Deep learning for early detection of cancer recurrence in patients with glioblastoma through imaging

Brain tumours are cancer of unmet need and are the most common cause of cancer death in under the 40s. Glioblastoma is the most common primary adult brain tumour and carries one of the worst prognoses amongst human cancers, with a median survival time of about 15 months.

The fundamental question to be answered in this project is: can deep learning be used for early detection and prediction of glioblastoma recurrence through imaging? To this end, we will extend the state-of-the-art using Bayesian recurrent variational auto-encoders (VAE) that will be conditioned on the patient meta-data.

An LSTM-RNN will be trained to approximate the predictive distribution of the next set of MR images, given current images and patient meta-data. We will devise an end-to-end training mechanism that will jointly learn the encoding-decoding maps along with the predictions of the spatiotemporal maps.