RESEARCH ARTICLE Deep learning model integrating features and novel classifiers fusion for brain tumor segmentation Sajid Iqbal 1,2 | Muhammad U. Ghani Khan 2 | Tanzila Saba 3 | Zahid Mehmood 4 | Nadeem Javaid 5 | Amjad Rehman 6 | Rashid Abbasi 7 1 Department of Computer Science, Bahauddin Zakariya University, Multan, Pakistan 2 Department of Computer Science and Engineering, University of Engineering and Technology, Lahore, Pakistan 3 College of Computer and Information Sciences, Prince Sultan University, Riyadh, Saudi Arabia 4 Department of Computer Engineering, University of Engineering and Technology, Taxila, Pakistan 5 Department of Computer Science, COMSATS University, Islamabad, Pakistan 6 College of Computer and Information Systems, Al Yamamah University, Riyadh, Saudi Arabia 7 School of Computer and Technology, Anhui University, Hefei, China Correspondence Sajid Iqbal, Department of Computer Science, Bahauddin Zakariya University, Multan, Pakistan. Email: sajidiqbal@bzu.edu.pk Review Editor: Peter Saggau Abstract Automatic and precise segmentation and classification of tumor area in medical images is still a challenging task in medical research. Most of the conventional neural network based models usefully connected or convolutional neural networks to per- form segmentation and classification. In this research, we present deep learning models using long short term memory (LSTM) and convolutional neural networks (ConvNet) for accurate brain tumor delineation from benchmark medical images. The two different models, that is, ConvNet and LSTM networks are trained using the same data set and combined to form an ensemble to improve the results. We used publicly available MICCAI BRATS 2015 brain cancer data set consisting of MRI images of four modalities T1, T2, T1c, and FLAIR. To enhance the quality of input images, multiple combinations of preprocessing methods such as noise removal, his- togram equalization, and edge enhancement are formulated and best performer com- bination is applied. To cope with the class imbalance problem, class weighting is used in proposed models. The trained models are tested on validation data set taken from the same image set and results obtained from each model are reported. The individ- ual score (accuracy) of ConvNet is found 75% whereas for LSTM based network pro- duced 80% and ensemble fusion produced 82.29% accuracy. KEYWORDS brain tumor segmentation, convolutional neural networks, ensemble neural networks, LSTM 1 | INTRODUCTION Uncontrolled growth of tissue cells results in a tumor that may be benign or malignant/cancerous. Brain cancer is among the different causes for an increase in mortality rate in the world. The death rate, due to a brain tumor, has been increased to 300% in the last few decades (Saba et al., 2019; Saba, Bokhari, Sharif, Yasmin, & Raza, 2018c; Tahir et al., 2019). According to National Cancer Institute America (NCIA) and American Cancer Society (ACS), around 1.5 mil- lion cancer cases are identified in 2013 and more than 1.6 million can- cer patients are registered in 2014 in America. The count is increased from 1.6 to 1.65 in 2015. By looking at statistics of 2014, 33% (0.6 million) of diagnosed patients have lost their lives (Khan, Akram, et al., 2019b; Khan, Lali, et al., 2019c). A benign tumor is a mass that does not spread and does not cause any harm whereas a malignant tumor is harmful and is also known as cancer. A malignant tumor is categorized into different classes based on various characteristics like type of originating tissue, location, and level of severity. Brain tumors are also categorized based on their ori- gin: primary brain tumors are the lesions that emerge within the brain whereas secondary brain tumors (metastatic tumors) originate at a dif- ferent location of the body and move to the brain. There are different categorization schemes designed by researchers but the scheme designed by the World Health Organization (WHO) is considered to Received: 21 January 2019 Revised: 24 March 2019 Accepted: 12 April 2019 DOI: 10.1002/jemt.23281 Microsc Res Tech. 2019;1–14. wileyonlinelibrary.com/journal/jemt © 2019 Wiley Periodicals, Inc. 1