Background
Infertility poses a significant challenge in the United States, affecting approximately 12% of reproductive-age women who struggle with achieving pregnancy or maintaining it to full term. Assisted Reproduction Technology (ART) is the preferred treatment for infertile couples, involving the fertilization of eggs in a petri dish, their cultivation for a few days, and subsequent transfer back to the patient with the goal of achieving pregnancy. While ART is a common procedure in the US, the success rate of clinical ART transfers remains uncomfortably low at around 30%. One of the primary hurdles in ART lies in the selection of embryos for transfer. In recent years, we have made significant progress in developing machine learning-based algorithms that can detect visual biomarkers and developmental stages of pre-implantation embryos using microscopy images and videos. These algorithms generate a score indicating the most viable embryo for implantation.
Research Goal
The objective of this project is to create an interactive visual analysis tool that incorporates the expertise of domain experts throughout the analysis process. The tool will enable scientists to explore the original video and image data, examine the segmentation and viability scores generated by automated machine learning methods, and explore and visualize high-dimensional feature spaces. By doing so, they can gain a deeper understanding of and verify the results obtained from the automated scoring algorithm. By enhancing the analysis process with an interactive visual tool, we aim to improve the selection of embryos for ART transfers and enhance the overall success rate of the procedure. This project provides an excellent opportunity for computer science students to contribute to the development of cutting-edge technology in the field of reproductive medicine. Furthermore, the application can help in active learning approaches pursued by collaborators.
Requirements
- Programming skills (JavaScript)
- Basic knowledge about data visualization
- ML background a plus, but not necessary
Project Collaborators
This project is run by Johanna Beyer and Eric Mörth at the Harvard University VCG. See below contact information.