Advancing Breast Cancer Prevention: Integrating Interpretable Machine Learning and Expert Insights to Enhance Decision-Making in Prevention Therapy Strategy Development

Breast cancer, one of the most prevalent cancers globally and particularly in the UK, presents a significant healthcare challenge. Developing effective prevention therapy strategies is vital for reducing the risk of cancer occurrence and recurrence. In this light, our project is committed to pioneering advanced interpretable machine-learning methods to enhance decision-making in the prevention of high-risk breast cancers, encompassing two critical aspects:

1. Establishing a Relationship between Mammographic Density and Spatial Transcriptomics in Breast Cancer: Mammographic density (MD), a measure visible in X-ray images, represents the proportion of fibroglandular tissue in the breast and is easily obtainable. It is contrasted with spatial transcriptomics, a more complex and costly method that maps gene expression data to the exact locations in the tissue, providing detailed insights into the cellular environment. By linking these two, we aim to unravel the deeper causes of breast cancer using the more readily accessible MD data. This approach not only has the potential to uncover the intricate biological underpinnings of breast cancer but also holds promise for widespread application in medical practice. To achieve this, we will utilize advanced interpretable learning techniques like Graph neural networks, designed to decipher complex relationships in extensive datasets.

2. Studying and Evaluating Various Prevention Therapy Strategies: Our focus extends to the examination and assessment of diverse prevention therapy strategies to identify the most effective methods for high-risk breast cancer cases. This will involve analysing data that reflects changes in breast tissue composition and mammographic patterns under different prevention strategies. A comprehensive risk prediction model will be developed, integrating multi-omics data, including X-ray imaging, spatial transcriptomics, etc. This model will be empowered by advanced machine learning algorithms and will crucially incorporate human-in-the-loop methodologies, involving clinician experts in the modelling and decision-making process. This is essential in navigating the complexities and nuances of breast cancer prevention strategies.

At the core of our methodology is the utilization of probabilistic machine learning techniques, adept at handling the inherent uncertainties and variabilities in medical data. The interpretability of our models, especially in a clinical context, is a priority. As such, we are dedicated to enhancing the interpretability of our machine-learning models through techniques like uncertainty quantification, interpretable modelling approaches and human-in-loop methods. This dual focus on sophisticated analytics and human expertise highlights our commitment to developing dependable, comprehensible, and clinically applicable tools for breast cancer prevention. Consequently, our project represents a synergistic fusion of cutting-edge technology and clinical insight, poised to make a significant impact in the personalized healthcare landscape, particularly in the realm of cancer prevention.

References

1. Risom, Tyler, et al. “Transition to invasive breast cancer is associated with progressive changes in the structure and composition of tumor stroma.” Cell 185.2 (2022): 299-310. https://pubmed.ncbi.nlm.nih.gov/35063072/
2. Romanov, Stepan, et al. “Artificial intelligence for image-based breast cancer risk prediction using attention.” Tomography 9.6 (2023): 2103-2115. https://www.mdpi.com/2379-139X/9/6/165
3. Squires, Steven, et al. “Automatic assessment of mammographic density using a deep transfer learning method.” Journal of Medical Imaging 10.2 (2023): 024502-024502. https://www.spiedigitallibrary.org/journals/journal-of-medical-imaging/volume-10/issue-2/024502/Automatic-assessment-of-mammographic-density-using-a-deep-transfer-learning/10.1117/1.JMI.10.2.024502.full
4. Joshi, Chaitanya K., et al. “On the expressive power of geometric graph neural networks.” arXiv preprint arXiv:2301.09308 (2023). https://arxiv.org/abs/2301.09308
5. Georgaka, Sokratia, et al. “CellPie: a fast spatial transcriptomics topic discovery method via joint factorization of gene expression and imaging data.” bioRxiv (2023): 2023-09. https://www.biorxiv.org/content/10.1101/2023.09.29.560213v1.abstract

Additional information

Additional information in Find A PhD