Anatomical Segmentation of Prostate MR Images

Mihail Bors


Supervised by Frank C Langbein; Moderated by Jing Wu

The aim of this project is to investigate approaches towards automatically segmenting the prostate in MRI datasets, based on the PROMISE12 grand challenge and our own, internal data set (other data sets are also available). This is part of a research project for early-stage prostate cancer detection. You may develop your own approach or test and then extent already published approaches. Which technique you are using is your choice, but it is likely that a deep learning approach (e.g. U-NET and variants thereof) will perform well.

You must have excellent mathematical and programming skills for this project, an understanding of image processing and related segmentation or machine learning algorithms. Algorithms should be implemented in Python. Access to a reasonably powerful GPU with enough memory is required; this is available via the Linux lab and SCW. To be useful for our applications, we need any developed software to run under Linux, and it must be released under the GNU AGPL v3 or a compatible licence.

Initial Plan (06/02/2023) [Zip Archive]

Final Report (19/05/2023) [Zip Archive]

Publication Form