Interactive segmentation of cancer tissue images

  • Swimlane: 2022-2023
  • Column: Draft
  • Position: 42
  • Assignee: Olivier Debeir
  • Creator: Olivier Debeir
  • Started:
  • Created: 15/06/2018 16:37
  • Modified: 18/05/2020 10:22
  • Moved: 18/05/2020 10:22
Description

Malignant tumors are made of multiple types of cells and heterogeneous morphological structures. This intra-tumoral heterogeneity is key to understand cancer progression and the anti-tumoral immune response.
The various morphological structures present within a tumor are readily visible under the microscope. Yet their characterization is currently resting on visual inspection, which is subjective, qualitative and time-consuming. A tool to interactively assist doctors and researchers segmentation and labeling of morphological structures within tumors microscope images would be extremely useful. It would also be a step towards ‘digital pathology’, the automated diagnostic of diseases from the images of sick tissues.

Specific morphological structures have distinctive textures that could exploitable for interactive machine learning and a dramatic speed up of tissues segmentation task (see figure).

The purpose of the thesis is to develop a prototype of computer-assisted tissue segmentation, and demonstrate how its application to a collection of cancer tissues yield quantitative insights on tumor morphology not accessible otherwise.
This interdisciplinary work will be supervised by researchers from LISA (image analysis), IRIBHM (bioinformatics and oncology, project initiator) and the pathology department of the Bordet Institute).

contacts:

Olivier Debeir, LISA (Olivier.Debeir@ulb.ac.be)
Vincent Detours, IRIBHM (vdetours@ulb.ac.be)

Sub-Tasks
Internal links
Comments