Mahmoud Aldraimli joined the University of Westminster in September 2017 as a Doctoral Research student in Health Data Science and Machine Learning funded by a Quintin Hogg Trust scholarship and supported by the Health Innovation Ecosystem. He is part of the Health & Social Care Modelling Group within the School of Computer Science and Engineering and an associate member of the Cancer Research Group within the School of Life Sciences, investigating how Artificial Intelligence and Big Data can be leveraged to predict breast cancer occurrence using data obtained from the UK Biobank. Mahmoud is also part of the UK STFC-funded Radiotherapy Machine Learning Network, based at the University of Manchester, leading a small group of experts working together on the application of Machine Learning to predict radiotherapy toxicity in cancer patients.
Programming Principles with Python
Programming Principles with Java
1. Building Machine Learning Models to Predict Breast Cancer Occurrence
Reducing mortality rates from major NCDs is at the heart of the World Health Organisation (WHO) agenda. The UK NCD profile published by the WHO in 2014 showed that NCDs account for approximately 89% of all mortality. Public Health England published in 2014 that a quarter of the UK population has a long-term medical condition (including major NCDs) and the number of people with multiple conditions is expected to rise.
This study aims to identify factors which influence the risk of breast cancer occurrence. This is the primary aim as in the long term this could lead to better understanding of the interplay between different illnesses and breast cancer occurrence.
Our approach will be to use transparent and opaque Artificial Intelligence (AI) to identify whether there is a link between factors associated with the increased risk of diabetes, obesity and CVD and breast cancer. In particular, we will use a Machine Learning (ML) approach for intelligent data analysis supported by the current digital revolution in collecting and storing data.
Following a systematic review of the UK Biobank, we plan to analyse raw and derived medical and non-medical variables in big datasets. The analyses will examine whether these variables are correlated with the occurrence of breast cancer, whether these relationships persist in the presence of other variables, and the potential role of obesity, diabetes and CVD in the breast cancer risk prediction.
We will impute missing values while accounting for uncertainty and come up with a predicted risk value of breast cancer. Although the new model will be used to predict breast cancer risk, we will examine our approach for suitability of predicting other NCDs such as obesity, diabetes and CVD. The performance of each model will be assessed mathematically and clinically.
The results may influence a substantial review of our current public health measures to prevent and diagnose breast cancer. This will help us to identify any variations in risk prediction and will allow us to commence investigating public health measures to prevent breast cancer or advise earlier screening. Also, it might initiate laboratory work to study the interplay between different factors on cellular or whole human body level.
2. Building Machine Learning Models to Predict Radiotherapy Toxicity for Cancer Patients
Radiation therapy treats many types of cancer effectively. But like other treatments, it often causes side effects. These are different for each person. They depend on the type of cancer, its location, the radiation therapy dose, and your general health.
For some people, radiation therapy causes few or no side effects. For others, the side effects are more severe. Reactions often start during the second or third week of treatment. They may last for several weeks after the final treatment.
I am leading a collaboration formed via the Radiotherapy Machine Learning Network Society from Leeds Cancer Centre, St Thomas’ Hospital Trust, Imperial College NHS Trust, University of Cambridge, Edinburgh Cancer Centre, Queen’s University Belfast, the University of Leicester and the University of Manchester. Our mission is to exploit machine learning classification techniques to predict specific side effects in cancer patients before receiving radiotherapy treatment.