Detection of brain tumor from MRI images using EfficientNet: A deep learning approach
Geetika Sharma, Moksh Sarda, Nikhil Malik, Harnoor Singh, Rahul Mishra, Somil Grover
Abstract
Brain tumors are either cancerous or noncancerous masses of abnormal brain cells. Meningioma tumor, glioma tumor, and pituitary tumor are the three forms of brain cancers that are examined in this research. One essential diagnostic technique for the identification and diagnosis of brain tumors is brain tumor segmentation. Although manual brain tumor segmentation using MRI is possible, the accuracy and identification are not very good. The classification of abnormalities is a time-consuming task for physicians because it is not predictable and straightforward. However, tumor detection is difficult because tumors have complex appearances and boundary characteristics. In this work, we report on an automated method for detecting and diagnosing brain tumors from MRI images using EfficientNet. Specifically, we developed a preliminary report generation system that automatically provides the type of tumor, its corresponding ICD code, the tumor's diameter and area, and the original input image with the tumor highlighted for visualization. This will be accomplished through the classification and segmentation of MRI images, with classification using Dense EfficinetNet to identify the tumor type and ICD code and segmentation using EfficientNet UNet to visualize the tumor and determine its diameter and area. A preliminary report for our paper is being created, which will use one MRI image of a brain tumor to diagnose tumor diameter and tumor area.