# Segmentation of 2D Brain MR Images

**Authors:** Angad Ripudaman Singh Bajwa  
**arXiv:** [2111.03370](https://arxiv.org/abs/2111.03370)  
**Date:** 2021-11-05  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/2111.03370v1/

## Key points

- Brain tumour segmentation is crucial in medical image processing for early diagnosis and improved treatment possibilities.
- Manual segmentation of brain tumours from MRI images is challenging and time-consuming.
- An automatic brain tumour image segmentation method is needed to address this issue.
- The project aims to develop a method using deep neural architectures to accurately and quickly locate the tumour in 2D brain MR images.
- The dataset used for evaluation includes 3064 slices from 233 patients, with each image labeled according to the type of tumour.
- The Dice coefficient is used to evaluate the segmentation results, along with a Dice loss function to assess model performance.
- The deep architecture incorporates convolutional neural networks (CNNs) and other techniques to improve accuracy.
- Optimizations such as batch-wise processing and data augmentation are implemented during training for efficient computation.
- This project provides an automated solution for brain tumour segmentation in MRI images, potentially aiding medical professionals in timely diagnoses and improving treatment outcomes.

## Summary

, , , , 

Brain tumour segmentation is a crucial task in medical image processing, enabling early diagnosis and improving treatment possibilities for patients. However, manually segmenting brain tumours from a large number of MRI images is challenging and time-consuming. To address this issue, an automatic brain tumour image segmentation method is needed. This project aims to develop such a method that can accurately and quickly locate the tumour using deep neural architectures to analyze 2D brain MR images. The dataset used for evaluation includes 3064 slices from 233 patients, with each image labeled according to the type of tumour. The Dice coefficient is used to evaluate the segmentation results, along with a Dice loss function to assess model performance. The deep architecture incorporates convolutional neural networks (CNNs) and other techniques to improve accuracy. Due to limited resources, optimizations such as batch-wise processing and data augmentation are implemented during training for efficient computation. Overall, this project provides an automated solution for brain tumour segmentation in MRI images, potentially aiding medical professionals in timely diagnoses and improving treatment outcomes for patients with brain tumours.