Research Topic: The ARCANE Study: Assessment of lymph node metastasis in Rectal Cancer pAtients using machiNe lEarning
Supervisors: Dr Aaron Quyn, Prof David Jayne, Dr Samuel Relton
About Ben: I graduated from the University of Leeds with an integrated master’s in Mathematics and Statistics in 2021. For my undergrad dissertation project I used Convolutional Neural Networks (CNNs) to classify COVID-19 from chest X-rays and for the master’s project I used Recurrent Neural Networks (RNNs) for sentiment analysis of film reviews. This year I will be the joint student representative for the CDT with Lewis.
Project description: Surgery is the cornerstone of rectal cancer management however it is associated with complications and long-term functional consequences. The key to avoiding unnecessary surgery is better prediction of lymph node involvement (metastasis) as clinical tools are currently around 50% accurate, and better prediction of how a patient will respond to treatments such as radiotherapy and chemotherapy. This study will use MRI and CT scans pre- and post-treatment, patient information and tumour characteristics. The computational methods I am using are primarily based around Variational Autoencoders (VAEs) along with methods for object detection and segmentation (of lymph nodes/tumours in medical images) and sequential analysis using RNNs (for patient response to treatment). I aim to use VAEs to perform latent space manipulation of features of the tumours and lymph nodes to infer shape/texture that influences predictions of metastasis or treatment response.