Integration of artificial intelligence (AI) and machine learning (ML) algorithms, and assessments from radiologists could help in the overall accuracy of breast cancer screenings.
The latest study by Jama Network Open, titled, “Evaluation of Combined Artificial Intelligence and Radiologist Assessment to Interpret Screening Mammograms” came up with some interesting findings. The study was focused on whether AI can tackle human mammography screening limitations with an unbiased evaluation of ML algorithms.
Breast cancer strikes one out of every eight women globally. This research is based on results from the “The Digital Mammography DREAM Challenge” by IBM Research, which was set in motion three years ago. It was a competition to capture an international scientific community for assessing whether AI and ML algorithms could meet the radiologist interpretive accuracy. IBM Research, in participation with Kaiser Permanente Washington Health Research Institute, Sage Bionetworks, and the University Of Washington School Of Medicine has confirmed how AI can assist radiologists to enhance the accuracy of mammogram screenings.
Christoph Lee, MD, MS and Professor of Radiology at the University of Washington School of Medicine and author of the paper as reported to have said in a statement – “Based on our findings, adding AI to radiologists’ interpretation could potentially prevent 500,000 unnecessary diagnostic workups each year in the United States. Robust clinical validation is necessary, however, before any AI algorithm can be adopted broadly.”
According to the American Cancer Society, about 20% of the breast cancers are missed by radiologists in mammogram screenings – and often diagnostics medical report is erroneous. Hence, combining the best-performing AI to mammography interpretation in the single-radiologist settings could provide remarkable performance improvements. It even has the potential to lower health care expenditures and addressing resource scarcity in the screening programs.
The detailed findings from the report:
“Overall, 144 231 screening mammograms from 85 580 US women (952 cancer positive 12 months from screening) were used for algorithm training and validation. A second independent validation cohort included 166 578 examinations from 68 008 Swedish women (780 cancer positive). The top-performing algorithm achieved an area under the curve of 0.858 (United States) and 0.903 (Sweden) and 66.2% (United States) and 81.2% (Sweden) specificity at the radiologists’ sensitivity, lower than community-practice radiologists’ specificity of 90.5% (United States) and 98.5% (Sweden). Combining top-performing algorithms and US radiologist assessments resulted in a higher area under the curve of 0.942 and achieved a significantly improved specificity (92.0%) at the same sensitivity.”
The study revealed that no AI algorithm could perform better than radiologists’ benchmark and combining them resulted in increased specificity. Thus, the underlying potential of machine learning methods used to enhance mammography screening interpretation needs to be taken seriously!