Subject

Breast cancer (early) detection and diagnosis rely heavily on the analysis of mammograms, particularly on breast microcalcifications (MCs) analysis. These tiny calcium deposits are reported to be early indicators of breast cancer. The dynamic nature of MCs, in terms of their appearance and/or evolution over time, can provide important insights – which nowadays are barely studied/analysed. Additionally, sequential mammography image datasets are scare and hence we are limited to performing in depth analyses. For existing datasets, one of the significant challenges is the accurate registration and tracking of MCs across different scans. Challenges in such analysis arise also due to the daily changes in the breast caused by natural factors. Given all this, there is a need to analyse the evolution of MCs over time and develop methodologies that can accurately register them.

Kind of work

Objective: The objective of this thesis is to develop and evaluate traditional (and/or DL) image registration algorithms to register sequential mammography images, enabling the accurate tracking and analysis of MCs over time. Additionally, in this thesis we also seek to investigate whether improved classification results of (individual MCs and/or clusters of MCs) can be obtained considering also MCs evolution info.

Description of work:

Literature Review (ETOC: 2 months): Literature review of existing methods for mammography image registration, MCs analysis and existing datasets of sequential mammography images. The literature review must also cover DL methods used to detect and classify breast MCs.

Dataset Familiarization (ETOC: 1 month) - Gain knowledge of available datasets for sequential mammography images, with a specific focus on the dataset specified here. You should clearly gain insight about the dataset's structure, the number of images available, their suitability for tracking MC evolution, etc.

Algorithm Implementations (ETOC: 6 months) - Develop traditional (and/or DL based) framework for registering sequential mammography images across different scans. Additionally, the analysis must extend to also evaluating the implications of individual and cluster MCs evolution on breast cancer diagnosis (or BIRADs classification).

Framework of the Thesis

Related work:

[1] Brahimetaj, R., Willekens, I., Massart, A.?et al.?Improved automated early detection of breast cancer based on high resolution 3D micro-CT microcalcification images.?BMC Cancer?22, 162 (2022). https://doi.org/10.1186/s12885-021-09133-4

[2] Loizidou, K., Skouroumouni, G., Pitris, C. et al. Digital subtraction of temporally sequential mammograms for improved detection and classification of microcalcifications. Eur Radiol Exp 5, 40 (2021). https://doi.org/10.1186/s41747-021-00238-w

[3] Loizidou K, Skouroumouni G, Nikolaou C, Pitris C. A Review of Computer-Aided Breast Cancer Diagnosis Using Sequential Mammograms. Tomography. 2022 8(6):2874-2892. https://doi.org/10.3390/tomography8060241

[4] Yujun Guo, Jasjit Suri and R. Sivaramakrishna, "Image Registration for Breast Imaging: A Review," 2005 IEEE Engineering in Medicine and Biology 27th Annual Conference, Shanghai, China, 2005, pp. 3379-3382, doi: 10.1109/IEMBS.2005.1617202.

Number of Students

1

Expected Student Profile

Following an MSc in a field related to one or more of the following: Computer Science, Biomedical Engineering, Applied Computer Science - Digital Health.

Strong programming skills (Python).

Experience with image processing and DL.

Interest/Motivation in developing state-of-the-art DL methods and conduct experiments.

Ability to write scientific reports and communicate research results at conferences in English.