Automatic Brain Tumour Detection using MRI Slice with MobileNet: An Analysis

Computerized disease examination is a common clinical practice and examination of the medical-data based mental-health using chosen algorithm is widely adopted in multispecialty hospitals. This process will considerably reduce the diagnostic time found in the traditional mental-health screening practice. This research aims to develop a deep-learning (DL) technique to detect the Brain Tumour (BT) in chosen Magnetic Resonance Imaging (MRI) database. This work employs the MobileNetV1 and MobilenetV2 models to detect the BT with improved accuracy. To increase the detection results, this work proposes feature reduction using $\mathbf{5 0 \%}$ dropout and features fusion. Then the merit of the develop system is confirmed using the binary classification with a 3 -fold cross validation. Initially, the classification is performed using the SoftMax and then the conventional classifiers widely adopted in the machine-learning approaches are also considered. The outcome confirms that this technique helps to achieve a detection accuracy of 99.5% with the chosen scheme. This confirms that the proposed technique works well on the chosen MRI database.