XClose

UCL Centre for Medical Image Computing

Home
Menu

Jin Chen - Olivier Jaubert - CMIC/WEISS joint seminar series

09 September 2020, 1:00 pm–2:00 pm

Jin Chen - Olivier Jaubert - talks as part of the CMIC/WEISS joint seminar series

Event Information

Open to

All

Organiser

cmic-seminars-request@cs.ucl.ac.uk

Chen Jin

Title - Foveation for Segmentation of Mega-pixel Histology Images

Abstract - Segmenting histology images is challenging because of the sheer size of the images with millions or even billions of pixels. Typical solutions pre-process each histology image by dividing it into patches of fixed size and/or down-sampling to meet memory constraints. Such operations incur information loss in the field-of-view (FoV) (i.e., spatial coverage) and the image resolution. The impact on segmentation performance is, however, as yet understudied. In this work, we first show under typical memory constraints (e.g., 10G GPU memory) that the trade-off between FoV and resolution considerably affects segmentation performance on histology images, and its influence also varies spatially according to local patterns in different areas. Based on this insight, we then introduce foveation module, a learnable \dataloader" which, for a given histology image, adaptively chooses the appropriate configuration (FoV/resolution trade-off) of the input patch to feed to the downstream segmentation model at each spatial location. The foveation module is jointly trained with the segmentation network to maximise the task performance. We demonstrate, on the Gleason2019 challenge dataset for histopathology segmentation, that the foveation module improves segmentation performance over the cases trained with patches of fixed FoV/resolution trade-off. Moreover, our model achieves better segmentation accuracy for the two most clinically important and ambiguous classes (Gleason Grade 3 and 4) than the top performers in the challenge by 13.1% and 7.5%, and improves on the average performance of 6 human experts by 6.5% and 7.5%.

Olivier Jaubert

Title - Multi-parametric mapping using Magnetic Resonance Fingerprinting for cardiac and liver tissue characterization

Abstract - MRI images are usually qualitative however developments towards quantitative MRI enabled the acquisition of parameter maps to quantify pathophysiologies such as fibrosis or edema, and robustly assess disease including myocardial infarction, amyloidosis or liver cirrhosis. However, quantitative MRI suffers from longer scan times than qualitative imaging. Cardiac and liver MR examinations are often difficult to carry out successfully in patient cohorts (such as pediatric, elderly and very ill patients) due to these long scan times and the high number of necessary breathholds. Magnetic Resonance Fingerprinting (MRF) was proposed recently for fast multi-parametric mapping. Early cardiac and liver MRF developments showed the potential to reduce the number of breath-holds necessary while providing co-registered T1, T2 and M0 maps in a single breath-hold. The aim of my previous works were to extend cardiac and liver MRF to map additional parameters and provide a more comprehensive multi-parametric tissue characterization from a single breath-hold scan.

In this talk, i'll quickly go over the free-running cardiac MRF and water-fat MRF techniques developped for heart and liver diagnosis during my PhD.