Brain Tumor Detection using MLops and Hybrid Multi-Cloud

Abstract

The research of brain tumor classification by hand is a time-consuming operation with the potential for human mistakes. An accurate and automatic brain tumor classification technique is provided in this approach. To segment the brain part first, a thresholding approach is performed, followed by a morphological operation. Because the brain Magnetic Resonance Imaging(MRI) dataset is limited, the Convolutional Neural Network's (CNN) 'VGG19’ transfer learning network is used. With little weights and photos of benign, glioma, and meningioma, VGG19 classification layer is substituted by the softmax layer. After the model is created, it is saved and using Flask, a WebApp is created that will help us to detect the brain tumor using Hybrid Multi-Cloud and Machine Learning Operations(MLOps) tools. Identity and Access Management (IAM) is used to keep the data secure. This end-to-end system can be used for object detection applications in hospitals and local medical clinics.