Development of Deep Learning Model Base on Modified CNN Architectures for Brain Tumours Early Diagnosis

Brain cancer is classified as a malignant tumours that can spread to other parts of the human body, especially the spine and brain instantaneously. Brain tumour has been considered as the biggest problem for human health due to the high mortality rate it has. However, not all brain tumours are classified as malignant tumours and can be treated properly before they mutate to be malignant. Unfortunately, brain tumours is very difficult to detect early because it often resembles normal tissue. Therefore, this disease is often detected when it has reach a critical phase and difficult to treat. This study aims to develop a Deep Learning model for an automatic classification system in order to help early detection of brain tumours by utilizing the CNN algorithm with several popular architectures such as EfficientNetB0, MobileNetV2, and Xception. In addition, to improve the performance of the developed Deep Learning model, modifications were made to the CNN architecture by adding several additional layers. Based on the results obtained, this study managed to achieve very good performance and surpassed previous research benchmarks with accuracy, precision and recall values above 97%.