Cardiac substructure segmentation with deep learning for improved cardiac sparing

Medical Physics - Tập 47 Số 2 - Trang 576-586 - 2020
Eric D. Morris1,2, A.I. Ghanem3,1, Ming Dong4, Milan Pantelic5, Eleanor M. Walker1, Carri Glide‐Hurst1,2
1Department of Radiation Oncology, Henry Ford Cancer Institute, Detroit, MI, USA
2Department of Radiation Oncology, Wayne State University School of Medicine, Detroit, MI, USA
3Department of Clinical Oncology, Alexandria University, Alexandria, Egypt
4Department of Computer Science, Wayne State University, Detroit, MI, USA
5Department of Radiology, Henry Ford Cancer Institute, Detroit, MI, USA

Tóm tắt

Purpose

Radiation dose to cardiac substructures is related to radiation‐induced heart disease. However, substructures are not considered in radiation therapy planning (RTP) due to poor visualization on CT. Therefore, we developed a novel deep learning (DL) pipeline leveraging MRI’s soft tissue contrast coupled with CT for state‐of‐the‐art cardiac substructure segmentation requiring a single, non‐contrast CT input.

Materials/methods

Thirty‐two left‐sided whole‐breast cancer patients underwent cardiac T2 MRI and CT‐simulation. A rigid cardiac‐confined MR/CT registration enabled ground truth delineations of 12 substructures (chambers, great vessels (GVs), coronary arteries (CAs), etc.). Paired MRI/CT data (25 patients) were placed into separate image channels to train a three‐dimensional (3D) neural network using the entire 3D image. Deep supervision and a Dice‐weighted multi‐class loss function were applied. Results were assessed pre/post augmentation and post‐processing (3D conditional random field (CRF)). Results for 11 test CTs (seven unique patients) were compared to ground truth and a multi‐atlas method (MA) via Dice similarity coefficient (DSC), mean distance to agreement (MDA), and Wilcoxon signed‐ranks tests. Three physicians evaluated clinical acceptance via consensus scoring (5‐point scale).

Results

The model stabilized in ~19 h (200 epochs, training error <0.001). Augmentation and CRF increased DSC 5.0 ± 7.9% and 1.2 ± 2.5%, across substructures, respectively. DL provided accurate segmentations for chambers (DSC = 0.88 ± 0.03), GVs (DSC = 0.85 ± 0.03), and pulmonary veins (DSC = 0.77 ± 0.04). Combined DSC for CAs was 0.50 ± 0.14. MDA across substructures was <2.0 mm (GV MDA = 1.24 ± 0.31 mm). No substructures had statistical volume differences (P > 0.05) to ground truth. In four cases, DL yielded left main CA contours, whereas MA segmentation failed, and provided improved consensus scores in 44/60 comparisons to MA. DL provided clinically acceptable segmentations for all graded patients for 3/4 chambers. DL contour generation took ~14 s per patient.

Conclusions

These promising results suggest DL poses major efficiency and accuracy gains for cardiac substructure segmentation offering high potential for rapid implementation into RTP for improved cardiac sparing.

Từ khóa


Tài liệu tham khảo

10.1111/j.1365-2141.2011.08713.x

10.1016/j.radonc.2014.11.037

10.1093/annonc/mdq042

10.1056/NEJMoa1209825

10.1016/S1470-2045(14)71207-0

10.1001/jamaoncol.2015.3969

10.1016/j.ijrobp.2009.04.093

10.1200/JCO.2016.69.8480

10.1016/j.ijrobp.2017.04.026

10.4061/2011/317659

10.1186/1748-717X-2-20

10.1016/j.ijrobp.2018.06.091

10.1016/j.ijrobp.2018.11.025

10.1016/j.radonc.2016.11.016

10.1016/j.radonc.2018.07.013

10.1016/j.media.2016.02.006

10.1016/j.radonc.2019.03.013

RonnebergerO FischerP BroxT.U‐net: Convolutional networks for biomedical image segmentation. International Conference on Medical image computing and computer‐assisted intervention. Springer. 2015. pp. 234–241.

10.1016/j.media.2017.07.005

10.1007/978-3-319-75541-0_20

10.1007/978-3-319-75541-0_21

ZhangZ YangL ZhengY.Translating and segmenting multimodal medical volumes with cycle‐ and shape‐consistency generative adversarial network. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.2018. pp. 9242‐9251

10.1007/978-3-030-32486-5_20

10.1016/j.media.2016.10.004

10.1109/TMI.2003.815867

10.1016/j.ijrobp.2009.10.058

Baughman DR, 2014, Neural networks in bioprocessing and chemical engineering

10.1007/978-3-319-75238-9_25

LeeC‐Y XieS GallagherP et al.Deeply‐supervised nets. Artificial intelligence and statistics. 2015;562–570.

10.1007/978-3-319-67389-9_32

Mao X, 2016, Image restoration using very deep convolutional encoder‐decoder networks with symmetric skip connections, Adv Neural Inf Process Syst, 29, 2802

10.3389/fonc.2017.00315

MilletariF NavabN AhmadiS‐A.V‐net: Fully convolutional neural networks for volumetric medical image segmentation. 3D Vision (3DV) 2016 Fourth International Conference on. IEEE. 2016. pp. 565–571.

NairAA TranTD ReiterA et al.A deep learning based alternative to beamforming ultrasound images. 2018 IEEE International Conference on Acoustics Speech and Signal Processing (ICASSP). IEEE. 2018. pp. 3359–3363.

KingmaDP BaJ.Adam: A method for stochastic optimization. arXiv preprint arXiv.2014.

10.1002/mp.13047

LongJ ShelhamerE DarrellT.Fully convolutional networks for semantic segmentation. Proceedings of the IEEE conference on computer vision and pattern recognition. 2015. pp. 3431‐3440

10.1007/978-3-319-59129-2_3

Krähenbühl P, 2011, Efficient inference in fully connected CRFs with gaussian edge potentials, Adv Neural Inf Process Syst, 24, 109

10.1007/11744085_44

10.1007/s11263-007-0109-1

ZhengS JayasumanaS Romera‐ParedesB et al.Conditional random fields as recurrent neural networks. Proceedings of the IEEE international conference on computer vision.2015. pp.1529–1537

10.1016/j.media.2017.10.002

ZhangJ NieH.A post‐processing method based on fully connected CRFs for chronic wound images segmentation and identification. NUDT.2018.

10.1016/j.radonc.2017.11.012

10.2307/1932409

10.1186/1748-717X-9-173

10.1118/1.4901409

Aljabar P, 2001, The cutting edge: delineating contours with deep learning, Mach Learn, 2005, 2013

Ben‐CohenA KlangE AmitaiMM et al.Anatomical data augmentation for cnn based pixel‐wise classification. Biomedical Imaging (ISBI 2018) 2018 IEEE 15th International Symposium on. IEEE.2018. pp.1096–1099.

10.1016/j.neucom.2018.09.013

KjerlandØ.Segmentation of coronary arteries from ct‐scans of the heart using deep learning: MS Thesis. NTNU.2017.

10.1109/TMI.2006.880587

Sudre CH, 2017, Proceedings of the MICCAI workshop on Deep Learning in Medical Image Analysis (DLMIA), 240

10.1002/mp.13221

10.1109/TMI.2016.2621185

10.1007/978-3-319-75238-9_38

10.1016/j.neucom.2018.11.103

LiuZ LiX LuoP et al.Semantic image segmentation via deep parsing network. Proceedings of the IEEE international conference on computer vision. 2015. pp. 1377–1385

TrulloR PetitjeanC RuanS et al.Segmentation of organs at risk in thoracic ct images using a sharpmask architecture and conditional random fields. 2017 IEEE 14th International Symposium on Biomedical Imaging (ISBI 2017). IEEE. 2017. pp. 1003–1006.

ZhuJ‐Y ParkT IsolaP et al.Unpaired image‐to‐image translation using cycle‐consistent adversarial networks. Proceedings of the IEEE international conference on computer vision. 2017. pp. 2223‐2232

HuangX LiY PoursaeedO et al.Stacked generative adversarial networks. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2017. pp. 5077‐5086

10.2147/OTT.S52101

10.1016/j.radonc.2016.08.006

10.1016/j.ejmp.2018.05.008