International Journal of Electrical and Computer Engineering (IJECE) Vol. 11, No. 6, December 2021, pp. 5530~5540 ISSN: 2088-8708, DOI: 10.11591/ijece.v11i6.pp5530-5540  5530 Journal homepage: http://ijece.iaescore.com A hybrid deep learning approach towards building an intelligent system for pneumonia detection in chest X-ray images Ihssan S. Masad 1 , Amin Alqudah 2 , Ali Mohammad Alqudah 3 , Sami Almashaqbeh 4 1,3,4 Department of Biomedical Systems and Informatics Engineering, Yarmouk University, Irbid, Jordan 2 Department of Computer Engineering, Yarmouk University, Irbid, Jordan Article Info ABSTRACT Article history: Received Sep 27, 2020 Revised Jun 7, 2021 Accepted Jun 17, 2021 Pneumonia is a major cause for the death of children. In order to overcome the subjectivity and time consumption of the traditional detection of pneumonia from chest X-ray images; this work hypothesized that a hybrid deep learning system that consists of a convolutional neural network (CNN) model with another type of classifiers will improve the performance of the detection system. Three types of classifiers (support vector machine (SVM), k-nearest neighbor (KNN), and random forest (RF) were used along with the traditional CNN classification system (Softmax) to automatically detect pneumonia from chest X-ray images. The performance of the hybrid systems was comparable to that of the traditional CNN model with Softmax in terms of accuracy, precision, and specificity; except for the RF hybrid system which had less performance than the others. On the other hand, KNN hybrid system had the best consumption time, followed by the SVM, Softmax, and lastly the RF system. However, this improvement in consumption time (up to 4 folds) was in the expense of the sensitivity. A new hybrid artificial intelligence methodology for pneumonia detection has been implemented using small-sized chest X-ray images. The novel system achieved a very efficient performance with a short classification consumption time. Keywords: Convolutional neural network classification Deep learning This is an open access article under the CC BY-SA license. Corresponding Author: Ihssan S. Masad Department of Biomedical Systems and Informatics Engineering Hijjawi Faculty for Engineering Technology Yarmouk University 566 Shafiq Irshidat Street, Irbid 21163, Jordan Email: imasad@yu.edu.jo 1. INTRODUCTION Pneumonia is a respiratory condition in which lungs are affected by infection [1]. As the leading cause for the death of children under the age of 5, where it accounts for around 16% of all deaths of children [2]. Pneumonia kills over 800,000 children around the world every year [3], [4]. Adults can be affected by pneumonia as well, where over 50,000 people die every year, and more than one million people in the US (for example) are admitted to hospitals because of pneumonia; making it the most common cause of hospital admissions other than women giving birth [5]. Although pneumonia can be diagnosed by different imaging modalities such as magnetic resonance imaging (MRI) [6], [7] and computed tomography (CT) [8], [9], chest X-ray imaging is still the most common method for pneumonia diagnosis, because it is cheap, fast, and clinically more available.