Assessment of optimal deep learning configuration for vertebrae segmentation from CT images

Automated segmentation of vertebral bone from diagnostic Computed Tomography (CT) images has become an important part of clinical workflow today. There is an increasing need for computer aided diagnosis applications of various spine disorders including scoliosis, fracture detection and even automated reporting. While modelbased methods have been widely used, recent deep Learning methods have shown a great potential in this area. However, choice of optimal configuration of the network to get the best segmentation performance is challenging. In this work, we explore the impact of different training and inference options, including dimensions, activation function, batch normalization, kernel size, filters, patch size and patch selection strategy in U-Net architecture. 20 publicly available CT Spine datasets from Spineweb repository was used in this study divided into training/test datasets. Training with different DL configurations were repeated with these datasets. We used the best weights corresponding to each configuration for inference on the independent test dataset. These results on the test dataset with the best weights for each configurations were compared. 3D models performed consistently better than 2D approaches. Overlapped patch based inference had a big impact on enhancing performance accuracy. The selection of training patch size was also found to be crucial in improving the model performance. Moreover, the need for an effective balance of positive and negative training patches was found. The best performance in our study was obtained by using overlapped patch inference, training with RELU activation and batch normalization in a 3D U-Net architecture with training patch size of 128×128×32 that resulted in average values of precision= 97%, sensitivity= 96% and F1 (Dice)= 96% for the test dataset.

[1]  Serge J. Belongie,et al.  Normalized cuts in 3-D for spinal MRI segmentation , 2004, IEEE Transactions on Medical Imaging.

[2]  Jan S. Kirschke,et al.  Btrfly Net: Vertebrae Labelling with Energy-based Adversarial Learning of Local Spine Prior , 2018, MICCAI.

[3]  Thomas Brox,et al.  3D U-Net: Learning Dense Volumetric Segmentation from Sparse Annotation , 2016, MICCAI.

[4]  H. Labelle,et al.  Spine Segmentation in Medical Images Using Manifold Embeddings and Higher-Order MRFs , 2013, IEEE Transactions on Medical Imaging.

[5]  Bilwaj Gaonkar,et al.  A deep learning approach to spine segmentation using a feed-forward chain of pixel-wise convolutional networks , 2018, 2018 IEEE 15th International Symposium on Biomedical Imaging (ISBI 2018).

[6]  Shuo Li,et al.  Spine‐GAN: Semantic segmentation of multiple spinal structures , 2018, Medical Image Anal..

[7]  Alejandro F. Frangi,et al.  Medical Image Computing and Computer-Assisted Intervention -- MICCAI 2015 , 2015, Lecture Notes in Computer Science.

[8]  Ronald M. Summers,et al.  A multi-center milestone study of clinical vertebral CT segmentation , 2016, Comput. Medical Imaging Graph..

[9]  Purang Abolmaesumi,et al.  Fast Automatic Vertebrae Detection and Localization in Pathological CT Scans - A Deep Learning Approach , 2015, MICCAI.

[10]  Christopher Nimsky,et al.  Cube-Cut: Vertebral Body Segmentation in MRI-Data through Cubic-Shaped Divergences , 2014, PloS one.

[11]  Sandeep Dutta,et al.  Evaluation of the impact of deep learning architectural components selection and dataset size on a medical imaging task , 2018, Medical Imaging.

[12]  Hala M. Ebeid,et al.  Vertebrae segmentation techniques for spinal medical images , 2015, 2015 IEEE Seventh International Conference on Intelligent Computing and Information Systems (ICICIS).

[13]  Yoshitaka Masutani,et al.  Automatic detection of vertebral number abnormalities in body CT images , 2017, International Journal of Computer Assisted Radiology and Surgery.

[14]  Hao Chen,et al.  Automatic Localization and Identification of Vertebrae in Spine CT via a Joint Learning Model with Deep Neural Networks , 2015, MICCAI.

[15]  S Crozier,et al.  Automated detection, 3D segmentation and analysis of high resolution spine MR images using statistical shape models , 2012, Physics in medicine and biology.

[16]  Christopher Nimsky,et al.  Square-Cut: A Segmentation Algorithm on the Basis of a Rectangle Shape , 2012, PloS one.

[17]  Stefano Pedemonte,et al.  DeepSPINE: Automated Lumbar Vertebral Segmentation, Disc-level Designation, and Spinal Stenosis Grading Using Deep Learning , 2018, MLHC.

[18]  Daguang Xu,et al.  Deep Image-to-Image Recurrent Network with Shape Basis Learning for Automatic Vertebra Labeling in Large-Scale 3D CT Volumes , 2017, MICCAI.

[19]  Mohammed Benjelloun,et al.  Vertebra identification using template matching modelmp and $$K$$K-means clustering , 2014, International Journal of Computer Assisted Radiology and Surgery.

[20]  Chengwen Chu,et al.  Fully Automatic Localization and Segmentation of 3D Vertebral Bodies from CT/MR Images via a Learning-Based Method , 2015, PloS one.

[21]  Stefan Freitag,et al.  VolHOG: a volumetric object recognition approach based on bivariate histograms of oriented gradients for vertebra detection in cervical spine MRI. , 2014, Medical physics.

[22]  Bostjan Likar,et al.  A Framework for Automated Spine and Vertebrae Interpolation-Based Detection and Model-Based Segmentation , 2015, IEEE Transactions on Medical Imaging.

[23]  Aly A. Farag,et al.  3D vertebrae segmentation using graph cuts with shape prior constraints , 2010, 2010 IEEE International Conference on Image Processing.

[24]  Thomas Brox,et al.  U-Net: Convolutional Networks for Biomedical Image Segmentation , 2015, MICCAI.

[25]  Ben Glocker,et al.  Automatic Localization and Identification of Vertebrae in Arbitrary Field-of-View CT Scans , 2012, MICCAI.

[26]  Bram van Ginneken,et al.  Iterative convolutional neural networks for automatic vertebra identification and segmentation in CT images , 2018, Medical Imaging.

[27]  Mohammed Benjelloun,et al.  Spine Localization in X-ray Images Using Interest Point Detection , 2009, Journal of Digital Imaging.

[28]  Mohammed Benjelloun,et al.  A Framework of Vertebra Segmentation Using the Active Shape Model-Based Approach , 2011, Int. J. Biomed. Imaging.

[29]  Boštjan Likar,et al.  Parametric modelling and segmentation of vertebral bodies in 3D CT and MR spine images , 2011, Physics in medicine and biology.

[30]  Guoyan Zheng,et al.  Fully automatic segmentation of lumbar vertebrae from CT images using cascaded 3D fully convolutional networks , 2017, 2018 IEEE 15th International Symposium on Biomedical Imaging (ISBI 2018).