Machine Learning Algorithms for Radio-Frequency Breast Cancer Detection

The project will focus on the development of novel machine learning algorithms (classification, anomaly detection, and regression) that can provide an assessment of the uncertainty associated with decisions and can provide explanations for the decisions. The Ph.D. student will initially explore the development of Bayesian deep neural networks and non-reversible Markov Chain Monte Carlo algorithms. The project will focus on algorithmic development and theoretical characterization, but there will also be an applied aspect to the work, targeting the application of detection of breast cancer using radio-frequency (RF) measurements.

Breast cancer detection: Key to successful breast tumour elimination and treatment is early diagnosis. Some women perform self-examinations to detect lumps or changes in their breasts, but without medical training, it can be difficult to distinguish between potential tumours and normal variations in appearance and texture. Over the past eight years, our research team has been working towards the development of a wearable bra device that incorporates a networked array of RF sensors (http://www.compem.ece.mcgill.ca/breast-cancer-detection/). The goal is for women to use this in the home at monthly intervals to provide doctors with an early indicator of potential disease. We have conducted clinical trials with promising preliminary results. With regard to this application, this project will include the design of algorithms that provide a decision about the need for additional scans. These algorithms will address the longitudinal nature of the data, the class imbalance (many more healthy scans than cases with tumors), and incorporate aspects of transfer learning.

A candidate must have completed or be about to complete a M. Sc. or M. Eng. degree that focuses on machine learning or statistical inference. A good track record of publishing in top conferences and journals is a strong plus. Candidates must have strong mathematical skills and good programming skills for data analysis (Python/Matlab/R). Knowledge and experience with tools in the domain of Bayesian statistical machine learning and deep learning are important (e.g. Tensorflow/Theano/Pytorch/Stan/PyMC).

The successful candidate will join the Networks Research Lab in the Department of Electrical and Computer Engineering at McGill University. McGill is a leading research university in Canada and is consistently rated highly in world university rankings. McGill is located in the beautiful city of Montreal, a vibrant, bilingual, multicultural metropolis in the province of Quebec, Canada. Montreal has recently become a centre of artificial intelligence and machine learning research activity, with many leading international companies opening research labs in the city. There are exciting opportunities for any student working in this field.

Interested applicants should contact Prof. Mark Coates ([Email Address Removed]) by email, attaching a CV, and should consult http://www.mcgill.ca/ece/graduate for information about graduate studies in the Department of Electrical and Computer Engineering at McGill University.