Datasets
Standard Dataset
Brats-2018_2019-2D

- Citation Author(s):
- Submitted by:
- BICHUAN FENG
- Last updated:
- Thu, 03/13/2025 - 05:03
- DOI:
- 10.21227/kvmr-yg04
- Data Format:
- License:
- Categories:
- Keywords:
Abstract
Brain tumors are one of the most common diseases threatening human health. Early detection and precise segmentation are of great significance for clinical diagnosis and treatment. This paper presents a Learnable Wavelet Transform and Attention Mechanism network(LWTA-Net2D), based on 2D Convolutional Neural Networks (CNN), integrating Learnable Discrete Wavelet Transform (LDWT), combination of Monte Carlo Attention (MCattn) and Monte Carlo Bottleneck Layer (MCBottleneck), and a U-Net-based encoder-decoder architecture. By training and testing on 2D slices of the BRATS2018 and BRATS2019 datasets, the proposed model demonstrates superior performance in multi-scale feature capturing and spatial detail enhancement. Experimental results show that the improved model achieves Intersection over Union (IoU) and Dice Similarity Coefficient (Dice) of 0.7842 and 0.8766, respectively, representing a 5\% improvement over traditional U-Net methods, significantly enhancing segmentation accuracy and robustness. This study provides an efficient and precise solution for MRI brain tumor segmentation tasks, showing great potential in medical image processing.
Training and validation sets and mask