Optimized CNN And Mobilenet Models For Accurate Four-Class Tumor Cancer Detection

Brain tumors are a group of cancers that originate from various cells of the central nervous system or cancers from other tissues that spread to the brain. Excessive cell growth in the brain is referred to as a tumor. This information is vital for selecting the best treatment approach, including surgery and radiation therapy. Early detection paves the way for much more effective treatment. Therefore, accurate prediction of a brain tumor in the early stages is crucial for its diagnosis and treatment. One of the most important applications of image processing in medical science is in the field of medical imaging. Therefore, the aim of this article is the intelligent detection of brain tumors from three classes: meningioma, glioma, and pituitary, using deep learning, we present techniques based on deep learning and convolutional neural network. Also, because the CNN technique has been used many times in articles and theses. In this thesis, there is an improved CNN method, improved CNN method, Resnet technique, EfficientNetV۲L, ConvNeXtBase and new MobileNet technique. The results showed that there is a lot of challenge in the four-class dataset for tumor diagnosis and its types. And the use of object detection techniques in the image with CNN and MobileNet method has better accuracy than other methods.

Keywords:

Deep Learning , Convolution Neural Network , MobileNet , Brian Tumor , Cancer Detection