Xueping Jing1, Mirjam Wielema2, Ludo J Cornelissen3, Margo van Gent2, Willie M Iwema4, Sunyi Zheng5, Paul E Sijens2, Matthijs Oudkerk6, Monique D Dorrius2, Peter M A van Ooijen5. 1. Department of Radiation Oncology, and Data Science Center in Health (DASH), Machine Learning Lab, University Medical Center Groningen, University of Groningen, Hanzeplein 1, 9713, GZ, Groningen, The Netherlands. x.jing@umcg.nl. 2. Department of Radiology, University Medical Center Groningen, University of Groningen, Hanzeplein 1, 9713, GZ, Groningen, The Netherlands. 3. Department of Radiation Oncology, University Medical Center Groningen, University of Groningen, Hanzeplein 1, 9713, GZ, Groningen, The Netherlands. 4. Faculty of Medical Sciences, University of Groningen, Antonius Deusinglaan 1, 9713, AV, Groningen, The Netherlands. 5. Department of Radiation Oncology, and Data Science Center in Health (DASH), Machine Learning Lab, University Medical Center Groningen, University of Groningen, Hanzeplein 1, 9713, GZ, Groningen, The Netherlands. 6. Faculty of Medical Sciences, University of Groningen and Institute of Diagnostic Accuracy, Wiersmastraat 5, 9713, GH, Groningen, The Netherlands.
Abstract
OBJECTIVES: To investigate the feasibility of automatically identifying normal scans in ultrafast breast MRI with artificial intelligence (AI) to increase efficiency and reduce workload. METHODS: In this retrospective analysis, 837 breast MRI examinations performed on 438 women from April 2016 to October 2019 were included. The left and right breasts in each examination were labelled normal (without suspicious lesions) or abnormal (with suspicious lesions) based on final interpretation. Maximum intensity projection (MIP) images of each breast were then used to train a deep learning model. A high sensitivity threshold was calculated based on the detection trade - off (DET) curve on the validation set. The performance of the model was evaluated by receiver operating characteristic analysis of the independent test set. The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) with the high sensitivity threshold were calculated. RESULTS: The independent test set consisted of 178 examinations of 149 patients (mean age, 44 years ± 14 [standard deviation]). The trained model achieved an AUC of 0.81 (95% CI: 0.75-0.88) on the independent test set. Applying a threshold of 0.25 yielded a sensitivity of 98% (95% CI: 90%; 100%), an NPV of 98% (95% CI: 89%; 100%), a workload reduction of 15.7%, and a scan time reduction of 16.6%. CONCLUSION: This deep learning model has a high potential to help identify normal scans in ultrafast breast MRI and thereby reduce radiologists' workload and scan time. KEY POINTS: • Deep learning in TWIST may eliminate the necessity of additional sequences for identifying normal breasts during MRI screening. • Workload and scanning time reductions of 15.7% and 16.6%, respectively, could be achieved with the cost of 1 (1 of 55) false negative prediction.
OBJECTIVES: To investigate the feasibility of automatically identifying normal scans in ultrafast breast MRI with artificial intelligence (AI) to increase efficiency and reduce workload. METHODS: In this retrospective analysis, 837 breast MRI examinations performed on 438 women from April 2016 to October 2019 were included. The left and right breasts in each examination were labelled normal (without suspicious lesions) or abnormal (with suspicious lesions) based on final interpretation. Maximum intensity projection (MIP) images of each breast were then used to train a deep learning model. A high sensitivity threshold was calculated based on the detection trade - off (DET) curve on the validation set. The performance of the model was evaluated by receiver operating characteristic analysis of the independent test set. The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) with the high sensitivity threshold were calculated. RESULTS: The independent test set consisted of 178 examinations of 149 patients (mean age, 44 years ± 14 [standard deviation]). The trained model achieved an AUC of 0.81 (95% CI: 0.75-0.88) on the independent test set. Applying a threshold of 0.25 yielded a sensitivity of 98% (95% CI: 90%; 100%), an NPV of 98% (95% CI: 89%; 100%), a workload reduction of 15.7%, and a scan time reduction of 16.6%. CONCLUSION: This deep learning model has a high potential to help identify normal scans in ultrafast breast MRI and thereby reduce radiologists' workload and scan time. KEY POINTS: • Deep learning in TWIST may eliminate the necessity of additional sequences for identifying normal breasts during MRI screening. • Workload and scanning time reductions of 15.7% and 16.6%, respectively, could be achieved with the cost of 1 (1 of 55) false negative prediction.
Authors: Roel D Mus; Cristina Borelli; Peter Bult; Elisabeth Weiland; Nico Karssemeijer; Jelle O Barentsz; Albert Gubern-Mérida; Bram Platel; Ritse M Mann Journal: Eur J Radiol Date: 2017-01-20 Impact factor: 3.528
Authors: Christiane K Kuhl; Simone Schrading; Kevin Strobel; Hans H Schild; Ralf-Dieter Hilgers; Heribert B Bieling Journal: J Clin Oncol Date: 2014-06-23 Impact factor: 44.544
Authors: Marije F Bakker; Stéphanie V de Lange; Ruud M Pijnappel; Ritse M Mann; Petra H M Peeters; Evelyn M Monninkhof; Marleen J Emaus; Claudette E Loo; Robertus H C Bisschops; Marc B I Lobbes; Matthijn D F de Jong; Katya M Duvivier; Jeroen Veltman; Nico Karssemeijer; Harry J de Koning; Paul J van Diest; Willem P T M Mali; Maurice A A J van den Bosch; Wouter B Veldhuis; Carla H van Gils Journal: N Engl J Med Date: 2019-11-28 Impact factor: 91.245
Authors: Dorothy A Sippo; Kristine S Burk; Sarah F Mercaldo; Geoffrey M Rutledge; Christine Edmonds; Zoe Guan; Kevin S Hughes; Constance D Lehman Journal: Radiology Date: 2019-05-07 Impact factor: 11.105
Authors: Ritse M Mann; Roel D Mus; Jan van Zelst; Christian Geppert; Nico Karssemeijer; Bram Platel Journal: Invest Radiol Date: 2014-09 Impact factor: 6.016
Authors: Doris Leithner; Linda Moy; Elizabeth A Morris; Maria A Marino; Thomas H Helbich; Katja Pinker Journal: J Magn Reson Imaging Date: 2018-09-08 Impact factor: 4.813