Related Experiment Video
Updated: Sep 17, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.0K
U-Net-based architecture with attention mechanisms and Bayesian Optimization for brain tumor segmentation using MR
K Ramalakshmi1, L Krishna Kumari2
1Electronics and Communication Engineering, P.S.R.Engineering College, Sivakasi, 626140, Tamilnadu, India.
Computers in Biology and Medicine
|July 1, 2025
Summary
Artificial intelligence (AI) aids radiologists in diagnosing brain tumors (BT) using MRI scans. A novel U-Net model with attention mechanisms and Bayesian optimization significantly improves BT segmentation accuracy.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Computer Science
Background:
- Advanced computer technology enables artificial intelligence (AI) for diagnosing brain tumors (BT).
- Early disease identification is crucial for effective therapies, with AI offering significant time and cost savings.
- Conventional methods struggle with the complexities of Magnetic Resonance (MR) imaging for BT detection.
Purpose of the Study:
- To develop an advanced deep learning model for precise brain tumor segmentation in MR images.
- To enhance segmentation performance by integrating Attention Mechanisms and Hyperparameter Optimization (HPO) with a U-Net architecture.
- To validate the model's efficacy using diverse MRI brain tumor datasets and compare it against state-of-the-art methods.
Main Methods:
- A U-Net-based deep learning architecture was developed, incorporating Attention Mechanisms for improved feature focus.
- Bayesian Optimization Algorithm was employed for Hyperparameter Optimization (HPO) to fine-tune model variables.
- Region-Adaptive Thresholding was utilized for pinpointing tumor regions, with segmentation results validated against ground truth annotations.
Main Results:
- The proposed U-Net-based model achieved a high DICE score of 0.89687 for MRI-BT segmentation.
- Performance was evaluated using metrics including IoU, accuracy, and DICE Score across LGG, Healthcare, and BraTS 2021 datasets.
- The model demonstrated superior performance compared to existing state-of-the-art methods in brain tumor segmentation.
Conclusions:
- The developed U-Net model with Attention Mechanisms and Bayesian Optimization offers a robust solution for brain tumor segmentation in MRI.
- This AI-driven approach addresses limitations of traditional methods, enhancing diagnostic accuracy and efficiency.
- The findings highlight the potential of deep learning in improving medical image analysis for neuro-oncology.
