Coder Social home page Coder Social logo

zhangsongdmk / aide Goto Github PK

View Code? Open in Web Editor NEW

This project forked from lich0031/aide

0.0 1.0 0.0 7.33 MB

AIDE: Annotation-efficient deep learning for automatic medical image segmentation

License: GNU Lesser General Public License v2.1

Python 100.00%

aide's Introduction

AIDE

AIDE: Annotation-efficient deep learning for automatic medical image segmentation

Introduction

This is the official code of AIDE, a deep learning framework for automatic medical image segmentation with imperfect datasets, including those having limited annotations, lacking target domain annotations, and containing noisy annotations. Automatic segmentation of medical images plays an essential role in both scientific research and medical care. Deep learning approaches have presented encouraging performances, but existing high-performance methods typically rely on very large training datasets with high-quality manual annotations, which are normally difficult or even impossible to obtain in many clinical applications. We introduce AIDE, a novel annotation-efficient deep learning framework to handle imperfect training datasets.

Quick start

Install

  1. Install PyTorch=1.1.0 following the official instructions.
  2. git clone https://github.com/lich0031/AIDE.
  3. Install dependencies: pip install -r requirements.txt

Data preparation

For CHAOS, it should like this:

$inputs_chaos

|-- All_Sets
|--|--Case_No
|--|--|--T1DUAL
|--|--|--|--DICOM_anon
|--|--|--|--Ground

Train and evaluate

  • Please specify the configuration file.

  • For example, train the comparison model on CHAOS with a batch size of 4 on GPU 0:

    python train_files/trainchaos_comparison_1case.py --model_name fuseunet --batch_size 4 --gpu_order 0 --repetition 1

  • Model evaluation on the CHAOS dataset can utilize the file train_files/evalchaos_comparison_1cases.py by modifying the image and optimized model path and information accordingly.

Hardware and time complexities

  • To train the model, computers with GPUs should be utilized. The optimization time of the model depends on various factors, including the dataset, the batch size, the epoch number, and the hardware. For our implementation of the CHAOS data, it took the comparison model (984 training samples) around 300s to run one epoch, and it took our framework around 420s. Our framework is a little bit more complex as two models are trained in parallel.
  • Installation of the relevant dependencies (e.g. PyTorch) is very fast, taking less than half an hour.
  • The models can be evaluated on computers with or without GPU. Evaluation is very fast, and it takes only several seconds to evaluate one 3D image.

Example results

Example segmentation results on the CHAOS dataset can be found in train_files/examplesegmentationresults. Additional optimized models and segmentation results for the task can be downloaded here.

aide's People

Contributors

lich0031 avatar

Watchers

 avatar

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    ๐Ÿ–– Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. ๐Ÿ“Š๐Ÿ“ˆ๐ŸŽ‰

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google โค๏ธ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.