Estimating left ventricular volume with ROI-based convolutional neural network

The volume of the human left ventricular (LV) chamber is an important indicator for diagnosing heart disease. Although LV volume can be measured manually with cardiac magnetic resonance imaging (MRI), the process is difficult and time-consuming for experienced cardiologists. This paper presents an end-to-end segmentation-free method that estimates LV volume from MRI images directly. The method initially uses Fourier transform and a regression lter to calculate the region of interest that contains the LV chambers. Then convolutional neural networks are trained to estimate the end-diastolic volume (EDV) and end-systolic volume (ESV). The resulting models accurately estimate the EDV and ESV with a mean absolute error of 15.83 and 9.82 mL, respectively, and an ejection fraction with root mean square error of 5.56%. The comparison results show that the direct estimation methods possess attractive advantages over the previous segmentation-based estimation methods.

PDF

___

[1] Petitjean C, Dacher JN. A review of segmentation methods in short axis cardiac MR images. Med Image Anal 2011; 15: 169-184.
[2] Ayed BI, Li S, Ross I. Embedding overlap priors in variational left ventricle tracking. IEEE T Med Imaging 2009; 28: 1902-1913.
[3] Ayed BI, Punithakumar K, Li S, Islam A, Chong J. Left ventricle segmentation via graph cut distribution matching. In: ACM 2009 Medical Image Computing and Computer-Assisted Intervention; 20{24 September 2009; London, UK. New York, NY, USA: ACM. pp. 901-909.
[4] Afshin M, Ayed BI, Punithakumar K, Law M, Islam A, Goela A, Peters T, Li S. Regional assessment of cardiac left ventricular myocardial function via MRI statistical features. IEEE T Med Imaging 2014; 33: 481-494.
[5] Wang Z, Salah BM, Gu B, Islam A, Goela A, Li S. Direct estimation of cardiac bi-ventricular volumes with an adapted Bayesian formulation. IEEE T Bio-Med Eng 2014; 61: 1251-1260.
[6] Zhen X, Wang Z, Islam A, Bhaduri M, Chan I, Li S. Multi-scale deep networks and regression forests for direct bi-ventricular volume estimation. Med Image Anal 2016; 30: 120-129.
[7] Kabani AW, El-Sakka MR. Estimating ejection fraction and left ventricle volume using deep convolutional networks. Lect Notes Comp Sci 2016; 9730: 678-686.
[8] Krizhevsky A, Sutskever I, Hinton GE. ImageNet classi cation with deep convolutional neural networks. In: NIPS 2012 Advances in Neural Information Processing Systems Conference; 3{8 December 2012; Lake Tahoe, NV, USA. La Jolla, CA, USA: NIPS. pp. 1097-1105.
[9] Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition. arXiv preprint 2014; 1409.1556.
[10] Szegedy C, Liu W, Jia YQ, Sermanet P, Reed S, Anguelov D, Erhan D, Vanhoucke Vet, Rabinovich A. Going deeper with convolutions. In: IEEE 2015 Conference on Computer Vision and Pattern Recognition; 7{12 June 2015; Boston, MA, USA. New York, NY, USA: IEEE. pp. 1-9.
[11] Lin X, Cowan B, Young A. Automated detection of left ventricle in 4D MR images: experience from a large study. In: ACM 2006 Medical Image Computing and Computer-Assisted Intervention Conference; 1{6 October 2006; Copenhagen, Denmark. New York, NY, USA: ACM. pp. 728-735.
[12] LeCun Y, Bengio Y, Hinton G. Deep learning. Nature 2015; 521: 436-444.
[13] Kingma D, Ba J. Adam: A method for stochastic optimization. arXiv preprint 2014; 1412.6980.
[14] Srivastava N, Hinton GE, Krizhevsky A, Sutskever I, Salakhutdinov R. Dropout: a simple way to prevent neural networks from over tting. J Mach Learn Res 2014; 15: 1929-1958.
[15] Gal Y, Ghahramani Z. Dropout as a Bayesian approximation: Representation model uncertainty in deep learning. In: ACM 2016 International Conference on Machine Learning; 19{24 June 2016; New York, NY, USA. New York, NY, USA: ACM. pp. 1050-1059.