Paper reading (八)：Deep Learning and Its Applications in Biomedicine

论文题目：Deep Learning and Its Applications in Biomedicine

scholar 引用：59

页数：16

发表时间：2017.07

发表刊物：Genomics, Proteomics & Bioinformatics

作者：Chensi Cao , Feng Liu , Hai Tan , Deshou Song , Wenjie Shu ,
Weizhong , Yiming Zhou , Xiaochen Bo , Zhi Xie 

摘要：Keywords: Deep learning, Big data, Bioinformatics, Biomedical informatics, Medical image, High-throughout sequencing

Advances in  biological and medical technologies have providing us explosive volumes of biological and physiological data, such as medical images, electroencephalography(脑电图学), genomic and protein sequences. Learrning from these data facilitates the understanding of human health and disease. Developed from artificial neural networks, deep learning based algorithms show great promise in extracting features and learning patterns from complex data. The aim of these paper is to provide an overview of deep learning techniques and some of the state-of-the-art applications in the biomedical field. We first introduce the development of artificial nerual network and deep learning. We then describe two main components of deep learning, i.e., deep learning architechtures and model optimization. Subsequently, some examples are demonstrated for deep learning applications, including medical image classification, genomic sequence analysis, as well as protein structure classification and prediction. Finally, we offer our perspectives for the future directions in the field of deep learning.

结论：

• deep learning is becoming more and more popular in genomic sequence analysis.
• Deep learning is represented by a group of technologies, has been widely used in biomedical data.(跟上篇论文类似，可以列一个application的table)
• Manually labeling these medical images is laborious and requires professional experts.
• medical images are highly associated with privacy, so collecting and protecting the data is demanding.
• biomedical data are usually imbalanced because the quantity of data from normal classes is much larger than that from other classes.
• Unlike other images, subtle changes in medical images may indicate disease, deep learning requires technological improvements.

Introduction：

• Deep learning in general has two properties: (1) multiple layers of nonlinear processing units; (2) supervised or unsupervised learning of feature presentations on each layer.
• a massive growth in biomedical data, such as genomic sequences, protein structures, and medical images.
• 这篇paper对deep learning的内容讲得挺细的。

正文组织架构：

1. Introduction (为什么introduction中没有biomedical的内容？)

(1) The development of ANNs

(2) The development of deep learning

2.Brief description of deep learning

(1) Basic concepts

• Activation functions
• Optimization objective (Loss function and Regularization term)
• Optimization methods

(2) Deep learning structures

• AEs
• RBMs and DBNs
• Convolutional neural networks
• Recurrent neural networks

3.Applications in biomedicine

(1) Medical image classification and segmentation

(2) Genomic sequencing and gene expression analysis

(3) Prediction of protein structure

4.Conclusion and perspective

正文部分内容摘录：

• due to the complexity of ANNs, other simple machine learning algorithms, such as support vector machines(SVMs), random forest, and k-nearest neighbors algorithms(k-NN), graduallly overtook ANNs in popularity.
• Most DNNs use cross entropy between the training data and the model distribution as the loss function.
• A learning task is transformed to an optimization problem, to achieve the minima of the objective function by selecting appropriate hyperparameters.
• Table 1. Applications of deep learning frameworks in biomedical informatics(这不就是我在读上一篇paper的时候想列的table吗  哈哈)  非常赞！！！ 重点关注：Genomic sequencing and gene expression analysis
• Detection of lesion and abnormality is the major issue in medical image analysis.
• 每一个领域有这么多的deep learning相关的方法，那么描述这些方法的逻辑应该怎样确定呢？不是按时间关系这么简单吧？
• 后面讲述的很多方法中都提到了AUC指标或者accuracy，也可以将各方法的指标列一下？