Researchers at the University of Eastern Finland have developed an algorithm to improve mammogram density assessment, with the potential to increase screening accuracy.
The artificial intelligence (AI)-based tool, MV-Defeat, uses a deep evidential neural network approach to evaluate multiple mammogram views at the same time for mammogram density assessment, mirroring the decision-making process of radiologists.
In an article in the journal IEEE Access, the research team in Kuopio, Finland, detailed their approach to developing the algorithm, which incorporates elements of the Dempster-Shafer evidential theory and subjective logic to assess mammogram images from multiple views to provide a more comprehensive analysis.
As high breast tissue density is associated with an increased risk of breast cancer, precise and accurate interpretation of mammography is essential in screening. While breast tissue density can be estimated from mammograms, variability in radiological evaluations and an ongoing global shortage of radiologists have presented challenges for effective screening.
The MV-Defeat algorithm was developed to address these challenges; the team used data from four open-source datasets to improve the algorithm's applicability and accuracy across different populations. In testing, MV-Defeat showed improvement over current approaches: MV-Defeat yielded improved results in distinguishing between benign and malignant tumors of 31.46% and 50.78% on the DDSM and VinDr-Mammo datasets, respectively, against the current standard multiview deep-learning model. The algorithm's accuracy persisted across different datasets, displaying its ability to adapt to different patient demographics.
The team emphasized in an article on the University of Eastern Finland's website that the AI algorithm will need to be refined and validated to ensure its efficacy in real-world settings.
"To fully integrate AI like MV-DEFEAT into clinical practice, it is crucial to build trust among healthcare professionals through rigorous testing and validation. Indeed, our next steps involve further validation studies to establish MV-DEFEAT as a reliable tool for breast cancer diagnostics in Finland," said doctoral researcher and co-author Raju Gudhe.
