The on intelligence algorithms in diagnosis domain like
The most critical gland of the human body is the liver that is a vital organ due to its functionality and hepatitis is the inflammation of the liver. The hepatitis viruses represent a significant public health problem worldwide and replicate primarily in the liver causing related diseases with a broad range of clinical manifestations 1. According to world health statistics in 2015, the seventh cause of 1.34 million deaths was due to hepatitis2. Despite the importance of this disease, it neglects as a health factor. Hepatitis B and C viruses in chronic conditions can lead to other liver diseases including cirrhosis and liver cancer-chronic 3. A blood test would be helpful to achieve a correct diagnosis and check if specific antiviral antibodies are existing 4. The accurate diagnosis could be the impact factor in optimal treatments so a less diagnosis error can support patients to receive appropriate care.
Nowadays medicine society prefers to save time to the interpretation and communication of clinical decisions to support patients. The presence of new technologies like medical data analyses has been abundant, and intelligent algorithms produce significant results. Multilayer network can train a neural network to diagnosis diabetes 5, chronic obstructive pulmonary disease6 and Hepatitis 7. Learning method can extract features from ultrasound images to classify and diagnose the staging of the chronic liver diseases8 or for adjudging the degree of liver ?brosis in chronic hepatitis C 9 also, liver segmentation from CT volumes10 and MRI images11.
Machine Learning algorithm can learn from data and make consideration and prediction based on a set of data. It is usual to use the advantages of methods inside of each other such as benign or malignant classification breast cancer tumors by Genetically Optimized Neural Network 12. Also, extraction of morphologic features is to utilize as the input of multilayer feed-forward neural network13, Elsewhere; there could be a combination of many techniques such as support vector machine and simulated annealing to be used in both methods for detection of some hepatitis14. Experiences have shown that combination methods bring data problems to be covered by another part. Nevertheless, it still might be better to use data mining methods before data provision as the inputs of a learning algorithm. Some medical equipment base on intelligence algorithms in diagnosis domain like Elastography and application in diagnosis of hepatic fibrosis stage15. However, the Elastography extracts the features, whose relationship that discovered so far, but it is not sufficiently accurate. In this regard, deep learning is an admirable ability to gain powerful features with linear and nonlinear relationships. The conventional method used in the training of a restricted Boltzmann machine is the contrastive divergence16. This technique has an impact on training performance. The advantage of this is that a gradient estimate is definite and a more significant training rate can be used 17. In addition to, Principle Component Analysis (PCA) as a reduction method can be used to classify hepatitis 18, PCA extracts linear features, apparently, unable to take non-linear ones, contrarily, RBM can extract them19. Auto-encoder as one of the Deep learning methods in dimension reduction is much more favorable than previous methods for dimension reduction, including principal component analysis20. Though an auto-encoder can be very desirable to extract features using deep learning, this should take into consideration that good results require more time to optimize the model.
Solving many issues is simple for computers. Deep learning seeks to learn experiences and understands the world through a series of concepts21. AI is working on new learning methods to facilitate diagnosis processes with a reliable accuracy. Deep Learning applies in all scientific fields. In recent years, deep learning that is a branch of Machine Learning (ML) has successfully been working due to wonderful ability in automatic learning, which is the main feature of these methods.
There are five different types of hepatitis viruses including types A, B, C, D and E that among them types A, B, and C are more common. Features and symptoms of hepatitis B and C are so similar and choosing them to use in diagnoses is of great importance. Use of Deep Neural Network is expected to be helpful to achieve this goal.
The RBM has created an independent framework for competing in nonlinear classifications; Therefore, a discriminative RBM can also be prosperous and has a satisfactory analysis in a semi-supervised learning22. It can argue that this method use as a method of non-linear independent classifications that would be very useful in medical areas23. As respect, learning methods that are flexible, able to extract attributes automatically and obtain nonlinear dependencies would be essential in the field of medicine24.
This research studies the effectiveness of deep learning for classifying hepatitis B and hepatitis C virus. This method is efficient and fast to gain desirable results. It is clear that additional tests are needed to diagnose hepatitis, but this research tries to prove that deep learning can classify hepatitis B and hepatitis C by having a set of general laboratory tests for liver health and gain a satisfactory accuracy. The propose of Deep learning is to solve some of the problems that are understandable to man’s mind but cannot be described by computer model. This article shows that results of the medical test can be correctly modeled in deep learning, so, some of the additional laboratory tests for diagnosis of hepatitis may remove in some cases. Also, in comparison with other methods that use feature engineering for classification, DBM is more reliable because of it’s extreme ability to extract non-linear features.
This paper organized as follows. In section 2, we introduce RBM as training algorithm in Deep Boltzmann Machine (DBM) architecture. Database description is provided in Section 3. In Section 4, the quality of the presented passage is examined and investigate the effect of DBM to classify hepatitis B and hepatitis C, and the conclusion of this work appears in Section 5.