Deep Learning for the differentiation of STIC lesions

This is a student project vacancy.

Deep Learning for the differentiation of STIC lesions

Background

The fallopian tubes are two appendages, connected on either side of the uterus. They form the passageway to the ovaries and enable the egg cells to travel to the uterus. It is thought that in the epithelial tissue covering the inside of the fallopian tubes, pre-cancerous lesions form. These lesions, known as serous tubal intraepithelial carcinoma (STIC) are considered pre-cancerous lesions and may result in ovarian or pelvic carcinoma.

Research question: Is it possible to differentiate STIC lesions, from normal epithelium?

Tasks

The goal is to develop a deep learning algorithm for the differentiation of STIC lesions from normal fallopian tube epithelium. The digital pathology images of a cohort of both STIC lesions and normal tubal epithelium from BRCA1/2 carries will be available. The output will be an algorithm that can identify aberrant tubal epithelium with a high sensitivity.

Clinical relevance

Ovarian carcinoma, which is assumed to form out of STIC lesions in the fallopian tube, has the 7th highest incidence of carcinoma amongst woman in the Netherlands and has a poor prognosis, partly because it is often diagnosed at a late stage. Until now there are no successful screening methods for this type of cancer and therefore woman who are at high risk of developing ovarian cancer, such as BRCA1/2 mutation carriers and woman with a positive family history for ovarian or breast cancer are counseled on what is called a ‘Risk Reducing Salpingo Oophorectomy’ or RRSO. This is an operation in which the ovaries and fallopian tubes are resected. This decreases the risk of carcinoma, but has several negative side effects related to a premature menopause, which is induced by removing the ovaries. These negative side effects include a decreased quality of life, osteoporosis and an increased risk of cardiovascular disease. The TUBA study is a Dutch multicentre study examining an alternative surgical approach, which consists of initially only removing the fallopian tubes (where the potential STIC lesions form) and removing the ovaries at a later stage. By this approach, premature menopause is delayed, thus improving the quality of life for these woman. In this setting it becomes especially important to successfully identify pre-cancerous lesions, as woman with STIC lesions are advised to undergo an ovariectomy straight away, to ensure any potential early spread of (pre-) cancerous cells is removed.

Requirements

  • Students with a major in computer science, biomedical engineering, artificial intelligence, physics, or a related area in the final stage of master level studies are invited to apply.
  • Interest in image analysis and machine learning
  • Affinity with programming in Python is required.

Information

  • Project duration: 6 months
  • Location: Radboud University Nijmegen Medical Centre
  • For more information please contact Jeroen van der Laak

People

Jeroen van der Laak

Jeroen van der Laak

Associate professor/Group leader