Project 4: Observer-independent Screening of Breast Cancer

Background

Mammography screening is capable of reducing mortality caused by breast cancer, however within recent years the efficiency of screening programs has strongly been criticized in Switzerland mostly due to an unfavorable cost-benefit relationship. In this project, we aim to substantially reduce the costs of mammography screening by development of a software product for automatic assessment of digital mammographies using techniques of Artificial Intelligence (AI, deep artificial neural networks, convolutional neural networks). The intended software tool will not only detect a potential tumor on mammographies, classify the finding according to its BIRADS category, write a standardized radiological report but will also provide recommendations for recall or follow-up examinations if required.

Working hypothesis

  1. Using techniques of AI, the complete radiological task of assessing mammographies can be carried out in a standardized observer-independent way, thereby providing a valuable support system to the radiologist.
  2. A standardized assessment of mammographies will result in a lower number of recalls, unnecessary biopsies, and overlooked cancers, thereby lowering healthcare costs and improving patient care.

Specific aims for this research project

The aim is to develop an AI software tool capable to

  1. automatically evaluate mammographies,
  2. classify the findings according to its BIRADS category,
  3. write a standardized radiological report,
  4. provide recommendations for recall or follow-up examinations if required.

 

Breast image