P D Chang1,2, E Kuoy1, J Grinband3, B D Weinberg4, M Thompson1, R Homo1, J Chen5, H Abcede6, M Shafie6, L Sugrue2, C G Filippi7, M-Y Su1, W Yu6, C Hess2, D Chow8. 1. From the Departments of Radiology (P.D.C., E.K., M.T., R.H., M.-Y.S., D.C.). 2. Departments of Radiology (P.D.C., L.S., C.H.), University of California, San Francisco, California. 3. Department of Radiology (J.G.), Columbia University, New York, New York. 4. Department of Radiology (B.D.W.), Emory University School of Medicine, Atlanta, Georgia. 5. Neurosurgery (J.C.). 6. Neurology (H.A., M.S., W.Y.), University of California Irvine. 7. Department of Radiology (C.G.F.), North Shore University Hospital, Long Island, New York. 8. From the Departments of Radiology (P.D.C., E.K., M.T., R.H., M.-Y.S., D.C.) chowd3@uci.edu.
Abstract
BACKGROUND AND PURPOSE: Convolutional neural networks are a powerful technology for image recognition. This study evaluates a convolutional neural network optimized for the detection and quantification of intraparenchymal, epidural/subdural, and subarachnoid hemorrhages on noncontrast CT. MATERIALS AND METHODS: This study was performed in 2 phases. First, a training cohort of all NCCTs acquired at a single institution between January 1, 2017, and July 31, 2017, was used to develop and cross-validate a custom hybrid 3D/2D mask ROI-based convolutional neural network architecture for hemorrhage evaluation. Second, the trained network was applied prospectively to all NCCTs ordered from the emergency department between February 1, 2018, and February 28, 2018, in an automated inference pipeline. Hemorrhage-detection accuracy, area under the curve, sensitivity, specificity, positive predictive value, and negative predictive value were assessed for full and balanced datasets and were further stratified by hemorrhage type and size. Quantification was assessed by the Dice score coefficient and the Pearson correlation. RESULTS: A 10,159-examination training cohort (512,598 images; 901/8.1% hemorrhages) and an 862-examination test cohort (23,668 images; 82/12% hemorrhages) were used in this study. Accuracy, area under the curve, sensitivity, specificity, positive predictive value, and negative-predictive value for hemorrhage detection were 0.975, 0.983, 0.971, 0.975, 0.793, and 0.997 on training cohort cross-validation and 0.970, 0.981, 0.951, 0.973, 0.829, and 0.993 for the prospective test set. Dice scores for intraparenchymal hemorrhage, epidural/subdural hemorrhage, and SAH were 0.931, 0.863, and 0.772, respectively. CONCLUSIONS: A customized deep learning tool is accurate in the detection and quantification of hemorrhage on NCCT. Demonstrated high performance on prospective NCCTs ordered from the emergency department suggests the clinical viability of the proposed deep learning tool.
BACKGROUND AND PURPOSE: Convolutional neural networks are a powerful technology for image recognition. This study evaluates a convolutional neural network optimized for the detection and quantification of intraparenchymal, epidural/subdural, and subarachnoid hemorrhages on noncontrast CT. MATERIALS AND METHODS: This study was performed in 2 phases. First, a training cohort of all NCCTs acquired at a single institution between January 1, 2017, and July 31, 2017, was used to develop and cross-validate a custom hybrid 3D/2D mask ROI-based convolutional neural network architecture for hemorrhage evaluation. Second, the trained network was applied prospectively to all NCCTs ordered from the emergency department between February 1, 2018, and February 28, 2018, in an automated inference pipeline. Hemorrhage-detection accuracy, area under the curve, sensitivity, specificity, positive predictive value, and negative predictive value were assessed for full and balanced datasets and were further stratified by hemorrhage type and size. Quantification was assessed by the Dice score coefficient and the Pearson correlation. RESULTS: A 10,159-examination training cohort (512,598 images; 901/8.1% hemorrhages) and an 862-examination test cohort (23,668 images; 82/12% hemorrhages) were used in this study. Accuracy, area under the curve, sensitivity, specificity, positive predictive value, and negative-predictive value for hemorrhage detection were 0.975, 0.983, 0.971, 0.975, 0.793, and 0.997 on training cohort cross-validation and 0.970, 0.981, 0.951, 0.973, 0.829, and 0.993 for the prospective test set. Dice scores for intraparenchymal hemorrhage, epidural/subdural hemorrhage, and SAH were 0.931, 0.863, and 0.772, respectively. CONCLUSIONS: A customized deep learning tool is accurate in the detection and quantification of hemorrhage on NCCT. Demonstrated high performance on prospective NCCTs ordered from the emergency department suggests the clinical viability of the proposed deep learning tool.
Authors: S M Davis; J Broderick; M Hennerici; N C Brun; M N Diringer; S A Mayer; K Begtrup; T Steiner Journal: Neurology Date: 2006-04-25 Impact factor: 9.910
Authors: Moritz Scherer; Jonas Cordes; Alexander Younsi; Yasemin-Aylin Sahin; Michael Götz; Markus Möhlenbruch; Christian Stock; Julian Bösel; Andreas Unterberg; Klaus Maier-Hein; Berk Orakcioglu Journal: Stroke Date: 2016-10-04 Impact factor: 7.914
Authors: K N Bhanu Prakash; Shi Zhou; Tim C Morgan; Daniel F Hanley; Wieslaw L Nowinski Journal: Int J Comput Assist Radiol Surg Date: 2012-09 Impact factor: 2.924
Authors: M Ross Bullock; Randall Chesnut; Jamshid Ghajar; David Gordon; Roger Hartl; David W Newell; Franco Servadei; Beverly C Walters; Jack E Wilberger Journal: Neurosurgery Date: 2006-03 Impact factor: 4.654
Authors: Charlotte Jj van Asch; Merel Ja Luitse; Gabriël Je Rinkel; Ingeborg van der Tweel; Ale Algra; Catharina Jm Klijn Journal: Lancet Neurol Date: 2010-01-05 Impact factor: 44.182
Authors: Y W Lui; P D Chang; G Zaharchuk; D P Barboriak; A E Flanders; M Wintermark; C P Hess; C G Filippi Journal: AJNR Am J Neuroradiol Date: 2020-07-30 Impact factor: 3.825
Authors: Samuel W Remedios; Snehashis Roy; Camilo Bermudez; Mayur B Patel; John A Butman; Bennett A Landman; Dzung L Pham Journal: Med Phys Date: 2019-11-19 Impact factor: 4.071
Authors: Robert J Harris; Shwan Kim; Jerry Lohr; Steve Towey; Zeljko Velichkovich; Tim Kabachenko; Ian Driscoll; Brian Baker Journal: J Digit Imaging Date: 2019-12 Impact factor: 4.056
Authors: Andreas M Rauschecker; Jeffrey D Rudie; Long Xie; Jiancong Wang; Michael Tran Duong; Emmanuel J Botzolakis; Asha M Kovalovich; John Egan; Tessa C Cook; R Nick Bryan; Ilya M Nasrallah; Suyash Mohan; James C Gee Journal: Radiology Date: 2020-04-07 Impact factor: 11.105
Authors: Sarah M Hooper; Jared A Dunnmon; Matthew P Lungren; Domenico Mastrodicasa; Daniel L Rubin; Christopher Ré; Adam Wang; Bhavik N Patel Journal: Radiol Artif Intell Date: 2021-04-28