Literature-review.docx

Deep learning in 1-D data Sohail Manzoor (17-MS-SE-19) Hadeed Ullah (17-MS-SE-08) Unknown

Literature Review: Deep learning methods perform very well for large size data sets than any other type of methods. Because deep learning algorithms check results against all possible inputs while in simple classic machine methods only few inputs are examined. Deep learning algorithms used in many applications which we are using in medical field for diagnosis of many disease(Faust et al., 2018).Deep learning with convolutional neural network is examined for the purpose of showing that both pathological and EEG recordings are different. For this purpose two basic convent’s architectures named shallow and deep were used. The results were better in decoding pathology case. Automated methods were used to check performance of architectural hyperparameters and notice very large difference between these architectures. Experiment showed that for ConveNet decoding behavior they used spectral power and frequency values of delta and theta used in architects were changed also other features work exact according to expectations which they were exacting from the analysis of medical textual and the experiment showed accuracy is increased with increasing the contextual data for example subject age. The techniques of ConvNet used in this method become base for further automated diagnosis on the base of EEG(Schirrmeister et al., 2017). Encephalogram (EEG) is used for diagnosis of epilepsy and used ancillary test for this purpose. Information about the activity of brain is controlled or hold by EEG signals. In past for this purpose techniques used was not giving good results for example there was variety of results against similar data, techniques used was taking a lot of time for performing required operations, don’t have capabilities to show all abnormalities also there were some limitations in technical artifact sense. Keeping in mind all discussed limitations there was need of an automated in other words computer aided diagnosis (CAD) system which use different machine

learning techniques and has capability to distinguish the class of EEG signals. It was start of examination of EEG signals on the base of CNN. The experiment was help to distinguish different classes for example normal, preictal and seizure. For this purpose a 13-layer deep CNN were used. Technique proved good results for all perspectives for accuracy 88.67%, for specificity 90.00% and for sensitivity it displayed 95.00%(Acharya et al., 2018). To make human life good and free from seizure type disease there is need of Brain-computer interface which will predict presence of seizure In human and in this way a human can find ways to fight with this type of disease. It is very difficult to develop a system discussed above because it needs a lot’s of effort and info. The EEG signals are nonstationary in nature keeping mind this nature of EEG signals it is very difficult to detect Seizure patters because it vary from patient to patient. Very large and huge amount of data is produced while brain recording using implanted electrodes. This big data is a big problem because large amount of data need large amount of resources to process to store. For the solution of above problems there were BCI system was introduced. The purpose of that’s system was to analyze EEG signals. In this system first of all dimensionality-reduction techniques was introduced for the purpose of reducing classification error and increase classification capacity with decreasing processing time and bandwidth for communication. A stack auto encoder was introduced for following two steps first extraction of unsupervised feature and the other classification. For the analysis of big EEG signals a cloud-computing methodology was introduced. When proposed system examined it show good result and become a new useful BCI system for epilepsy patients(Hosseini et al., 2016). Automated system for the detection of Parkinson’s disease was introduced. The system used convolutional neural network for performing this operation. They used 20 EEG signals of the subjects facing Parkinson’s disease and 20 EEG signals of normal subjects and get excellent results for accuracy ,sensitivity and specify by implementing thirteen layer CNN(Oh et al., 2018). An automated system was implemented or developed for the detection of MI ECG beats and normal beats. The proposed system give good result for with noise and without noise signals. No other function like feature extraction and selection was performed in this system(Acharya et al., 2017b). An automated system was developed to detect Coronary artery disease (CAD) and to overcome all problems that old system were facing while detection of CAD for this purpose a 4 layer CNN was used. For diagnosis of CAD four max pooling layers and three fully contacted layers used. The proposed system Show great result for both nets in EEG signal observation(Acharya et al., 2017a). Using 1-D convolutional neural network a system is developed for the classification and detection of patients. Using 1-D CNN two major functions of ECG feature extraction and classification handled and ECG trained with patients and used for the classification of patients it was able to classify the patients according to trained data. The proposed system show excellent results than previous proposed systems but the system was invariant

nature and it can be used for any type of ECG datasets(Kiranyaz et al., 2016). For the segmentation of Cell in histopathological images a system was proposed using deep learning algorithms. The proposed system also use spatial relationship with deep learning algorithms to get better result for cell segmentation. For this purpose they collect cellular and extracellular samples from histopathological images and perform operations on collected samples and when get result it was noticed collected results show better segmentation details than old proposed methods(Hatipoglu and Bilgin, 2017). A simple and really quick technique of information acquisition, feature extraction and have house creation for convulsion detection. The scalp graphical record (EEG) dataset collected at the Children’s Hospital Boston from twenty two pediatric patients having 192 unmalleable seizures (available as CHBMIT database) is employed to assess this easy approach against existing ones with terribly positive results reaching up to 99.48% Sensitivity(Bugeja et al., 2016).

References: ACHARYA, U. R., FUJITA, H., LIH, O. S., ADAM, M., TAN, J. H. & CHUA, C. K. 2017a. Automated detection of coronary artery disease using different durations of ECG segments with convolutional neural network. Knowledge-Based Systems, 132, 62-71. ACHARYA, U. R., FUJITA, H., OH, S. L., HAGIWARA, Y., TAN, J. H. & ADAM, M. 2017b. Application of deep convolutional neural network for automated detection of myocardial infarction using ECG signals. Information Sciences, 415, 190-198. ACHARYA, U. R., OH, S. L., HAGIWARA, Y., TAN, J. H. & ADELI, H. 2018. Deep convolutional neural network for the automated detection and diagnosis of seizure using EEG signals. Computers in biology and medicine, 100, 270-278. BUGEJA, S., GARG, L. & AUDU, E. E. A novel method of EEG data acquisition, feature extraction and feature space creation for early detection of epileptic seizures. Engineering in Medicine and Biology Society (EMBC), 2016 IEEE 38th Annual International Conference of the, 2016. IEEE, 837-840.

FAUST, O., HAGIWARA, Y., HONG, T. J., LIH, O. S. & ACHARYA, U. R. 2018. Deep learning for healthcare applications based on physiological signals: a review. Computer methods and programs in biomedicine. HATIPOGLU, N. & BILGIN, G. 2017. Cell segmentation in histopathological images with deep learning algorithms by utilizing spatial relationships. Medical & biological engineering & computing, 55, 1829-1848. HOSSEINI, M.-P., SOLTANIAN-ZADEH, H., ELISEVICH, K. & POMPILI, D. Cloud-based deep learning of big eeg data for epileptic seizure prediction. 2016 IEEE Global Conference on Signal and Information Processing (GlobalSIP), 2016. IEEE, 1151-1155. KIRANYAZ, S., INCE, T. & GABBOUJ, M. 2016. Real-time patient-specific ECG classification by 1-D convolutional neural networks. IEEE Transactions on Biomedical Engineering, 63, 664-675. OH, S. L., HAGIWARA, Y., RAGHAVENDRA, U., YUVARAJ, R., ARUNKUMAR, N., MURUGAPPAN, M. & ACHARYA, U. R. 2018. A deep learning approach for Parkinson’s disease diagnosis from EEG signals. Neural Computing and Applications, 1-7. SCHIRRMEISTER, R., GEMEIN, L., EGGENSPERGER, K., HUTTER, F. & BALL, T. Deep learning with convolutional neural networks for decoding and visualization of EEG pathology. Signal Processing in Medicine and Biology Symposium (SPMB), 2017 IEEE, 2017. IEEE, 1-7.