this platform which quickly and accurately allows scientists and clinicians to analyse thousands of mammograms. This is particularly important given the backlog of breast cancer screening appointments.” says Stephen Taylor, Chief Scientific Officer of Zegami.

Steven Borg, Director of Medical Imaging at Microsoft Health said “Zegami’s imaging visualization platform, leveraging Microsoft’s open source Medical Imaging Server for DICOM, makes rapid visualization and exploration of medical images possible. This visualization of data allows researchers to quickly explore the interactions of underlying variables and analyze data in ways that visually validate a hypothesis. Zegami has built a powerful use-case, that will allow scientists and clinicians to develop better image analysis and AI solutions to improve medical outcomes for patients.”