Unraveling the Threads of Fertility

Table of Contents

Done under the supervision of Professor Philip Geevarghese from Chennai Mathematical Institute (CMI)

Abstract

This projects presents a comprehensive analysis of fertility treatment outcomes using the rich dataset from the Human Fertilisation and Embryology Authority (HFEA) Anonymized Register. Leveraging data spanning nearly three decades, from 1991 to 2018, this study delves into the factors influencing the success of various fertility treatments, including In Vitro Fertilization (IVF) and Donor Insemination (DI). The HFEA, mandated by the Human Fertilisation and Embryology Act 1990, meticulously maintains the longest-running database of its kind globally.

The dataset, organized into yearly CSV files, underwent meticulous preprocessing to align with a usable format. The research focuses on files from 2014 onwards, employing a systematic approach to column selection. The official ‘Data Dictionary’ provided by the HFEA serves as a crucial guide in understanding the nuanced variations in column names, ensuring the robustness of the subsequent analyses.

Our study employs probabilistic programming, specifically leveraging Pyro Probabilistic Programming Language (Pyro PPL), to construct a predictive model aimed at unraveling the intricate web of factors that impact the success rates of fertility treatments. By exploring the probabilistic relationships between treatment outcomes and various covariates such as patient demographics, treatment protocols, and donor characteristics, we aim to elucidate hidden patterns and provide valuable insights for clinicians and researchers in the field.

The project showcases not only the technical aspects of the modeling process but also the implications of our findings on the broader landscape of fertility research. Through a synthesis of statistical analyses, predictive modeling, and interpretability tools, our research contributes to a deeper understanding of the dynamics governing fertility treatments, paving the way for informed decision-making, personalized treatment strategies, and advancements in reproductive medicine. We anticipate that our insights will serve as a valuable resource for clinicians, policymakers, and researchers, fostering progress in the pursuit of improved fertility outcomes and patient care.

Python Libraries Used

• pyro
• pandas
• polars
• matplotlib
• csv

Techniques Used

• Probabilistic Programming
• Data Cleaning
• Principle Component Analysis
• One-Hot Encoding