Medical Image Segmentation

Project: Interactive Vessel Segmentation

Despite the numerous successes of machine learning in medical image segmentation, vessel segmentation remains unsolved through ML due to the problems of small data. In collaboration with University College London and INRIA, we curretly try to define an interactive learning framework that eases the process of brain vessel segmentation.

To address this problem we propose to use CAPTCHA-style annotation with automatic generation of weak labels to train a segmenter network. The code is available in Github.

Students

Ngoc-Vien Dang
Giuseppe Di Giacomo
Viola Marconetto

Collaborators

Ferran Prados - UCL
Rosa Cortese - UCL & University of Siena
Marco Lorenzi - INRIA

References

V Dang, F Galati, et al. Vessel-CAPTCHA: an efficient learning framework for vessel annotation and segmentation Medical Image Analysis 2022