Spatio-Temporal Classification of Lung Ventilation Patterns using 3D EIT Images: A General Approach for Individualized Lung Function Evaluation

Shuzhe Chen, Li Li, Zhichao Lin, Ke Zhang, Ying Gong, Lu Wang, Xu Wu, Yuanlin Song, Fan Yang, Shenheng XuarXiv 2023 · DOI · arXiv

Abstract

The Pulmonary Function Test (PFT) is a widely utilized and rigorous classification test for lung function evaluation, serving as a comprehensive tool for lung diagnosis. Meanwhile, Electrical Impedance Tomography (EIT) is a rapidly advancing clinical technique that visualizes conductivity distribution induced by ventilation. EIT provides additional spatial and temporal information on lung ventilation beyond traditional PFT. However, relying solely on conventional isolated interpretations of PFT results and EIT images overlooks the continuous dynamic aspects of lung ventilation. This study aims to classify lung ventilation patterns by extracting spatial and temporal features from the 3D EIT image series. The study uses a Variational Autoencoder network with a MultiRes block to compress the spatial distribution in a 3D image into a one-dimensional vector. These vectors are then concatenated to create a feature map for the exhibition of temporal features. A simple convolutional neural network is used for classification. Data collected from 137 subjects were finally used for training. The model is validated by ten-fold and leave-one-out cross-validation first. The accuracy and sensitivity of normal ventilation mode are 0.95 and 1.00, and the f1-score is 0.94. Furthermore, we check the reliability and feasibility of the proposed pipeline by testing it on newly recruited nine subjects. Our results show that the pipeline correctly predicts the ventilation mode of 8 out of 9 subjects. The study demonstrates the potential of using image series for lung ventilation mode classification, providing a feasible method for patient prescreening and presenting an alternative form of PFT.

Introduction

The current "gold standard" for assessing respiratory health is the Pulmonary Function Test (PFT). While clinically rigorous, PFT suffers from two significant limitations: a lack of spatial resolution and a reliance on population-wide demographic averages. PFT measures the total volume and velocity of air at the mouth, which obscures localized obstructions or regional ventilation disparities. Moreover, interpreting results against "predicted values" based on age, height, and ethnicity can fail to capture what is "normal" for a specific individual.

Electrical Impedance Tomography (EIT) addresses these gaps by using small electrical currents to visualize internal conductivity changes, providing non-invasive, radiation-free spatial data. However, clinical EIT has traditionally been limited to static 2D slices. This research proposes a novel deep learning pipeline that treats 3D EIT data as a dynamic time-series, capturing the full spatio-temporal signature of a breath to classify lung ventilation patterns more accurately.

Methodology

The proposed pipeline processes 3D EIT "movies" of forced expiration through several stages:

  1. 3D EIT Reconstruction: Using electrode belts, 3D lung images are reconstructed at a rate of 20 frames per second.
  2. Spatial Compression: To manage the high dimensionality of 3D data (48x32x48 pixels per frame), the researchers utilized a Variational Autoencoder (VAE) enhanced with MultiRes blocks. This network compresses each 3D frame into a compact 32-dimensional latent vector, capturing essential spatial features of lung shape and air distribution.
  3. Temporal Integration: These latent vectors are stacked chronologically to create a 2D "feature map." In this map, one axis represents the latent spatial features, while the other represents time, effectively summarizing the entire breathing cycle in a single image.
  4. Classification: A 2D Convolutional Neural Network (CNN) analyzes this feature map to classify ventilation patterns as normal or abnormal (obstructive, restrictive, or mixed).
The deep learning pipeline architecture, showing the VAE encoder-decoder structure and the MultiRes block used for spatial feature extraction.

Results and Discussion

The model was trained on a primary cohort of 137 subjects (70 normal, 67 abnormal). In binary classification tasks, the pipeline achieved an accuracy of 95.6% and an Area Under the Curve (AUC) of 0.96. For multi-class classification (Normal, Obstructive, Restrictive, Mixed), the model maintained a robust accuracy of 86.3%.

A 3D reconstruction of lung ventilation patterns (b) generated from EIT data, illustrating the spatial distribution of conductivity changes.

The model's real-world generalization was tested on a blind set of nine new subjects, where it correctly predicted the ventilation mode for eight individuals. The single error occurred in a restrictive case that the model struggled to distinguish from normal due to the patient's relatively high flow metrics. Furthermore, the researchers demonstrated that the VAE learned a physically meaningful representation of the lungs by showing smooth transitions in reconstructed images when interpolating between different lung states.

Conclusion

This study demonstrates the potential of using 3D EIT image series for individualized lung ventilation classification. By focusing on the internal dynamics of air movement rather than demographic averages, the pipeline offers a promising tool for patient prescreening and a viable alternative to traditional PFT. Future work will focus on expanding the dataset to improve multi-class performance, incorporating tidal breathing for severely ill patients, and enhancing the interpretability of the deep learning model to provide clinicians with clearer physiological insights.

Summary updated

Abstract

The Pulmonary Function Test (PFT) is an widely utilized and rigorous classification test for lung function evaluation, serving as a comprehensive tool for lung diagnosis. Meanwhile, Electrical Impedance Tomography (EIT) is a rapidly advancing clinical technique that visualizes conductivity distribution induced by ventilation. EIT provides additional spatial and temporal information on lung ventilation beyond traditional PFT. However, relying solely on conventional isolated interpretations of PFT results and EIT images overlooks the continuous dynamic aspects of lung ventilation. This study aims to classify lung ventilation patterns by extracting spatial and temporal features from the 3D EIT image series. The study uses a Variational Autoencoder network with a MultiRes block to compress the spatial distribution in a 3D image into a one-dimensional vector. These vectors are then concatenated to create a feature map for the exhibition of temporal features. A simple convolutional neural network is used for classification. Data collected from 137 subjects were finally used for training. The model is validated by ten-fold and leave-one-out cross-validation first. The accuracy and sensitivity of normal ventilation mode are 0.95 and 1.00, and the f1-score is 0.94. Furthermore, we check the reliability and feasibility of the proposed pipeline by testing it on newly recruited nine subjects. Our results show that the pipeline correctly predicts the ventilation mode of 8 out of 9 subjects. The study demonstrates the potential of using image series for lung ventilation mode classification, providing a feasible method for patient prescreening and presenting an alternative form of PFT.
BibTeX
@misc{chen2023spatio,
  author = {Shuzhe Chen and Li Li and Zhichao Lin and Ke Zhang and Ying Gong and Lu Wang and Xu Wu and Maokun Li and Yuanlin Song and Fan Yang and Shenheng Xu},
  title = {{Spatio-Temporal Classification of Lung Ventilation Patterns using 3D EIT Images: A General Approach for Individualized Lung Function Evaluation}},
  year = {2023},
  eprint = {2307.00307},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2307.00307},
}

Loading related content...

Figure