Imaging & Mathematics Network

The Centre for Mathematical Imaging in Healthcare (CMIH) and Cancer Research UK Cambridge Institute (CRUK-CI) have created a new network to bring together researchers in imaging and mathematics from across Cambridge. Talks and discussions will include topics including imaging technologies, tools and methodologies, open problems and challenges associated with large volume or complex forms of imaging data.

Network meetings will take place once a term and will include talks from invited speakers. We encourage all attendees to come with questions and to actively participate in the discussion and networking session afterward. Our intention is to bring imaging experts/users/researchers together to connect, share knowledge and to develop potential collaborations.

The next meeting of the Imaging and Mathematics Network will take place on Tuesday 3rd March from 3-4:30 pm in the Lecture Theatre at Cancer Research UK Cambridge Institute.

The speakers for this meeting are:

Dr Angelica Aviles-Rivero, Centre for Mathematical Imaging in Healthcare (CMIH)

Title: When Labelling Hurts: Learning to Classify Large-Scale Medical Data with Minimal Supervision

Abstract: The task of classifying when scarse number of labels are available is of great interest in real-world problems. Asthonising results in classification has been achieved by fully-supervised techniques. However, in the medical domain, this might be a strong assumption for a solution, as annotated data contains strong human bias. In this talk, we motivate the use of semi-supervised learning for classifying large-scale data. We also introduce our recent works in this area where we introduce the concept of hybrid models. In particular, we present a novel optimisation model, and also show that our new functional can be connected to deep nets. We demonstrate, through numerical and visual results, that our proposed approach can reach great accuracy to such unprecedented amount of unlabelled data.

Dr Carlos Gonzalez Fernandez and Dr Eduardo Gonzalez Solares, Institute of Astronomy 

Title: Image processing at the petabyte scale

Abstract: In this talk we’ll talk about the problems a large yield project like IMAXT faces, and we’ll showcase some of the solutions we are implementing for issues like image processing and visualisation for larger-than-memory datasets, multi-user concurrent access to data, parallel and distributed image processing pipelines and their deployment, and archive curation.

It’s free to attend but please register here so we have enough tea and cake!