Can we reduce the workload of mammographic screening by automatic identification of normal exams with artificial intelligence? A feasibility study.

Authors:
Kristina Lang
Kristina Lang
Lund University
Jonas Teuwen
Jonas Teuwen
Radboud University Medical Center
Mireille Broeders
Mireille Broeders
Radboud University Nijmegen Medical Centre
Netherlands
Gisella Gennaro
Gisella Gennaro
University of Padua
Italy
Paola Clauser
Paola Clauser
University of Udine
Italy
Thomas H Helbich
Thomas H Helbich
Medical University of Vienna
Austria

Eur Radiol 2019 Apr 16. Epub 2019 Apr 16.

Department of Radiology and Nuclear Medicine, Radboud University Medical Center, PO Box 9101, 6500 HB, Nijmegen, The Netherlands.

Purpose: To study the feasibility of automatically identifying normal digital mammography (DM) exams with artificial intelligence (AI) to reduce the breast cancer screening reading workload.

Methods And Materials: A total of 2652 DM exams (653 cancer) and interpretations by 101 radiologists were gathered from nine previously performed multi-reader multi-case receiver operating characteristic (MRMC ROC) studies. An AI system was used to obtain a score between 1 and 10 for each exam, representing the likelihood of cancer present. Using all AI scores between 1 and 9 as possible thresholds, the exams were divided into groups of low- and high likelihood of cancer present. It was assumed that, under the pre-selection scenario, only the high-likelihood group would be read by radiologists, while all low-likelihood exams would be reported as normal. The area under the reader-averaged ROC curve (AUC) was calculated for the original evaluations and for the pre-selection scenarios and compared using a non-inferiority hypothesis.

Results: Setting the low/high-likelihood threshold at an AI score of 5 (high likelihood > 5) results in a trade-off of approximately halving (- 47%) the workload to be read by radiologists while excluding 7% of true-positive exams. Using an AI score of 2 as threshold yields a workload reduction of 17% while only excluding 1% of true-positive exams. Pre-selection did not change the average AUC of radiologists (inferior 95% CI > - 0.05) for any threshold except at the extreme AI score of 9.

Conclusion: It is possible to automatically pre-select exams using AI to significantly reduce the breast cancer screening reading workload.

Key Points: • There is potential to use artificial intelligence to automatically reduce the breast cancer screening reading workload by excluding exams with a low likelihood of cancer. • The exclusion of exams with the lowest likelihood of cancer in screening might not change radiologists' breast cancer detection performance. • When excluding exams with the lowest likelihood of cancer, the decrease in true-positive recalls would be balanced by a simultaneous reduction in false-positive recalls.

Download full-text PDF

Source
http://link.springer.com/10.1007/s00330-019-06186-9
Publisher Site
http://dx.doi.org/10.1007/s00330-019-06186-9DOI Listing
April 2019
7 Reads

Publication Analysis

Top Keywords

likelihood cancer
20
breast cancer
16
cancer screening
16
reduce breast
12
screening reading
12
exams
11
cancer
10
lowest likelihood
8
excluding true-positive
8
read radiologists
8
artificial intelligence
8
high likelihood
8
true-positive exams
8
excluding exams
8
exams artificial
8
exams lowest
8
likelihood
6
screening
5
points •
4
normal area
4

References

(Supplied by CrossRef)
Article in CA Cancer J Clin
RA Smith et al.
CA Cancer J Clin 2010
Article in J Med Screen
M Broeders et al.
J Med Screen 2012
Article in Lancet
Independent UK Panel on Breast Cancer Screening et al.
Lancet 2012
Article in N Engl J Med
HG Welch et al.
N Engl J Med 2016
Article in Ann Intern Med
NT Brewer et al.
Ann Intern Med 2007
Article in Radiology
N Karssemeijer et al.
Radiology 2009
Article in PLoS One
KK Evans et al.
PLoS One 2013
Article in Radiographics
PT Huynh et al.
Radiographics 1998
Article in N Engl J Med
JJ Fenton et al.
N Engl J Med 2007
Article in JAMA Intern Med
CD Lehman et al.
JAMA Intern Med 2015
Article in Med Image Anal
G Litjens et al.
Med Image Anal 2017

Similar Publications