Spatial resolution enhancement using deep learning improves chest disease diagnosis based on thick slice CT (npj Digital Medicine)

Pengxin Yu, Haoyue Zhang, Dawei Wang, Rongguo Zhang, et al.

Abstract

CT is crucial for diagnosing chest diseases, with image quality affected by spatial resolution. Thick-slice CT remains prevalent in practice due to cost considerations, yet its coarse spatial resolution may hinder accurate diagnoses. Our multicenter study develops a deep learning synthetic model with Convolutional-Transformer hybrid encoder-decoder architecture for generating thin-slice CT from thick-slice CT on a single center (1576 participants) and access the synthetic CT on three cross-regional centers (1228 participants). The qualitative image quality of synthetic and real thin-slice CT is comparable (p= 0.16). Four radiologists’ accuracy in diagnosing community-acquired pneumonia using synthetic thin-slice CT surpasses thick-slice CT (p < 0.05), and matches real thin-slice CT (p > 0.99). For lung nodule detection, sensitivity with thin-slice CT outperforms thick-slice CT (p < 0.001) and comparable to real thin-slice CT (p > 0.05). These ndings indicate the potential of our model to generate high-quality synthetic thin-slice CT as a practical alternative when real thin-slice CT is preferred but unavailable.

Network Architecture

Visual Comparison Results

Code

Model train

python train.py train --path_key HD --gpu_idx 0 --model t3dv1 --net_idx CTH_net

Model val

python val.py val --path_key HD --gpu_idx 0 --model t3dv1 --net_idx CTH_net

Model test

python test.py test --path_key HD --gpu_idx 0 --model t3dv1 --net_idx CTH_net

Model

The well-trained model parameters can be downloaded in baidu cloud disk.
The model path is /model/xxx.pkl.

Data

The data used for this study are not publicly available due to hospital privacy restrictions.
In our previous paper, we made public a dataset RPLHR-CT, which contains 250 cases of data.
We build a tiny dataset with 40 cases to baidu cloud disk.
For data usage rights reasons, should you required the complete dataset, please contact ypengxin@infervision.com.

For this code, the data should be organized as follows.

data/
|-- HD_1mm/
|  |-- A.nii.gz
|  |-- B.nii.gz
|  |-- C.nii.gz
|-- HD_5mm/
|  |-- A.nii.gz
|  |-- B.nii.gz
|  |-- C.nii.gz

Citation

If you use our code or data, please consider citing:

@article{yu2024spatial,
  title={Spatial resolution enhancement using deep learning improves chest disease diagnosis based on thick slice CT},
  author={Yu, Pengxin and Zhang, Haoyue and Wang, Dawei and Zhang, Rongguo and Deng, Mei and Yang, Haoyu and Wu, Lijun and Liu, Xiaoxu and Oh, Andrea S and Abtin, Fereidoun G and others},
  journal={npj Digital Medicine},
  volume={7},
  number={1},
  pages={335},
  year={2024},
  publisher={Nature Publishing Group UK London}
}
@inproceedings{yu2022rplhr,
  title={RPLHR-CT Dataset and Transformer Baseline for Volumetric Super-Resolution from CT Scans},
  author={Yu, Pengxin and Zhang, Haoyue and Kang, Han and Tang, Wen and Arnold, Corey W and Zhang, Rongguo},
  booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention},
  pages={344--353},
  year={2022},
  organization={Springer}
}

Acknowledgment: This code is based on the RPLHR-CT Dataset and Transformer Baseline for Volumetric Super-Resolution from CT Scans.