Automatic Detection of Airway Invasion from Videofluoroscopy via Deep Learning Technology
Open accessThe DOI matches its Crossref record; 32 of the 32 references it cites were found in a registry or an index.
Abstract
In dysphagia, food materials frequently invade the laryngeal airway, potentially resulting in serious consequences, such as asphyxia or pneumonia. The VFSS (videofluoroscopic swallowing study) procedure can be used to visualize the occurrence of airway invasion, but its reliability is limited by human errors and fatigue. Deep learning technology may improve the efficiency and reliability of VFSS analysis by reducing the human effort required. A deep learning model has been developed that can detect airway invasion from VFSS images in a fully automated manner. The model consists of three phases: (1) image normalization, (2) dynamic ROI (region of interest) determination, and (3) airway invasion detection. Noise induced by movement and learning from unintended areas is minimized by defining a “dynamic” ROI with respect to the center of the cervical spinal column as segmented using U-Net. An Xception module, trained on a dataset consisting of 267,748 image frames obtained from 319 VFSS video files, is used for the detection of airway invasion. The present model shows an overall accuracy of 97.2% in classifying image frames and 93.2% in classifying video files. It is anticipated that the present model will enable more accurate analysis of VFSS data.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“Videofluoroscopy of swallowing was used as the reference examination in most studies.”
From Artificial intelligence in the diagnosis and management of dysphagia: a scoping review (2025) · cites this paper as [23]
In the citing paper
“Studies on AI in dysphagia primarily rely on imaging resources such as VFSS for comparative analysis due to its high reliability(16,23-28).”
From Artificial intelligence in the diagnosis and management of dysphagia: a scoping review (2025) · cites this paper as [23]
Citations
Cites
Deep Residual Learning for Image Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.90 · Cited by 235,474 · cited 2×
“U-Net is an end-to-end FCN (fully connected network)-based model consisting of a convolutional encoder and a convolutional decoder connected by skip connections [24].”
MethodU-Net: Convolutional Networks for Biomedical Image Segmentation
Ronneberger, O., Fischer, P., & Brox, T. (2015) · Lecture notes in computer science · doi:10.1007/978-3-319-24574-4_28 · Cited by 99,404
Rethinking the Inception Architecture for Computer Vision
Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J., & Wojna, Z. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.308 · Cited by 32,138
Focal Loss for Dense Object Detection
Lin, T.-Y., Goyal, P., Girshick, R., He, K., & Dollar, P. (2017) · 2017 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2017.324 · Cited by 30,581 · cited 1×
“To alleviate the problem of data imbalance between these classes, we used the focal loss function [31].”
MethodPan, S. J., & Yang, Q. (2010) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2009.191 · Cited by 23,387 · cited 2×
“To effectively train the network on a small amount of data, we used the transfer learning [25] technique.”
MethodXception: Deep Learning with Depthwise Separable Convolutions
Chollet, F. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.195 · Cited by 20,380 · cited 2×
“Therefore, among the models that can be effectively trained on limited data, in this paper, we choose a CNN based on an Xception module [27].”
MethodContrast Limited Adaptive Histogram Equalization
Zuiderveld, K. (1994) · Graphics Gems · doi:10.1016/b978-0-12-336156-1.50061-6 · Cited by 4,192 · cited 1×
“Therefore, to ensure that all other conditions in all images are identical and that the convolution filters in the deep learning network are effectively trained, we first normalized all of the input images using CLAHE (contrast limited adaptive histogram equalization) [22].”
MethodA penetration-aspiration scale
Rosenbek, J. C., Robbins, J. A., Roecker, E. B., Coyle, J. L., & Wood, J. L. (1996) · Dysphagia · doi:10.1007/bf00417897 · Cited by 2,803 · cited 1×
“…was found to be misinterpreted as normal, and all false negative images showed penetration corresponding to a PAS (penetration aspiration score) of two, in which the bolus invades the laryngeal vestibule, but does not cross the true vocal folds and is expelled immediately after swallowing [35].”
BackgroundAn overview of deep learning in medical imaging focusing on MRI
Lundervold, A. S., & Lundervold, A. (2019) · Zeitschrift für Medizinische Physik · doi:10.1016/j.zemedi.2018.11.002 · Cited by 2,058 · cited 1×
“Recently, advances in deep learning technology have dramatically improved the accuracy of image classification by computers to a level exceeding that of human eyes [17].”
BackgroundBag of Tricks for Image Classification with Convolutional Neural Networks
He, T., Zhang, Z., Zhang, H., Zhang, Z., Xie, J., & Li, M. (2019) · 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2019.00065 · Cited by 1,680 · cited 1×
“1 ∗ b/256 by using the linear scaling learning rate method [30] to determine the learning rate in accordance with the batch size (b is 32 for the U-Net model in Section 2.”
Method
Research works printed but not linked: 4
- Simonyan, K., and Zisserman, A. (2014). Very deep convolutional networks for large-scale image recognition. arXiv.
- Ioffe, S., and Szegedy, C. (2015). Batch normalization: Accelerating deep network training by reducing internal covariate shift. arXiv.
- Chen, L.C., Papandreou, G., Schroff, F., and Adam, H. (2017). Hartwig Rethinking atrous convolution for semantic image segmentation. arXiv.
- Zoph, B., Ghiasi, G., Lin, T.Y., Cui, Y., Liu, H., Cubuk, E.D., and Le, Q.V. (2020). Rethinking pre-training and self-training. arXiv.
Cited by
Kim, H.-I., Kim, Y., Kim, B., Shin, D. Y., Lee, S. J., & Choi, S. (2021) · Diagnostics · doi:10.3390/diagnostics11071147 · Cited by 35
“…D Rx , D Ry ) of ( H x , H y ) to ( C x , C y ) from the center point ( C x , C y ) and the position of the hyoid bone ( H x , H y ) as follows [21,25]: where D Rx and D Ry are the distances in the horizontal and vertical directions between the center points of the hyoid bone and the cervical…”
BackgroundSelf-citationJeong, S. Y., Kim, J. M., Park, J. E., Baek, S. J., & Yang, S. N. (2023) · Scientific Reports · doi:10.1038/s41598-023-44802-3 · Cited by 31
“Hence, studies have been conducted using deep learning to analyze VFSS videos 22–26 .”
BackgroundMethodGirardi, A. M., Cardell, E. A., & Bird, S. P. (2023) · Big Data and Cognitive Computing · doi:10.3390/bdcc7040178 · Cited by 16
Dysphagia foods: Current research, technological innovations, and future directions
Dong, Y., Dai, J., Han, J., Zhang, Y., Xu, Z., Jiang, L., Lan, T., & Sui, X. (2025) · Trends in Food Science & Technology · doi:10.1016/j.tifs.2025.105330 · Cited by 18
Jeong, C., Lee, C.-S., Lim, D.-W., Noh, S.-H., Moon, H. K., Park, C., & Kim, M.-S. (2024) · Brain Sciences · doi:10.3390/brainsci14060546 · Cited by 16
Shu, K., Mao, S., Zhang, Z., Coyle, J. L., & Sejdić, E. (2024) · Computer Methods and Programs in Biomedicine · doi:10.1016/j.cmpb.2024.108505 · Cited by 16
Kim, Y., Kim, H.-I., Park, G. S., Kim, S. Y., Choi, S.-I., & Lee, S. J. (2021) · Journal of Clinical Medicine · doi:10.3390/jcm10122681 · Cited by 12
“The video files were examined for the presence of laryngeal penetration or aspiration using the computer model described in a previous study [20].”
MethodSelf-citationArtificial Intelligence in Videofluoroscopy Swallow Study Analysis: A Comprehensive Review
Sanjeevi, G., Gopalakrishnan, U., Pathinarupothi, R. K., & Iyer, K. S. (2025) · Dysphagia · doi:10.1007/s00455-025-10812-8 · Cited by 12
“Key research findings: ● Lee et al. [47] achieved state-of-the-art results in their study on penetration-aspiration detection (Accuracy: 97.2%).”
InfluentialNam, K., Lee, C., Lee, T., Shin, M., Kim, B. H., & Park, J. (2024) · Diagnostics · doi:10.3390/diagnostics14131444 · Cited by 10
“The remarkable performance of previous studies may be attributed to the extraction of image characteristics through multiple convolutions with various kernels specialized in detecting single pathological events such as aspiration [10].”
BackgroundMethodMachine learning in the evaluation of voice and swallowing in the head and neck cancer patient
Srinivasan, Y., Liu, A. Y., & Rameau, A. (2023) · Current Opinion in Otolaryngology & Head & Neck Surgery · doi:10.1097/moo.0000000000000948 · Cited by 10
Fakhry, A., Antony, S. M., Park, E., & Lee, J. T. (2025) · IEEE Access · doi:10.1109/access.2025.3573282 · Cited by 8
“Lee et al. [37] proposed a two-stage deep learning framework for this task, combining ROI localization with frame-level classification.”
Kwak, S., Kim, J. K., Moon, J. S., Lee, G. W., Kim, S., & Chang, M. C. (2025) · Scientific Reports · doi:10.1038/s41598-025-10397-0 · Cited by 8
“While sensitivity for penetration and aspiration remained low, the high accuracy in detecting normal swallowing can potentially reduce the time for interpreting VFSS and support clinical decision-making in dysphagia management 18,27 .”
Artificial intelligence in the diagnosis and management of dysphagia: a scoping review
Silva, R. D. D., Almeida, S. B., Gonçalves, F. M., Zeigelboim, B. S., Stechman-Neto, J., Schroder, A. G. D., Nascimento, W. V., Santos, R. S., & Araujo, C. M. D. (2025) · CoDAS · doi:10.1590/2317-1782/e20240305en · Cited by 6
“Videofluoroscopy swallowing study (VFSS) was utilized in studies (16,23-28) , with four of them concurrently using high-resolution manometry (27,29-31) , only two studies (32,33) used fiberoptic endoscopic evaluation of swallowing (FEES), and 2 studies reported electromyography use (3,17) .”
Lai, C.-J., Jhuang, J., Tu, Y., Liu, C.-Y., Ching-Ying, T. S. A. I., & Chien, K. (2024) · Journal of Nursing Research · doi:10.1097/jnr.0000000000000608 · Cited by 6
Artificial intelligence in dysphagia since the 21st century: a bibliometric and visualization study
Liu, T., Rong, Y., Li, D., Zhang, H., Li, B., Cui, G., & Shang, S. (2025) · Frontiers in Medicine · doi:10.3389/fmed.2025.1624381 · Cited by 5
Bandini, A., Lasala, A., Péladeau-Pigeon, M., Dharmarathna, I., Gandhi, P., Meng, D., Milne, H., Panes, V., Simmons, M. M., & Steele, C. M. (2025) · Computers in Biology and Medicine · doi:10.1016/j.compbiomed.2025.111361 · Cited by 3
Rapid Detection of Penetration-Aspiration from Fluoroscopic Videos in Dysphagia Patients
Sanjeevi, G., Gopalakrishnan, U., Pathinarupothi, R. K., Arya, C. J., & Subramania Iyer, K. (2023) · IEEE International Conference on Bioinformatics and Bioengineering (BIBE) · doi:10.1109/bibe60311.2023.00073 · Cited by 2
Evaluation and Management of Dysphagia Based on Digital Health Technologies
Lee, W. H. (2021) · Journal of the Korean Dysphagia Society · doi:10.34160/jkds.2021.11.2.004 · Cited by 2
Jeong, C., Lim, D.-W., Noh, S.-H., Moon, H., Park, C., Ko, N., & Kim, M. (2025) · Diagnostics · doi:10.3390/diagnostics16010045 · Cited by 1
“A deep convolutional neural network (CNN)-based classifier was designed to identify the occurrence of airway invasion using a total of 179 swallowing motion videos manually being separated, and this model was reported to achieve 74.2% recall, 97.2% accuracy, and 59.1% precision [20].”
InfluentialClassification of dysphagia severity after lateral medullary infarction with deep learning
Lee, T., Kim, B. H., Nam, K., & Park, J.-W. (2026) · Scientific Reports · doi:10.1038/s41598-026-40751-9 · Cited by 1
Artificial intelligence tools for the assessment and management of dysphagia: a scoping review
Sreedevi, E. V., Subramania Iyer, K., Thankappan, K., Roe, J. W. G., Karuveettil, V., Janakiram, C., Guntha, R., Sanjeevi, G., Menon, J., & Pathinarupothi, R. K. (2026) · BMJ Open · doi:10.1136/bmjopen-2026-119582 · Cited by 1
Lee, S., Ni, J.-C., Shen, Y.-C., Ku, H.-C., Yang, C., Huang, K.-W., & Chen, C.-H. (2023) · Communications in computer and information science · doi:10.1007/978-3-031-42430-4_19 · Cited by 1
Self-citationAutomated Oropharyngeal Dysphagia Assessment with Mask R-CNN and Kinematic Measures
Riaz, Z., Dilawari, A., Iqbal, S., & Alyahya, A. (2025) · Scientific Journal of King Faisal University Basic and Applied Sciences · doi:10.37575/b/med/250007
da Rocha, M. M. L., van der Molen, L., Neijman, M., van Alphen, M. J. A., van den Brekel, M. W. M., & Siepel, F. J. (2025) · International Journal of Computer Assisted Radiology and Surgery · doi:10.1007/s11548-025-03556-1 · Cited by 1
Abdolmotalleby, H., Reinhardt, J. M., & Daele, D. J. V. (2025) · Diagnostics · doi:10.3390/diagnostics15243180
“For instance, several image-based methods extract features from manually identified frames: Lee et al. [12] employed a Xception-based CNN achieving 97.2% frame-level accuracy and 93.2% video-level accuracy, while Kim et al. [8] used a U-Net architecture to classify scans based on a coarse…”
InfluentialSmart Swallows: The Role of Artificial Intelligence in Dysphagia Evaluation and Treatment
Vora, J. (2025) · Journal of Indian Speech Language & Hearing Association · doi:10.4103/jisha.jisha_22_25
Chuang, W.-K., Lin, B., Lee, Y., Su, P.-H., Kao, Y., & Lu, C.-F. (2026) · Archives of Physical Medicine and Rehabilitation · doi:10.1016/j.apmr.2026.03.016
Metadata
- Authors
- Lee, S. J., Ko, J. Y., Kim, H. I., & Choi, S.-I.
- Affiliations
- Dankook University Hospital (KR)
- Dankook University (KR)
- Corresponding author
- Choi, Sang-Il
- Venue
- Applied Sciences
- Publisher
- MDPI AG
- Published
- 2020-09-05
- Volume
- 10
- Issue
- 18
- Pages
- 6179
- ISSN
- 2076-3417 (ISSN-L)
- OpenAlex
- W3083842555
- Cited by
- 28
- References
- 32
- Citations by year
- 2026: 3 · 2025: 11 · 2024: 4 · 2023: 7 · 2021: 3
- FWCI
- 3.55
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- Locations
- Applied Sciences (publishedVersion, cc-by)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Funders
- National Research Foundation of Korea (2018R1A2B6001400, 2018R1D1A3B07049300)
- Topics
- Health Sciences › Health Professions › Speech and Hearing › Dysphagia Assessment and Management
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Medical Image Segmentation Techniques
- Physical Sciences › Computer Science › Artificial Intelligence › AI in cancer detection
- Keywords
- videofluoroscopic swallowing study, dysphagia, laryngeal airway, region of interest segmentation, U-Net, Xception, image denoising
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.