Brain MRI Image Classification and Cells Localization Analysis

Abstract

Brain Tumor is the type of ailment in which the brain cell attains irregular growth and, unless diagnosed at an early stage, it will cause an individual to get injured. But once it is diagnosed, one does not know where the cells are at times. Should it be diagnosed with, then the patient may be advised with a suitable cure to eradicate the ailment. This is so because the brain is the most significant organ in our system of organs whose failure to treat well can leave an individual in a coma or even worse. The most important aspect of the diagnosis of brain tumor is to be capable of identifying the correct location. AI can be used to detect tumors and here, it would assist in locating cell as well. The initial analysis involved training the model to identify the presence of a tumor therein with a combination of several image preprocessing techniques such as enhancement, denoising, and resizing among others were implemented so as to provide adequate performance and learning. To find the location of the cell, many images were annotated manually both using Roboflow as well as using YoloV5 with the parameters and a few adjustments only and then the model was trained to determine the location of the cell. They followed the process in the sense that when the image is suspected to be classified as tumor class, the location will be subsequently identified. Into it BrainDetModel1 is made with the aid of several distinct layers and pre-trained model give the accuracy score of 94% and yolov5 model object detection is made which has the accuracy score of 96.2% on the given validation dataset.