Skip to content

🫁 Chest X-ray abnormalities localization via ensemble of deep convolutional neural networks

License

Notifications You must be signed in to change notification settings

pvti/CXRAbnormalityLocalization

Folders and files

NameName
Last commit message
Last commit date

Latest commit

Β 

History

15 Commits
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 

Repository files navigation

Visitors

🫁 Chest X-ray abnormalities localization via ensemble of deep convolutional neural networks

1Modelling and Simulation Centre, Viettel High Technology Industries Corporation, Vietnam  2Kings Distributed Systems, Canada  3Eutech Systems Ltd, Bangladesh  βœ‰Corresponding Author
Convolutional neural networks have been applied widely in chest X-ray interpretation thanks to the availability of high-quality datasets. Among them, VinDr-CXR is one of the latest public datasets including 18000 expert-annotated images labeled into 22 local position-specific abnormalities and 6 globally suspected diseases. A proposed deep learning algorithm based on Faster-RCNN, Yolov5, and EfficientDet frameworks were developed and investigated in the task of multi-class clinical detection from chest radiography. The ground truth was defined by a radiologist-adjudicated image review. Their performance was evaluated by the mean average precision. The results show that the best performance belonged to object detection models ensembled with an EfficientNet classifier, resulting in a peak mAP of 0.292. As a trade-off, ensembling object detection models was much slower, increasing computing time by 3.75, 5, and 2.25 times compared to FasterRCNN, Yolov5, and EfficientDet individually. Overall, the classifiers show constant improvement on all detector models, which is recommended for further research. All of these aspects should be considered to address real-world CXR diagnosis where the accuracy and computing cost are of concern.

πŸ‘¨β€βš•οΈ News

πŸ“‹ Main results

Evaluation of the proposed framework on the VinDr-CXR test dataset.

Detector Accuracy ([email protected]) Performance
Single model Resnet50 EficientNet-B7 Speed GPU memory requirement (MB) Training time (hour)
YOLOv5 0.21 0.246 0.269 15 3291 7
FasterRCNN 0.248 0.263 0.278 20 2076 9.5
EfficientDet 0.269 0.28 0.273 9 3685 12
Ensemble 0.272 0.285 0.292 4 3685 30.5

πŸ’» Installation

Please refer to INSTALL.md for installation instructions.

🩺 Model zoo

Trained models are available in the MODEL_ZOO.md.

πŸ” Dataset zoo

Please see DATASET_ZOO.md for detailed description of the training/evaluation datasets.

πŸ’‰ Getting Started

Follow the aforementioned instructions to install environments and download models and datasets.

GETTING_STARTED.md provides a brief intro of the usage of builtin command-line tools.

πŸ’Š Citing

If you use this work in your research or wish to refer to the results, please use the following BibTeX entry.

@inproceedings{pham2021chest,
  title={Chest x-ray abnormalities localization via ensemble of deep convolutional neural networks},
  author={Pham, Van-Tien and Tran, Cong-Minh and Zheng, Stanley and Vu, Tri-Minh and Nath, Shantanu},
  booktitle={2021 International Conference on Advanced Technologies for Communications (ATC)},
  pages={125--130},
  year={2021},
  organization={IEEE}
}