9 March 2017 Using non-specialist observers in 4AFC human observer studies
Author Affiliations +
Virtual clinical trials (VCTs) are an emergent approach for rapid evaluation and comparison of various breast imaging technologies and techniques using computer-based modeling tools. Increasingly 4AFC (Four alternative forced choice) virtual clinical trials are used to compare detection performances of different breast imaging modalities. Most prior studies have used physicists and/or radiologists and physicists interchangeably. However, large scale use of statistically significant 4AFC observer studies is challenged by the individual time commitment and cost of such observers, often drawn from a limited local pool of specialists. This work aims to investigate whether non-specialist observers can be used to supplement such studies. A team of five specialist observers (medical physicists) and five non-specialists participated in a 4AFC study containing simulated 2D-mammography and DBT (digital breast tomosynthesis) images, produced using the OPTIMAM toolbox for VCTs. The images contained 4mm irregular solid masses and 4mm spherical targets at a range of contrast levels embedded in a realistic breast phantom background. There was no statistically significant difference between the detection performance of medical physicists and non-specialists (p>0.05). However, non-specialists took longer to complete the study than their physicist counterparts, which was statistically significant (p<0.05). Overall, the results from both observer groups indicate that DBT has a lower detectable threshold contrast than 2D-mammography for both masses and spheres, and both groups found spheres easier to detect than irregular solid masses.
© (2017) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Premkumar Elangovan, Alistair Mackenzie, David R. Dance, Kenneth C. Young, Kevin Wells, "Using non-specialist observers in 4AFC human observer studies", Proc. SPIE 10132, Medical Imaging 2017: Physics of Medical Imaging, 1013256 (9 March 2017); doi: 10.1117/12.2255560; https://doi.org/10.1117/12.2255560

Back to Top