AI-based denoising of phase-contrast CT for breast cancer imaging

Breast cancer is the most common cancer in women worldwide with more than two million new cases diagnosed annually. It is also one of the two leading causes of cancer-related deaths in women.

Digital Mammography (DM) is the prevalent imaging method for breast screening. Yet significant limitations persist despite recent innovations such as digital breast tomosynthesis (DBT). These limitations include low sensitivity and specificity, radiation doses and painful compression of the breast. Unlike DM, DBT is capable of 3D imaging of the breast, albeit with limited depth resolution.

In contrast, computed tomography (CT) based methods, including phase-contrast CT (PCT) technique employed by our team, are superior to both DM and DBT in their ability to create high-fidelity 3D representations of the tissue density distribution in the breast.

We propose to investigate AI and ML based methods for denoising CT images of the human breast. Initially, we will analyse PCT scans of mastectomy samples collected by our team over the last 10 years at the Imaging and Medical beamline of the Australian Synchrotron.

While image denoising is a well-established area of information technology, denoising images without loss of spatial resolution, as required in medical CT, is a challenging problem. This has been the subject of recent research in Australia and overseas using AI and DL, with promising results published in scientific literature in the last few years.

We will draw on a significant existing dataset of PCT scans of mastectomy samples, collected under different imaging conditions, including different doses / noise levels. This can be used for effective training and fine-tuning of AI models.

Working with MDAP will provide expertise in working with complex GPU infrastructure, AI, algorithm and software development, and data analysis.

The outcomes of this project will greatly benefit our PCT research and will be applied to the world-first synchrotron-based breast cancer PCT scans of live patients scheduled at the Australian Synchrotron for 2024.

In the longer term, the methods developed as a part of this project could be expanded further to imaging of other organs which have been already demonstrated to benefit from X-ray phase contrast, such as joints (in connection with osteoarthritis), the heart and brain.