A deep learning approach to fetal-ECG signal reconstruction

Fetal electrocardiogram (FECG) monitoring has become essential due to the current increase in the relative number of cardiac patients worldwide. This paper proposes to use a deep learning approach to compress/recover FECG signals, improving the computation speed in a telemonitoring system. The problem is analogous to the reconstruction of a non-sparse signal in compressive sensing (CS) framework. The architecture incorporates a non-linear mapping using a stacked denoising autoencoder (SDAE). The compression of the raw non-sparse FECG data takes place at the transmitter side using a deep neural network. After pre-training, the whole deep SDAE can be further fine tuned by the mini-batch gradient descent-based back-propagation algorithm. Although the training for SDAE is usually time-consuming, it does not affect the performance due to the one-time off-line training process. The real-time FECG reconstruction is faster due to a few matrix-vector multiplications at the receiver end. The simulations performed by employing standard non-invasive FECG databases shows promising results in terms of the reconstruction quality.

[1]  Geoffrey E. Hinton,et al.  Reducing the Dimensionality of Data with Neural Networks , 2006, Science.

[2]  Antonio Torralba,et al.  Small codes and large image databases for recognition , 2008, 2008 IEEE Conference on Computer Vision and Pattern Recognition.

[3]  Quoc V. Le,et al.  Recurrent Neural Networks for Noise Reduction in Robust ASR , 2012, INTERSPEECH.

[4]  Yoshua. Bengio,et al.  Learning Deep Architectures for AI , 2007, Found. Trends Mach. Learn..

[5]  M. Elad,et al.  $rm K$-SVD: An Algorithm for Designing Overcomplete Dictionaries for Sparse Representation , 2006, IEEE Transactions on Signal Processing.

[6]  Yoshua Bengio,et al.  Greedy Layer-Wise Training of Deep Networks , 2006, NIPS.

[7]  Pierre Vandergheynst,et al.  Compressed Sensing for Real-Time Energy-Efficient ECG Compression on Wireless Body Sensor Nodes , 2011, IEEE Transactions on Biomedical Engineering.

[8]  Kjersti Engan,et al.  Method of optimal directions for frame design , 1999, 1999 IEEE International Conference on Acoustics, Speech, and Signal Processing. Proceedings. ICASSP99 (Cat. No.99CH36258).

[9]  Yoshua Bengio,et al.  Extracting and composing robust features with denoising autoencoders , 2008, ICML '08.

[10]  Jason Weston,et al.  A unified architecture for natural language processing: deep neural networks with multitask learning , 2008, ICML '08.

[11]  Tzyy-Ping Jung,et al.  Compressed Sensing for Energy-Efficient Wireless Telemonitoring of Noninvasive Fetal ECG Via Block Sparse Bayesian Learning , 2012, IEEE Transactions on Biomedical Engineering.

[12]  Marc'Aurelio Ranzato,et al.  Semi-supervised learning of compact document representations with deep networks , 2008, ICML '08.

[13]  A. Bruckstein,et al.  K-SVD : An Algorithm for Designing of Overcomplete Dictionaries for Sparse Representation , 2005 .

[14]  Bart De Schutter,et al.  DAISY : A database for identification of systems , 1997 .

[15]  Geoffrey E. Hinton.  A Practical Guide to Training Restricted Boltzmann Machines , 2012, Neural Networks: Tricks of the Trade.

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

[17]  Yoshua Bengio,et al.  Gradient-based learning applied to document recognition , 1998, Proc. IEEE.

[18]  Geoffrey E. Hinton,et al.  Restricted Boltzmann machines for collaborative filtering , 2007, ICML '07.

[19]  Yee Whye Teh,et al.  A Fast Learning Algorithm for Deep Belief Nets , 2006, Neural Computation.

[20]  Geoffrey E. Hinton,et al.  Learning a Nonlinear Embedding by Preserving Class Neighbourhood Structure , 2007, AISTATS.

[21]  Geoffrey E. Hinton,et al.  Learning internal representations by error propagation , 1986 .

[22]  Jeffrey M. Hausdorff,et al.  Physionet: Components of a New Research Resource for Complex Physiologic Signals". Circu-lation Vol , 2000 .

[23]  Nitish Srivastava,et al.  Dropout: a simple way to prevent neural networks from overfitting , 2014, J. Mach. Learn. Res..

[24]  Rita Paradiso,et al.  A wearable health care system based on knitted integrated sensors , 2005, IEEE Transactions on Information Technology in Biomedicine.

[25]  Razvan Pascanu,et al.  On the difficulty of training recurrent neural networks , 2012, ICML.