ARU project team
Raj Mani Shukla
Lecturer
This project aims to create a tool that enhances women's ability to select the contraceptive method best suited to their needs, and to improve healthcare professionals' ability to provide more accurate advice regarding the effects and side effects of individual contraceptive methods, by leveraging the power of routinely collected GP data using a machine-learning algorithm.
At a molecular level, the individual effects of contraceptive methods stem from differences in the expression of various enzymes and receptors that process contraceptive hormones. Additionally, they are influenced by factors such as a woman’s weight, age, previous pregnancy experience, smoking habits, and other poorly understood factors.
Because of the multitude of factors affecting the side effects experienced by an individual woman, predicting them using traditional statistical methods is challenging. To tackle this challenge, we plan to employ artificial intelligence (AI) algorithms, as they can extract relevant information from large-scale historical datasets to predict potential side effects.
Given the sensitive nature of healthcare applications, we aim to incorporate responsible and trustworthy AI techniques that provide accurate predictions to instil confidence in both patients and doctors in using such tools.
Most modern contraceptives are highly effective, but many women struggle to find a method that suits them without causing troublesome side effects.
Inconvenient side effects of hormonal contraception such as headaches, unpredictable vaginal bleeding, low mood, and breast pain can significantly reduce quality of life and well-being for women, and reduces the acceptability of the contraceptive method.
The side effects of most modern contraceptive methods stem from their artificial hormone content. These hormonal effects are highly individualised and influenced by various attributes in each woman.
As a result, one woman might find that an effective method causes unacceptable variation in her menstrual bleeding, whereas another will find that the method does not have this side effect, and as a result is very acceptable. Some women experience low mood and depression because of a method which has particular hormonal ingredients, whereas others find that their mood and well-being improves whilst using the same method.
As a result, when a woman and her doctor discuss contraceptive options, it is challenging for the woman to receive reliable advice on the likely side effects of a specific method. It is very difficult to predict whether an individual woman will experience a particular side effect.
This means that women often try a variety of methods consecutively over months or years before finding one that is most suitable. This is the problem this project aims to address, by harnessing the power of AI to spot patterns in large datasets that are invisible to human or standard statistical analysis.
This project is developing an algorithm to utilise various attributes known about a woman, including her experiences with previous contraceptive methods, as well as her weight, height, age, previous pregnancies, and other relevant factors.
It will use synthetic clinical records from the Clinical Practice Research Database, which contain such comprehensive information, to train and test various AI models. The best-performing model can then generate an individualised prediction of the likelihood of a given woman experiencing a particular side effect.
Lecturer