Machine learning to assist radiologists in BreastScreen Norway

-          A research project aimed at investigating advantages and disadvantages of integrating a machine learning system for automatic interpretation of screening mammograms in BreastScreen Norway
Last updated: 8/6/2020

Background

BreastScreen Norway is the national population-based screening program against breast cancer in Norway, offered for women aged 50 to 69 every other year. The goal of the program is to find breast cancer at an early stage so that fewer women die from the disease.

Breast radiologists working in BreastScreen Norway spend a substantial amount of time on screen-reading mammograms of healthy women.

With new advances within machine learning, there is a potential to improve today`s breast cancer screening program by integrating a machine learning system that automatically interprets screening mammograms. By doing this, the breast radiologists can potentially spend less time on healthy women and more time focusing on women with breast symptoms or in other areas of the health care system.

Objective

The aim of the study is to explore the advantages and disadvantages of implementing a machine learning system for automatically interpreting screening mammograms in BreastScreen Norway. The long-term goal is to offer women attending screening an improved offer by integrating machine learning in the interpretation process.

The machine learning system to be tested, Transpara, is developed by ScreenPoint Medical B.V. and is FDA approved. The study is a retrospective cohort study that will compare the performance of Transpara with independent radiologists interpretation score from BreastScreen Norway. The study consist of five sub-studies, with data from various breast centers in Norway.

The following research question will be explored: “Can the machine learning system be used to reduce workload, increase the efficiency and supplement the work of radiologists associated with mammographic screening, without decreasing the sensitivity of the program or increasing the rate of false positive screening-examinations?”

Data

The project will use data from screening examinations that have already been conducted in BreastScreen Norway, consisting of mammograms (image data) and information on radiological assessments, including any positive and negative findings from the screening program (screening data). In addition, we will use Transpara for reinterpreting the mammograms and providing an automatic interpretation score indicating the likelihood of cancer.

The project will include data from women screened at various breast centers in BreastScreen Norway. We will include image data and screening data from all breast cancer cases, and from a control group with negative screening results. This is equivalent to about 135 000 screening examinations of about 120 000 women.

The project will only use data from women who have allowed that their personal data related to negative screening results be permanently stored at the Cancer Registry, in accordance with the Cancer Registry Regulations (Kreftregisterforskriften). It will not be possible to identify individuals from any published study results.

Organisation

The Cancer Registry is the project leader and responsible for obtaining ethical approcal(s) for the project, data collection, using the machine learning system Transpara, as well as analyzing results.

ScreenPoint Medical is the company that has developed Transpara. They will contribute with installation of Transpara, as well as training for using the system.

The breast centers are the regional specialists on breast cancer screening and diagnosis. They will contribute with radiological expertise and practical knowledge about screening, as well as facilitating the delivery of image data to the project.