Brain tumor magnetic resonance images enhanced by hybrid method based on deep learning paradigm

Abstract

The development of software engineering has given very successful results in the field of medical diagnosis in recent years. Deep learning and machine learning applications give remarkable results in the detection, monitoring, diagnosis, and treatment of possible tumoral regions with the analysis of the obtained medical images and data mining. Studies to diagnose brain tumors are essential because of the wide variety of brain tumors, the importance of the patient's survival time, and the brain tumor's aggressive nature. Brain tumors are defined as a disease with destructive and lethal features. Detection of a brain tumor is an essential process because of the difficulty in distinguishing between abnormal and normal tissues. With the right diagnosis, the patient can get excellent treatment, extending their lifespan. Despite all the research, there are still significant limitations in detecting tumor areas because of abnormal lesion distribution. It may be challenging to locate an area with very few tumor cells because areas with such small areas frequently appear healthy. Studies are becoming more common in which automated classification of early-stage brain tumors is performed using deep learning or machine learning approaches. This study proposes a hybrid deep learning model for the detection and early diagnosis of brain tumors via magnetic resonance imaging. The dataset images were subjected to Local Binary Pattern (LBP) and Long Short-Term Memory (LSTM) algorithms. The highest accuracy rate obtained in the hybrid model created is 98.66%.