The impact of simulated motion blur on lesion detection performance in Full Field Digital Mammography Ahmed K. Abdullah1,2, John Thompson2, Claire Mercer2, Rob Aspin2, Judith Kelly3, Peter Hogg2 1University of Diyala, Iraq; 2University of Salford, United Kingdom; 3Countess of Chester Hospital NHS Foundation Trust, UK
Background and Purpose Motion blur is a known phenomenon in Full Field Digital Mammography (FFDM) The impact on cancer detection performance is unknown This is the first study to investigate FFDM breast cancer detection performance with different magnitudes of simulated motion blur
Method – Simulated Motion Blur Mathematical simulation of motion blur applied globally to the image1 Cancer detection evaluated under three conditions 0 mm simulated blurring (no blurring applied) 0.7 mm simulated blurring 1.5 mm simulated blurring 1. Ma WK, Aspin R, Kelly J, Millington S, Hogg P. What is the minimum amount of simulated breast movement required for visual detection of blurring? An exploratory investigation. Br J Radiol. 2015 Aug;88(1052):20150126.
Example - Microcalcification FFDM CC view Simulated Motion Blur Applied 0 mm 0.7 mm 1.5 mm Example - Microcalcification
Method – Observer Study Free-response receiver operating characteristic study Statistic power ensured1 7 observers (mean 15±5 years’ FFDM reporting experience) 248 cases from the PROCAS dataset2 62 malignant masses 62 malignant micro-calcifications 124 normal cases wJAFROC analysis performed; test alpha set at 0.05 1. Obuchowski N. Sample size tables for receiver operating characteristic studies. Am J Roentgenol :603–8. Available from: http://www.ajronline.org/doi/abs/10.2214/ajr.175.3.1750603 2. A study looking at breast cancer risk during screening (PROCAS), http://www.cancerresearchuk.org/about-cancer/find-a-clinical-trial/a-study-looking-breast-cancer-risk-screening-procas#undefined. Accessed 31-2-2017
Results - Overall Statistically significant difference for the detection of: Masses F(2,49) = 6.01 P=0.0084 Microcalcifications: F(2,49) = 23.14 P<0.0001
Results - Masses |----------------- P = 0.012 -----------------| |--------------------------------------- P = 0.004 ---------------------------------------| P = 0.641
Results - Microcalcifications |----------------- P < 0.001 -----------------| |--------------------------------------- ---------------------------------------| P = 0.004
Conclusion Simulated motion blur has a negative and statistically significant impact on cancer detection performance This is likely to have implications for clinical practice TAKE HOME MESSAGE: If you can see blur then consider repeating the image because you might miss a cancer
Future Work Assess the impact of double reporting on cancer detection in the presence of simulated motion blur Assess the impact of regional simulated motion blur on cancer detection performance