**2.4 Deep learning method**

Deep learning method is a contemporary expansion of the traditional neural network method. **Figure 4** represents deep learning like a neural network with multicovers.

Rapid growth of current computing allowed deep learning for constructing the neural networks along with huge amount of covers, which is impossible for traditional neural networks. Basically, this technique helps to investigate many critical nonlinear models in the information. One more cause for the recent acceptance of deep learning techniques is owing to the enhancement of the critical and volume of data [15]. Dissimilar to the traditional neural network, this process generally applies more hidden levels in order that the algorithms can handle critical data with different structures [5]. In the healthcare applications, the generally applied deep learning algorithms consist of recurrent neural network, convolution neural network technique, deep neural network, and deep belief network. Convolution neural network is the most accepted one in 2016. The convolution neural network is extended to analyzing the ineptitude of the traditional machine

**127**

**Figure 4.**

*Multilayer neural network.*

disease stages.

*Application of Artificial Intelligence in Modern Healthcare System*

learning algorithms when conducting high dimensional data, that is, data with a huge number of characteristics. Conventionally, the machine learning algorithms are considered to examine data when the number of characteristics is little. The image data are physically high dimensional because each image generally includes thousands of pixels as characteristics. One explanation is to present dimension decrease: primarily preselect an object of pixels as elements and then complete the machine learning algorithms on the ensuing lower dimensional traits. However, heuristic feature selection events may drop the information in the images. Unsupervised learning methods such as clustering or PCA can be applied for data-determined dimension decrease. The convolution neural network was first projected the high-dimensional image investigation [16], where the inputs for convolution neural network are the accurately regulated pixel values on the images. The convolution neural network then transmitted the pixel values in the image throughout weighting in the difficulty layers and variety in the subsampling layers instead. The ultimate output is a recursive purpose of the weighted input values. The weights are skilled to reduce the average error involving the predictions and the outcomes. The performance of convolution neural network has been incorporated in trendy software packages such as Caffe from Berkeley AI Research [17] and Tensor Flow from Google [18]. Recently, the convolution neural network has been effectively executed in the healthcare area to help disease identification. It is used to identify the congenital cataract disease throughout learning the ocular images [19], though it has over 90% accuracy on identification and treatment implication. Convolution neural network was performed to identify skin cancer from clinical images [20]. Convolution neural network is applied to identify referable diabetic retinopathy via the retinal fundus photographs [21]. The specificity and sensitivity of the algorithm are both over 90%, which expressed the usefulness of using the method in the analysis of diabetes. It is importance to declare that in all this type of applications, the presentation of the convolution neural network is competitive beside an experienced physician in the truthfulness for categorizes both usual and

*DOI: http://dx.doi.org/10.5772/intechopen.90454*

## *Application of Artificial Intelligence in Modern Healthcare System DOI: http://dx.doi.org/10.5772/intechopen.90454*

*Alginates - Recent Uses of This Natural Polymer*

is always global optimum. Additionally, many obtainable rounded optimization technique applications are readily available for the SVM performance. SVM has been widely applied in healthcare research. For example, SVM was used to recognize imaging biomarkers of psychiatric and neurological disease [11]. SVM was also applied in cancer diagnosis [12]. SVM and other statistical methods can also be used to reach early detection of Alzheimer's syndrome [13]. SVM was applied to analyze the power of an offline human and device interface, which can control the upper-

Deep learning method is a contemporary expansion of the traditional neural network method. **Figure 4** represents deep learning like a neural network with

Rapid growth of current computing allowed deep learning for constructing the neural networks along with huge amount of covers, which is impossible for traditional neural networks. Basically, this technique helps to investigate many critical nonlinear models in the information. One more cause for the recent acceptance of deep learning techniques is owing to the enhancement of the critical and volume of data [15]. Dissimilar to the traditional neural network, this process generally applies more hidden levels in order that the algorithms can handle critical data with different structures [5]. In the healthcare applications, the generally applied deep learning algorithms consist of recurrent neural network, convolution neural network technique, deep neural network, and deep belief network. Convolution neural network is the most accepted one in 2016. The convolution neural network is extended to analyzing the ineptitude of the traditional machine

**126**

limb prostheses [14].

*Neural network diagram.*

multicovers.

**Figure 3.**

**2.4 Deep learning method**

learning algorithms when conducting high dimensional data, that is, data with a huge number of characteristics. Conventionally, the machine learning algorithms are considered to examine data when the number of characteristics is little. The image data are physically high dimensional because each image generally includes thousands of pixels as characteristics. One explanation is to present dimension decrease: primarily preselect an object of pixels as elements and then complete the machine learning algorithms on the ensuing lower dimensional traits. However, heuristic feature selection events may drop the information in the images. Unsupervised learning methods such as clustering or PCA can be applied for data-determined dimension decrease. The convolution neural network was first projected the high-dimensional image investigation [16], where the inputs for convolution neural network are the accurately regulated pixel values on the images. The convolution neural network then transmitted the pixel values in the image throughout weighting in the difficulty layers and variety in the subsampling layers instead. The ultimate output is a recursive purpose of the weighted input values. The weights are skilled to reduce the average error involving the predictions and the outcomes. The performance of convolution neural network has been incorporated in trendy software packages such as Caffe from Berkeley AI Research [17] and Tensor Flow from Google [18]. Recently, the convolution neural network has been effectively executed in the healthcare area to help disease identification. It is used to identify the congenital cataract disease throughout learning the ocular images [19], though it has over 90% accuracy on identification and treatment implication. Convolution neural network was performed to identify skin cancer from clinical images [20]. Convolution neural network is applied to identify referable diabetic retinopathy via the retinal fundus photographs [21]. The specificity and sensitivity of the algorithm are both over 90%, which expressed the usefulness of using the method in the analysis of diabetes. It is importance to declare that in all this type of applications, the presentation of the convolution neural network is competitive beside an experienced physician in the truthfulness for categorizes both usual and disease stages.

**Figure 4.** *Multilayer neural network.*

## **2.5 Natural language processing**

Genetic data and EP plus image are all machine-comprehensible, that is why the machine learning (ML) algorithms can be straightly presented after quality control processes or appropriate preprocessing. Though huge extents of medical data are like descriptive content, like a substantial examination, operative notes, and an experimental laboratory reports and release abstracts, these are formless and inconceivable for computer programming. Below this background, natural language processing (NLP) targets removing helpful data from the descriptive text to support the medical conclusion making [3]. A natural language processing (NLP) pipeline includes two main components: (i) classification and (ii) text processing. During text processing, the natural language processing (NLP) recognizes a sequence of disease-appropriate keywords at clinical remarks related to the past records [22]. After that, keyword subsets are preferred during analyzing their achievements in the arrangement in the normal abnormal cases. The authorized keywords then enter and enhance the controlled information to support medical choice making. The natural language processing pipelines have been developed to help the medical choice making on attentive treatment preparations and monitoring critical effects. For instance, it was showed that establishment of natural language processing, for analyzing the chest X-ray reports would help the antibiotic assistant system to aware physicians for the probable necessitate for anti-infective therapy [23]. Natural language processing was used to mechanically monitor laboratory-based difficult effects. Moreover, the natural language processing pipelines can also assist with disease analysis [24]. A recognized of 14 cerebral aneurysm disease-associated changeable during executing natural language processing (NLP), based on the clinical remarks [25]. Resulting variables are effectively applied for categorizing the common patients and the patients with cerebral problems, with 86% to 95% accuracy rates on the validation and training trials correspondingly. A natural language processing was implemented to extort the peripheral arterial diseaseallied keywords from description clinical remarks. The keywords are then applied to categorize the common patients and the patients who have peripheral arterial disease, which reaches over 90% accurate [22].
