Loading [a11y]/accessibility-menu.js
Vision Intelligence Assisted Lung Function Estimation Based on Transformer Encoder–Decoder Network With Invertible Modeling | IEEE Journals & Magazine | IEEE Xplore

Vision Intelligence Assisted Lung Function Estimation Based on Transformer Encoder–Decoder Network With Invertible Modeling


Impact Statement:Invertible neural networks possess unique advantages over other generative models due to their full reversibility and minimal information loss. This article introduces an...Show More

Abstract:

Lung function evaluation is important to many medical applications, but conducting pulmonary function tests is constrained by different conditions. This article presents ...Show More
Impact Statement:
Invertible neural networks possess unique advantages over other generative models due to their full reversibility and minimal information loss. This article introduces an invertible model to represent the bidirectional relationship between lung function and CT scans with limited labeled data, meanwhile minimizing the computational resource requirement. Moreover, this approach can be extended to address similar image regression problems and their corresponding generative tasks. Additionally, it offers a streamlined process for utilizing NF models, simplifying their usage without compromising accuracy and reducing computational overhead.

Abstract:

Lung function evaluation is important to many medical applications, but conducting pulmonary function tests is constrained by different conditions. This article presents a pioneer study of an integrated invertible deep learning method for lung function estimation via using computed tomography (CT) images. First, the projection method is proposed to flatten the three-dimensional (3-D) image onto a two-dimensional (2-D) plane, with preserving location information in 3-D. Next, the MBConv transformer-based encoder–decoder structure is developed to extract latent features. Finally, we develop an invertible normalizing flow (NF) model to infer lung function based on the extracted features and design two loss functions for two directions. The method enables both estimating the lung function based on CT images and metadata as well as generating the corresponding simulated CT image according to the lung function. Computational studies show that the proposed regression model outperforms all sta...
Published in: IEEE Transactions on Artificial Intelligence ( Volume: 5, Issue: 7, July 2024)
Page(s): 3336 - 3349
Date of Publication: 01 January 2024
Electronic ISSN: 2691-4581

Funding Agency:


References

References is not available for this document.