Page 261 - Handbook of Deep Learning in Biomedical Engineering Techniques and Applications

252 Chapter 9 Applications of deep learning in biomedical engineering

Figure 9.5 Recurrent neural network.

dealing with static and dynamic situations. The architecture can be
highly recommended to classify time series data.
RNN can provide insight feature, which can be utilized to
modeling biological signals such as ECG and EEG. Some of the
recent applications of RNN in biomedicine are as follows:
1. Reconstruction of missing data from time series signals
2. Finding the relationship between muscle activity and arm
kinematics
3. Real-time applications such as detecting glucose level of
children
4. Classification of biological signals
5. Classifying snoring and nonsnoring signals
6. Blood cell classification methods
7. Sleep-stage segmentation [12]

12. Generative adversarial networks
GAN is an emerging technique of generative models. It can be
employed for both semisupervised and unsupervised learning. It
has been attained by modeling distributions of data with high
dimension. GAN aims to generate new samples by learning the
input data patterns. It consists of two types of network
[13].They are as follows.

12.1 Generator network

It focuses on image generation. It takes the input as fixed-
length random vector and produces samples. This vector acts