https://github.com/dawnofthedebayan/vssegcode
https://github.com/dawnofthedebayan/vssegcode
Last synced: 11 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/dawnofthedebayan/vssegcode
- Owner: dawnofthedebayan
- Created: 2023-11-01T11:48:21.000Z (over 2 years ago)
- Default Branch: master
- Last Pushed: 2023-11-10T08:43:02.000Z (over 2 years ago)
- Last Synced: 2025-02-25T08:52:12.241Z (over 1 year ago)
- Language: Python
- Size: 1.86 MB
- Stars: 0
- Watchers: 1
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Automatic Segmentation of Vestibular Schwannoma from MRI
## Table of Contents
1. [Overview](#1-Overview)
2. [Installation](#2-installation)
3. [Usage](#3-usage)
4. [Results](#4-results)
5. [Dataset](#5-dataset)
6. [Model Architecture](#6-model-architecture)
7. [Images](#7-images)
8. [Citation](#8-citation)
9. [License](#9-license)
## 1. Overview
This repository contains the implementation of our two-stage Convolutional Neural Network (CNN) model for the automatic segmentation of Vestibular Schwannoma (VS) from T1 and T2-weighted Magnetic Resonance Images (MRI). This project aims to improve the accuracy of tumor size estimation in VS, which is crucial for patient management and clinical workflow.
## 2. Installation
To use our network, you will need the following software and libraries:
You can install these dependencies using the following command:
```bash
pip install -r requirements.txt
```
## 3. Usage
To run the model for Vestibular Schwannoma segmentation, follow these steps:
1. **Data Preprocessing:**
- Prepare your MRI data in the appropriate format.
2. **Model Training:**
- Train the model on your dataset.
3. **Model Evaluation:**
- Evaluate the model using publicly available or in-house datasets.
For more detailed instructions, please refer to the documentation or code provided in this repository.
## 4. Results
CASCADENET's performance has been evaluated on different datasets and MRI variations. Here are some key results:
- 3D CNN achieved a Dice score of 0.85 for T1-weighted MRI using a public dataset.
- Our proposed method using T1-weighted MRI on the public dataset achieved a Dice score of 0.89.
- In our in-house dataset, the Dice score using a 2D CNN was 0.79.
- Our method achieved a Dice score of 0.83 on the same in-house dataset.
## 5. Dataset
We used publicly available and in-house datasets for model evaluation. Please refer to the "Dataset" section of the code or documentation for more information on these datasets and their sources.
## 6. Model Architecture
For a detailed understanding of CASCADENET's architecture, please refer to the "Model Architecture" section of the code or documentation.
## 7. Images
*Architecture of CascadeNet*
*Qualitative Results*
## 8. Citation
If you use our model in your research or find it useful, please cite the following paper:
```
[Insert Citation Here]
```
## 9. License
For any questions or issues, please feel free to contact Debayan Bhattacharya (debayan.bhattacharya@tuhh.de).