Master’s student position

Context integration in whole-slide histopathology images

Ref. 2019-28

Job Description

In the IBM Research Computational Pathology team, one of our main points of focus is analyzing digitized histopathology images.

To perform a sure test for cancer diagnosis, tissues biopsies are sampled from the organ and stained to find tumor aggressiveness. Pathologists analyze and estimate the stage and grade of cancer for diagnosis and prognosis purposes by looking at the tissue samples under a microscope, zooming in and out to identify different features responsible for grading the cancer. However, this is a tedious manual and time-consuming task. Therefore the tissues samples are digitized to have the potential to be shared (tele-pathology) and analyzed using machine-learning algorithms.

The digital form of an image, called the whole slide image (WSI), is such that one has access to different resolutions of the image to be able zoom in and out as pathologists do. A WSI is a gigapixel-sized image and therefore difficult to be processed all at once. In the literature, most deep-learning based methods focus on extracting smaller regions or patches from a specific resolution. However, this leads to a loss of context information and limits the use of only a specific resolution. It was recently proved that including context from other resolutions and from patches surrounding the region helps improve performance.

[ References ]

K. Sirinukunwattanar /> “Improving Whole Slide Segmentation Through Visual Context-A Systematic Study
International Conference on Medical Image Computing and Computer-Assisted Intervention. Springer, Cham, 2018.

B. Kong et al.
Cancer metastasis detection via spatially structured deep network
International Conference on Information Processing in Medical Imaging. Springer, Cham, 2017.

A. BenTaieb, H. Ghassan
Predicting Cancer with a Recurrent Visual Attention Model for Histopathology Images
International Conference on Medical Image Computing and Computer-Assisted Intervention. Springer, Cham, 2018.

Goals of the Master’s thesis:

  • Extensive literature review of existing approaches to include context
  • If code is not provided, re-implement some of the existing methods
  • Develop a framework using reinforcement learning or graph-based approaches—or any other approaches—to identify which combinations of resolutions and surrounding regions make sense specific to a task and compare the results with existing methods
  • Understand the results to try to find a reasoning to how the new method learns to combine information from different levels and if it corresponds to how pathologists analyze the images.

Requirements

  • Working knowledge of Python
  • Strong theoretical understanding of classical computer vision techniques and machine-learning architectures
  • Practical experience with deep-learning frameworks
  • A background in either histopathology, biology, bioinformatics or healthcare sciences could be beneficial.

Diversity

IBM is committed to diversity at the workplace. With us you will find an open, multicultural environment. Excellent, flexible working arrangements enable both women and men to strike the desired balance between their professional development and their personal lives.

How to apply

If you are interested, please send your application to (fra@zurich.ibm.com) including a brief description, motivation and CV.

Preferable starting date: August – October 2019.

Note: The position is only for Master’ thesis students. Internships are not available at the moment.