A convolutional neural network Cascade for plantar pressure images registration.

BACKGROUND Plantar pressure image (PPI) recorded in high spatial and temporal resolution is very useful in clinical gait analysis. For functional analysis of PPI, image registration is often performed to maximally correlate source image with a template image. Previous methods estimate the registration parameters by iteratively optimizing different objective functions. These methods are often computational expensive to achieve satisfactory registration accuracy. RESEARCH QUESTION Can we develop a single PPI registration technique that performs more rapidly than previous methods, and that also maintains adequate PPI correspondence as defined by various (dis)similarity metrics? METHODS A cascaded convolutional neural network (CNN) was proposed for the registration of PPIs. Our model was trained to learn a regression from the difference between the template and misaligned images to the registration parameters. The registration performance was evaluated by three different metrics, i.e. the mean squared error (MSE), the exclusive or (XOR), and the mutual information (MI). For comparison, four previous methods were also implemented. These included the principal axes (PA) method, the center of pressure trajectory (COP) method, the MSE method, and the XOR method. RESULTS Experimental results on a dataset with 71 PPI template-source pairs showed that the proposed CNN-based method could obtain comparable registration accuracy to the MSE and XOR method. With regards to the registration speed, registration durations (mean ± sd in seconds) per image pair were: MSE (30.584 ± 2.171), XOR (24.245 ± 1.596), PA (0.016 ± 0.003), COP (25.614 ± 0.341), and the proposed model (0.054 ± 0.007). SIGNIFICANCE Our findings indicate that the proposed registration approach can achieve high accuracy but less computational time. Thus, it is more practical to utilize our pre-trained CNN-based model to develop near-real time applications for plantar pressure images registration.

[1]  Xiaogang Wang,et al.  Structured Feature Learning for Pose Estimation , 2016, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[2]  J. Friedman.  Greedy function approximation: A gradient boosting machine. , 2001 .

[3]  Michael E. Tipping.  Sparse Bayesian Learning and the Relevance Vector Machine , 2001, J. Mach. Learn. Res..

[4]  João Manuel R S Tavares,et al.  Rapid pedobarographic image registration based on contour curvature and optimization. , 2009, Journal of biomechanics.

[5]  Shohei Tateishi,et al.  Nonlinear regression modeling via the lasso-type regularization , 2010 .

[6]  Xin Liu,et al.  AgeNet: Deeply Learned Regressor and Classifier for Robust Apparent Age Estimation , 2015, 2015 IEEE International Conference on Computer Vision Workshop (ICCVW).

[7]  K. Kendrick,et al.  Gait Influence Diagrams in Parkinson’s Disease , 2017, IEEE Transactions on Neural Systems and Rehabilitation Engineering.

[8]  Kaiming He,et al.  Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks , 2015, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[9]  Mohammad Reza Daliri,et al.  Chi-square distance kernel of the gaits for the diagnosis of Parkinson's disease , 2013, Biomed. Signal Process. Control..

[10]  S. Zeger,et al.  Longitudinal data analysis using generalized linear models , 1986 .

[11]  Mohammad Reza Daliri,et al.  Automatic diagnosis of neuro-degenerative diseases using gait dynamics , 2012 .

[12]  T. Horstmann,et al.  Changes in plantar pressure distribution after Achilles tendon augmentation with flexor hallucis longus transfer. , 2008, Clinical biomechanics.

[13]  A Karlsson,et al.  Correlations between force plate measures for assessment of balance. , 2000, Clinical biomechanics.

[14]  Geoffrey E. Hinton,et al.  ImageNet classification with deep convolutional neural networks , 2012, Commun. ACM.

[15]  Qiao Li,et al.  In-Shoe Plantar Pressure Measurement and Analysis System Based on Fabric Pressure Sensing Array , 2010, IEEE Transactions on Information Technology in Biomedicine.

[16]  J. Y. Goulermas,et al.  A comparison of seven methods of within-subjects rigid-body pedobarographic image registration. , 2008, Journal of biomechanics.

[17]  Michael N. Vrahatis,et al.  Recent approaches to global optimization problems through Particle Swarm Optimization , 2002, Natural Computing.

[18]  Max A. Viergever,et al.  Mutual-information-based registration of medical images: a survey , 2003, IEEE Transactions on Medical Imaging.

[19]  Robin M Queen,et al.  Effect of shoe type on plantar pressure: a gender comparison. , 2010, Gait & posture.

[20]  Francisco P.M. Oliveira,et al.  Registration of pedobarographic image data in the frequency domain , 2010, Computer methods in biomechanics and biomedical engineering.

[21]  C. H. M. van Schie,et al.  A review of the biomechanics of the diabetic foot. , 2005 .

[22]  J Y Goulermas,et al.  Nonlinear spatial warping for between-subjects pedobarographic image registration. , 2009, Gait & posture.

[23]  João Manuel R. S. Tavares,et al.  Enhanced spatio-temporal alignment of plantar pressure image sequences using B-splines , 2013, Medical & Biological Engineering & Computing.

[24]  Susannah J Sample,et al.  Variance associated with walking velocity during force platform gait analysis of a heterogeneous sample of clinically normal dogs. , 2017, American journal of veterinary research.

[25]  Qingshan Liu,et al.  Adaptive Cascade Regression Model For Robust Face Alignment , 2017, IEEE Transactions on Image Processing.

[26]  A J Harrison,et al.  A moment-based technique for the automatic spatial alignment of plantar pressure data , 2000, Proceedings of the Institution of Mechanical Engineers. Part H, Journal of engineering in medicine.