Deep-learning model associating lateral cervical radiographic features with Cormack-Lehane grade 3 or 4 glottic view
This repository contains two Python scripts for a deep-learning model:
train.py
: A script for training the model on a dataset.test.py
: A script for testing the trained model on a DICOM file and generating a class activation map (CAM) with an overlay on the input image.
This project requires the following dependencies to be installed:
- Python 3.7+
- OpenCV
- Numpy
- Tensorflow 1.14
- argparse
This script trains the model on a given dataset.
python train.py --path <path_to_dataset> --input_size <input_size> <input_size> --batch_size <batch_size> --total_epoch <total_epoch> --save_path <save_path> --save_interval <save_interval> --learning_rate <learning_rate> --load_path <load_path>
Parameter | Description | Default Value |
---|---|---|
--path |
Path to the dataset | ./dataset |
--input_size |
Input size of the images | 256 256 |
--batch_size |
Batch size for training | 15 |
--total_epoch |
Total number of epochs | 100 |
--save_path |
Path to save the trained model | ./weights |
--save_interval |
Interval between model checkpoints | 1 |
--learning_rate |
Learning rate for the optimizer | 1e-3 |
--load_path |
Path to load the weights of a pre-trained model | None |
This script tests the trained model on a DICOM file and generates a class activation map (CAM) with an overlay on the input image.
python test.py <dicom_file_path> --input_size <input_size> <input_size> --load_path <load_path> --overlay_weight <overlay_weight>
Parameter | Description | Default Value |
---|---|---|
dicom_file_path |
Path to the DICOM file | - |
--input_size |
Input size of the images | 256 256 |
--load_path |
Path to load the trained model | ./weights |
--overlay_weight |
Weight of the CAM overlay on the image | 0.8 |
This project is licensed under the MIT License. See the LICENSE file for more information.
[PAPER] Deep-learning model associating lateral cervical radiographic features with Cormack–Lehane grade 3 or 4 glottic view
H.-Y.Cho, K.Lee. H-J.Kong, H.L-Yang. C.W. Jung, H.-P.Park, J.Y.Hwang, and H.-C.Lee
Unanticipated difficult laryngoscopy is associated with serious airway-related complications. We aimed to develop and test a convolutional neural network-based deep-learning model that uses lateral cervical spine radiographs to predict Cormack–Lehane grade 3 or 4 direct laryngoscopy views of the glottis. We analysed the radiographs of 5939 thyroid surgery patients at our hospital, 253 (4%) of whom had grade 3 or 4 glottic views. We used 10 randomly sampled datasets to train a model. We compared the new model with six similar models (VGG, ResNet, Xception, ResNext, DenseNet and SENet). The Brier score (95%CI) of the new model, 0.023 (0.021–0.025), was lower (`better´) than the other models: VGG, 0.034 (0.034–0.035); ResNet, 0.033 (0.033–0.035); Xception, 0.032 (0.031–0.033); ResNext, 0.033 (0.032–0.033); DenseNet, 0.030 (0.029–0.032); SENet, 0.031 (0.029–0.032), all p < 0.001. We calculated mean (95%CI) of the new model for: R2, 0.428 (0.388–0.468); mean squared error, 0.023 (0.021–0.025); mean absolute error, 0.048 (0.046–0.049); balanced accuracy, 0.713 (0.684–0.742); and area under the receiver operating characteristic curve, 0.965 (0.962–0.969). Radiographic features around the hyoid bone, pharynx and cervical spine were associated with grade 3 and 4 glottic views.
Figure. Convolutional neural network-based deep-learning model with a convolutional layer, pooling layer, self-attention layer, and final fully connected layer to predict difficult laryngoscopy.
If you find our work useful in your research, please consider citing our paper:
@article{cho2023deep,
title={Deep-learning model associating lateral cervical radiographic features with Cormack--Lehane grade 3 or 4 glottic view},
author={Cho, H-Y and Lee, Kyungsu and Kong, H-J and Yang, H-L and Jung, C-W and Park, H-P and Hwang, Jae Youn and Lee, H-C},
journal={Anaesthesia},
volume={78},
number={1},
pages={64--72},
year={2023},
publisher={Wiley Online Library}
}