Supervised learning for bone shape and cortical thickness estimation from CT images for finite element analysis

HighlightsA bone shape regression produces surface meshes for cortical and trabecular bone.Local corrections are done based on supervised learning between QCT and HRpQCT images.For FEA purposes, those corrections are regularised by Gaussian process model. Cortical thickness is automatically estimated with high accuracy without segmentation. Graphical abstract Figure. No caption available. ABSTRACT Knowledge about the thickness of the cortical bone is of high interest for fracture risk assessment. Most finite element model solutions overlook this information because of the coarse resolution of the CT images. To circumvent this limitation, a three‐steps approach is proposed. 1) Two initial surface meshes approximating the outer and inner cortical surfaces are generated via a shape regression based on morphometric features and statistical shape model parameters. 2) The meshes are then corrected locally using a supervised learning model build from image features extracted from pairs of QCT (0.3‐1 mm resolution) and HRpQCT images (82 &mgr;m resolution). As the resulting meshes better follow the cortical surfaces, the cortical thickness can be estimated at sub‐voxel precision. 3) The meshes are finally regularized by a Gaussian process model featuring a two‐kernel model, which seamlessly enables smoothness and shape‐awareness priors during regularization. The resulting meshes yield high‐quality mesh element properties, suitable for construction of tetrahedral meshes and finite element simulations. This pipeline was applied to 36 pairs of proximal femurs (17 males, 19 females, 76 ± 12 years) scanned under QCT and HRpQCT modalities. On a set of leave‐one‐out experiments, we quantified accuracy (root mean square error = 0.36 ± 0.29 mm) and robustness (Hausdorff distance = 3.90 ± 1.57 mm) of the outer surface meshes. The error in the estimated cortical thickness (0.05 ± 0.40 mm), and the tetrahedral mesh quality (aspect ratio = 1.4 ± 0.02) are also reported. The proposed pipeline produces finite element meshes with patient‐specific bone shape and sub‐voxel cortical thickness directly from CT scans. It also ensures that the nodes and elements numbering remains consistent and independent of the morphology, which is a distinct advantage in population studies.

[1]  Mauricio Reyes,et al.  A novel registration-based methodology for prediction of trabecular bone fabric from clinical QCT: A comprehensive analysis , 2017, PloS one.

[2]  Alexandre Terrier,et al.  Identification of elastic properties of human patellae using micro-finite element analysis. , 2016, Journal of biomechanics.

[3]  Dieter H. Pahr,et al.  From high-resolution CT data to finite element models: development of an integrated modular framework , 2009 .

[4]  F. Deconinck,et al.  Information Processing in Medical Imaging , 1984, Springer Netherlands.

[5]  Pierre Geurts,et al.  Extremely randomized trees , 2006, Machine Learning.

[6]  Yohan Payan,et al.  A fast and robust patient specific Finite Element mesh registration technique: Application to 60 clinical cases , 2010, Medical Image Anal..

[7]  P. Zysset,et al.  Validation of an anatomy specific finite element model of Colles' fracture. , 2008, Journal of biomechanics.

[8]  Timothy F. Cootes,et al.  Learning-Based Shape Model Matching: Training Accurate Models with Minimal Manual Input , 2015, MICCAI.

[9]  Thor F Besier,et al.  Predictive statistical models of baseline variations in 3-D femoral cortex morphology. , 2016, Medical engineering & physics.

[10]  Cyril Flaig,et al.  On Smoothing Surfaces in Voxel Based Finite Element Analysis of Trabecular Bone , 2009, LSSC.

[11]  Ling Qin,et al.  Clinical Use of Quantitative Computed Tomography (QCT) of the Hip in the Management of Osteoporosis in Adults: the 2015 ISCD Official Positions-Part I. , 2015, Journal of clinical densitometry : the official journal of the International Society for Clinical Densitometry.

[12]  S. Cummings,et al.  BMD at Multiple Sites and Risk of Fracture of Multiple Types: Long‐Term Results From the Study of Osteoporotic Fractures , 2003, Journal of bone and mineral research : the official journal of the American Society for Bone and Mineral Research.

[13]  Mauricio Reyes,et al.  Statistical analysis of the inter-individual variations of the bone shape, volume fraction and fabric and their correlations in the proximal femur. , 2017, Bone.

[14]  Andrew H Gee,et al.  Predicting Hip Fracture Type With Cortical Bone Mapping (CBM) in the Osteoporotic Fractures in Men (MrOS) Study , 2015, Journal of bone and mineral research : the official journal of the American Society for Bone and Mineral Research.

[15]  Enrico Dall'Ara,et al.  Clinical versus pre-clinical FE models for vertebral body strength predictions. , 2014, Journal of the mechanical behavior of biomedical materials.

[16]  Gabriel Taubin,et al.  Curve and surface smoothing without shrinkage , 1995, Proceedings of IEEE International Conference on Computer Vision.

[17]  Robert M. Haralick,et al.  Textural Features for Image Classification , 1973, IEEE Trans. Syst. Man Cybern..

[18]  F. Kainberger,et al.  A nonlinear QCT-based finite element model validation study for the human femur tested in two configurations in vitro. , 2013, Bone.

[19]  Yan Kang,et al.  A new accurate and precise 3-D segmentation method for skeletal structures in volumetric CT data , 2003, IEEE Transactions on Medical Imaging.

[20]  H. Genant,et al.  Comparison of proximal femur and vertebral body strength improvements in the FREEDOM trial using an alternative finite element methodology. , 2015, Bone.

[21]  Andrew H. Gee,et al.  Imaging the femoral cortex: Thickness, density and mass from clinical CT , 2012, Medical Image Anal..

[22]  E. Dall’Ara,et al.  Orthotropic HR-pQCT-based FE models improve strength predictions for stance but not for side-way fall loading compared to isotropic QCT-based FE models of human femurs. , 2014, Journal of the mechanical behavior of biomedical materials.

[23]  Timothy F. Cootes,et al.  Active Shape Models-Their Training and Application , 1995, Comput. Vis. Image Underst..

[24]  Andrew H. Gee,et al.  Independent measurement of femoral cortical thickness and cortical bone density using clinical CT , 2015, Medical Image Anal..

[25]  Moritz Tannast,et al.  Head-Neck Osteoplasty has Minor Effect on the Strength of an Ovine Cam-FAI Model: In Vitro and Finite Element Analyses , 2016, Clinical orthopaedics and related research.

[26]  Gábor Székely,et al.  Statistical model based shape prediction from a combination of direct observations and various surrogates: Application to orthopaedic research , 2012, Medical Image Anal..

[27]  J. Cauley,et al.  Public health impact of osteoporosis. , 2013, The journals of gerontology. Series A, Biological sciences and medical sciences.

[28]  Andrew H. Gee,et al.  The Effects on the Femoral Cortex of a 24 Month Treatment Compared to an 18 Month Treatment with Teriparatide: A Multi-Trial Retrospective Analysis , 2016, PloS one.

[29]  Thomas Gerig,et al.  Gaussian Process Morphable Models , 2016, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[30]  M. Viceconti,et al.  Experimental validation of DXA-based finite element models for prediction of femoral strength. , 2016, Journal of the mechanical behavior of biomedical materials.

[31]  R. Harnish,et al.  Automatic multi-parametric quantification of the proximal femur with quantitative computed tomography. , 2015, Quantitative imaging in medicine and surgery.

[32]  Andrew H. Gee,et al.  High resolution cortical bone thickness measurement from clinical CT data , 2010, Medical Image Anal..

[33]  Brent C Taylor,et al.  Risk Factors for Hip Fracture in Older Men: The Osteoporotic Fractures in Men Study (MrOS) , 2016, Journal of bone and mineral research : the official journal of the American Society for Bone and Mineral Research.

[34]  Kate A. Gavaghan,et al.  Facial nerve image enhancement from CBCT using supervised learning technique , 2015, 2015 37th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC).

[35]  Yutaka Ohtake,et al.  A comparison of mesh smoothing methods , 2003 .

[36]  F. Taddei,et al.  Can CT image deblurring improve finite element predictions at the proximal femur? , 2016, Journal of the mechanical behavior of biomedical materials.

[37]  William E. Lorensen,et al.  Marching cubes: A high resolution 3D surface construction algorithm , 1987, SIGGRAPH.

[38]  Stefan Weber,et al.  Highly Accurate Facial Nerve Segmentation Refinement From CBCT/CT Imaging Using a Super-Resolution Classification Approach , 2018, IEEE Transactions on Biomedical Engineering.

[39]  Mauricio Reyes,et al.  Prediction of Trabecular Bone Anisotropy from Quantitative Computed Tomography Using Supervised Learning and a Novel Morphometric Feature Descriptor , 2015, MICCAI.

[40]  D. C. Bauer,et al.  Erratum to: Clinical utility of routine laboratory testing to identify possible secondary causes in older men with osteoporosis: the osteoporotic fractures in men (MrOS) study , 2016, Osteoporosis International.

[41]  Max A. Viergever,et al.  elastix: A Toolbox for Intensity-Based Medical Image Registration , 2010, IEEE Transactions on Medical Imaging.

[42]  Richard J. Cook,et al.  Risk factors for hip fracture in older home care clients. , 2009, The journals of gerontology. Series A, Biological sciences and medical sciences.

[43]  H. Genant,et al.  Accuracy of CT-based thickness measurement of thin structures: modeling of limited spatial resolution in all three dimensions. , 2002, Medical physics.

[44]  Raphaël Marée,et al.  Extremely Randomized Trees and Random Subwindows for Image Classification, Annotation, and Retrieval , 2013 .

[45]  Ludovic Humbert,et al.  Technical Note: Cortical thickness and density estimation from clinical CT using a prior thickness-density relationship. , 2016, Medical physics.

[46]  C. Thomas,et al.  Relation between age, femoral neck cortical stability, and hip fracture risk , 2005, The Lancet.

[47]  Normand Robert,et al.  Generalized method for computation of true thickness and x-ray intensity information in highly blurred sub-millimeter bone features in clinical CT images , 2012, Physics in medicine and biology.

[48]  Antonio Criminisi,et al.  Image Quality Transfer via Random Forest Regression: Applications in Diffusion MRI , 2014, MICCAI.

[49]  Willi A Kalender,et al.  An anatomic coordinate system of the femoral neck for highly reproducible BMD measurements using 3D QCT. , 2005, Computerized Medical Imaging and Graphics.

[50]  Thomas Vetter,et al.  A Unified Approach to Shape Model Fitting and Non-rigid Registration , 2013, MLMI.

[51]  Andrew H. Gee,et al.  Focal osteoporosis defects play a key role in hip fracture , 2017, Bone.

[52]  Philippe C. Cattin,et al.  Prediction of Cranio-Maxillofacial Surgical Planning Using an Inverse Soft Tissue Modelling Approach , 2013, MICCAI.

[53]  Dorin Comaniciu,et al.  Shape Regression Machine , 2007, IPMI.

[54]  William H. Press,et al.  Numerical Recipes 3rd Edition: The Art of Scientific Computing , 2007 .

[55]  Philippe K. Zysset,et al.  FEA to Measure Bone Strength: A Review , 2016, Clinical Reviews in Bone and Mineral Metabolism.

[56]  Robert C. Bolles,et al.  Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography , 1981, CACM.

[57]  Klaus Engelke,et al.  FEA to measure bone strength , 2016 .

[58]  W. Kalender,et al.  Accuracy limits for the determination of cortical width and density: the influence of object size and CT imaging parameters. , 1999, Physics in medicine and biology.

[59]  V. Gudnason,et al.  Interactive graph-cut segmentation for fast creation of finite element models from clinical ct data for hip fracture prediction , 2016, Computer methods in biomechanics and biomedical engineering.

[60]  Ahmed Bouridane,et al.  Classification of Prostatic Tissues using Feature Selection Methods , 2007 .

[61]  K. Y. Dai,et al.  A Smoothed Finite Element Method for Mechanics Problems , 2007 .

[62]  Christophe Geuzaine,et al.  Gmsh: A 3‐D finite element mesh generator with built‐in pre‐ and post‐processing facilities , 2009 .

[63]  Leo Joskowicz,et al.  Can a partial volume edge effect reduction algorithm improve the repeatability of subject-specific finite element models of femurs obtained from CT data? , 2014, Computer methods in biomechanics and biomedical engineering.

[64]  Raphaël Marée,et al.  Fast Multi-class Image Annotation with Random Subwindows and Multiple Output Randomized Trees , 2009, VISAPP.