Research program of deep learning for medical image analysis

Research program of deep learning for medical image analysis

This is a study on "deep learning in Medical image processing," The first section of the report, mainly includes research objects, common methods, deep learning introduction, research status, research focus.
When I wrote my report, I found two reviews of deep learning for bioinformation/biomedicine and shared them:
Deep Learning in bioinformatic
Applications of deep learning in biomedicine

1. Research background and significance

Medical image analysis is an interdisciplinary field of integrated medical imaging, mathematical modeling, Digital image processing, artificial intelligence and numerical algorithms. Medical images include ultrasound, X-ray computed tomography (CT), nuclear magnetic resonance (MRI), digital blood vessel Silhouette (DSA), Positron tomography (PET), etc. In the field of medical image analysis, image segmentation, image registration and information Fusion, time series image analysis and content-based image retrieval are the main research directions. With the rapid development of medical imaging technology, medical image analysis into the era of big data, how to excavate the useful knowledge from the massive medical image data, so as to provide a more sufficient basis for clinical diagnosis and treatment and scientific research, has become a research hotspot in academia and industry.
Machine learning methods are widely used in medical image analysis to accomplish specific tasks on new data, such as classification, recognition and segmentation, by training models on a given set of data. The commonly used algorithms are support vector machine (SVM), Hidden Markov (HMM) and artificial neural network. However, traditional machine learning algorithms need to use prior knowledge to extract features from raw data to train the model. Because the feature selection is difficult, the model may have over-fitting problems, the generalization ability is difficult to guarantee; On the other hand, the traditional model is difficult to adapt to large scale data sets and the model is poor in expansibility.
Deep learning is a new field in machine learning, and its motive is to build and simulate the human brain for analytical study. Deep learning is a data-driven model that simulates the visual mechanism of the human brain and automatically learns the abstract features of all levels of data to better reflect the essential characteristics of the data. The research of Deep Neural network since 2006 Hinton has developed a multi-layer limited Boltzmann Machine level (RBM) structure based on probabilistic graph model, and has achieved corresponding success in the field of big data applications such as visual processing, speech processing, natural language processing and information retrieval in recent years. The good effect of the deep learning model in various fields has aroused the upsurge of data mining and analysis in more fields, and has also attracted much attention in the field of medicine and biology cognition. At present, deep learning has been involved in pathological classification of medical images [1-3], segmentation [4-5], recognition [6] and brain function research [7] and so on. Enlitic, a deep learning startup, developed a cancer detection system based on deep learning to detect lung cancer in chest CT images over doctors. [8] IBM proposes Watson for oncology, which analyzes patients ' medical information by learning a lot of data and experience to help doctors develop reliable medical solutions. [9] Alphago's Google subsidiary DeepMind recently announced the DeepMind Health project, using deep learning to further develop effective healthcare technology. [10]
Computational complexity has become the biggest obstacle in the research and application of deep neural network, and the deep neural network must be realized effectively by using new hardware structure. The deep Neural Network algorithm has the typical computation-intensive application characteristic, the speed bottleneck increasingly becomes the deep neural network theory research and the application development obstacle. At the same time, with the increase of network layer and the number of neurons in each layer, the computational complexity will increase exponentially with the system scale. This paper mainly puts forward the research scheme of deep learning algorithm in medical image analysis, constructs the deep neural network for the specific task, and optimizes its implementation according to the architecture characteristics while improving the effect, thus realizing the improvement of the effect and performance of the deep neural network.

1 Plis SM, Hjelm DR, Salakhutdinov R et al deep learning for neuroimaging:a validation study. Frontiers in Neuroscience 2014;8.
2 Li Q, Cai W, Wang X et al Medical image classification with convolutional neural network. In:control Automation Robotics & Vision (ICARCV), 13th International Conference on. P. 844-8. Ieee.
3 Ypsilantis-P, Siddique M, Sohn H-m et al. predicting Response to neoadjuvant chemotherapy with PET Imaging Using convo Lutional Neural Networks. PloS one 2015;10 (9): e0137036.
4 Turaga SC, Murray JF, Jain V et al convolutional networks can learn to generate affinity graphs for image segmentation. Neural Computation 2010;22 (2): 511-38.
5 Roth HR, Lu L, Farag A et al deeporgan:multi-level deep convolutional Networks for automated pancreas segmentation. Medical Image Computing and computer-assisted Intervention–miccai 2015. Springer, 2015, 556-64.
[6] Roth HR, Lu L, Seff A et al. A New 2.5 D representation for lymph node detection using the random sets of deep convolutional neural network observations. Medical Image Computing and computer-assisted Intervention–miccai 2014. Springer, 2014, 520-7.
[7] Koyamada S, Shikauchi Y, Nakae K et al. Deep learning of FMRI big Data:a novel
Approach to Subject-transfer decoding. ArXiv preprint arxiv:1502.00093 2015.
[8] enlitic technology detected lung cancer nodules in chest CT images
Http://www.enlitic.com/science.html#deep-learning.
[9] IBM Watson for oncology. Ibm.
Http://www.ibm.com/smarterplanet/us/en/ibmwatson/watson-oncology.html, 2016.
[Ten] DeepMind health. Google DeepMind. Https://www.deepmind.com/health, 2016.

Research program of deep learning for medical image analysis