吞嚥困難患者之舌骨前移與上抬為咽喉期吞嚥中攸關氣道保護與食團清除的重要生物力學事件，然而傳統舌骨運動量測多仰賴人工逐幀標記，耗時且不易規模化。本研究旨在建立並驗證一套全自動螢光吞嚥攝影檢查舌骨運動學分析流程，並探討自動萃取之舌骨運動學特徵與吸入風險及吞嚥障礙嚴重度之關聯。
本研究回溯性納入 42 位吞嚥困難高風險患者，共 89 段液體吞嚥攝影片段。分析流程包含 digital imaging and communications in medicine (DICOM) 影像前處理、YOLOv11l 舌骨偵測、differentiable spatial-to-numerical transform (DSNT) 頸椎標記點偵測、以 C2–C4 距離標準化之頸椎相對座標轉換、自動吞嚥事件偵測，以及病人層級舌骨運動學特徵萃取。自動偵測所得吞嚥事件均由復健科醫師逐一人工複核後納入分析。
舌骨偵測模型於獨立測試集之平均 intersection over union (IoU) 為 0.834，頸椎DSNT 模型整體中位像素誤差為 3.7 px。自動吞嚥偵測以舌骨移動幅度(excursion)作為主要活動訊號。舌骨移動幅度定義為滑動時間窗內舌骨前後向與上下向相對座標變化全距之乘積，用以量化短時間內二維同步位移強度。結合 Otsu 法與一秒起始點合併後，逐事件偵測 F1 為 0.792。全樣本共偵測 437 個候選吞嚥事件，經人工複核後 302 個真實吞嚥納入後續分析。
本研究納入的 42 位患者中，因其中 1 名患者其唯一液體片段經人工複核後未擷取 到可供分析之吞嚥而被排除，故最終進行病人層級運動學分析僅 41 位患者。此 41 位患者中，被歸類為吸入者的有 17 位、非吸入者有 24 位。單變項分析顯示，吸 入者呈現整體舌骨運動幅度降低，其中又以上抬峰值位置與前向位移幅度之判別效 能最佳，area under curve (AUC) 分別為 0.716 與 0.750。兩者合併後，以邏輯斯迴 歸 (logistic regression) 建模，並經 leave-one-out cross-validation (LOO-CV) 計算之 AUC 為 0.75。嚴重度分析顯示，前向位移幅度與會厭谷殘留呈負相關，Spearman's ρ 為 −0.49。上抬峰值位置與 Penetration–Aspiration Scale (PAS) 分級呈負相關，ρ 為 −0.31。探索性波形層級分析則顯示吸入者具有較高之跨吞嚥形狀變異，且上抬 軌跡差異主要集中於吞嚥中段。
整體而言，本研究建立之全自動分析流程可有效量化螢光吞嚥攝影檢查中之舌骨運 動，並顯示吸入風險較高者具有前向位移不足與上抬高度受限之特徵。該流程具備 作為臨床吞嚥生物力學輔助評估工具之潛力，惟其對於吸入風險的診斷效度仍需更 大樣本與外部資料驗證。

Anterior displacement and superior elevation of the hyoid bone are key biomechanical events during pharyngeal swallowing and contribute to airway protection and bolus clearance. However, conventional hyoid kinematic analysis of videofluoroscopic swallowing studies (VFSS) typically requires manual frame-by-frame annotation, which is time-consuming and difficult to implement at scale. This retrospective study aimed to develop and validate a fully automated VFSS-based hyoid kinematic analysis pipeline and to examine the associations among automatically extracted hyoid kinematic features, aspiration status, and dysphagia severity.
A total of 89 liquid-swallow VFSS clips from 42 patients at high risk for dysphagia were included. The pipeline comprised digital imaging and communications in medicine (DICOM) image preprocessing, hyoid detection using YOLOv11l, cervical landmark detection using a differentiable spatial-to-numerical transform (DSNT), C2-C4-normalized cervical coordinate transformation, automated swallow-event detection, and extraction of patient-level hyoid kinematic features. All automatically detected swallow events were manually reviewed by a physiatrist before inclusion in the kinematic analysis.
The YOLOv11l model achieved a mean intersection over union (IoU) of 0.834 on the test set, whereas the DSNT cervical landmark model achieved an overall median localization error of 3.7 pixels. For swallow-event detection, excursion was defined as the product of the anterior-posterior and superior-inferior coordinate ranges within a sliding temporal window, thereby quantifying short-term two-dimensional hyoid movement intensity. When combined with Otsu's method and a one-second onset-merging rule, this approach yielded an event-level F1 score of 0.792.
Across 89 liquid-bolus swallowing video clips, 437 candidate swallow events were detected, of which 302 were confirmed as true swallows after manual review and retained for analysis. The final patient-level kinematic analysis included 41 patients, after excluding one patient whose only liquid-bolus clip yielded no valid swallow event for analysis after manual review. Among these patients, 17 were classified as aspirators and 24 as non- aspirators. In univariable analysis, aspirators demonstrated globally reduced hyoid movement amplitude. The strongest individual discriminators were peak superior position and anterior range, with areas under the receiver operating characteristic curves (AUCs) of 0.716 and 0.750, respectively. Combining these two complementary features using logistic regression yielded a leave-one-out cross-validated AUC of 0.75. Severity analyses showed that anterior range was negatively correlated with vallecular residue severity, with a Spearman's ρ of −0.49, whereas peak superior position was most strongly associated with PAS severity, with a Spearman's ρ of −0.31. Exploratory waveform-level analyses further suggested greater cross-swallow shape variability among aspirators and revealed that group differences in superior hyoid trajectory were concentrated in the mid-swallow phase.
In summary, this study developed an automated, clinician-verified VFSS-based hyoid kinematic analysis pipeline and provided preliminary evidence that reduced anterior displacement and limited superior hyoid elevation were associated with aspiration risk and dysphagia severity. The proposed pipeline may serve as an objective adjunct to biomechanical assessment of swallowing, although its diagnostic utility requires validation in larger external cohorts.

Bendsen, B. B., Jensen, D., Westmark, S., Krarup, A. L., Riis, J., & Melgaard, D. (2022). The Quality of Life in Citizens with Oropharyngeal Dysphagia-A Cross-Sectional Study. Journal of Clinical Medicine, 11(14), 4212. https://doi.org/10.3390/jcm11144212
Blanař, V., Hödl, M., Lohrmann, C., Amir, Y., & Eglseer, D. (2019). Dysphagia and factors associated with malnutrition risk: A 5-year multicentre study. Journal of Advanced Nursing, 75(12), 3566–3576. https://doi.org/10.1111/jan.14188
Brates, D., Steele, C. M., & Molfenter, S. M. (2020). Measuring Hyoid Excursion Across the Life Span: Anatomical Scaling to Control for Variation. Journal of Speech, Language, and Hearing Research, 63(1), 125–134. https://doi.org/10.1044/2019_JSLHR-19-00007
Cliff, N. (1993). Dominance statistics: Ordinal analyses to answer ordinal questions. Psychological Bulletin, 114(3), 494–509. https://doi.org/10.1037/0033-2909.114.3.494
Hossain, I., Roberts-South, A., Jog, M., & El-Sakka, M. R. (2014). Semi-automatic assessment of hyoid bone motion in digital videofluoroscopic images. Computer Methods in Biomechanics and Biomedical Engineering: Imaging & Visualization, 2(1), 25–37. https://doi.org/10.1080/21681163.2013.833859
Hsiao, M.-Y., Weng, C.-H., Wang, Y.-C., Cheng, S.-H., Wei, K.-C., Tung, P.-Y., Chen, J.- Y., Yeh, C.-Y., & Wang, T.-G. (2023). Deep Learning for Automatic Hyoid Tracking in Videofluoroscopic Swallow Studies. Dysphagia, 38(1), 171–180. https://doi.org/10.1007/s00455-022-10438-0
Khanam, R., & Hussain, M. (2024). YOLOv11: An Overview of the Key Architectural Enhancements (arXiv:2410.17725). arXiv. https://doi.org/10.48550/arXiv.2410.17725
Kim, H.-I., Kim, Y., Kim, B., Shin, D. Y., Lee, S. J., & Choi, S.-I. (2021). Hyoid Bone Tracking in a Videofluoroscopic Swallowing Study Using a Deep-Learning-Based Segmentation Network. Diagnostics, 11(7), 1147. https://doi.org/10.3390/diagnostics11071147
Kim, W.-S., Zeng, P., Shi, J. Q., Lee, Y., & Paik, N.-J. (2017). Semi-automatic tracking, smoothing and segmentation of hyoid bone motion from videofluoroscopic swallowing study. PLOS ONE, 12(11), e0188684. https://doi.org/10.1371/journal.pone.0188684
Kim, Y. H., Han, T. R., Nam, H. S., Seo, H. G., & Oh, B.-M. (2019). Temporal characteristics of laryngeal penetration and aspiration in stroke patients. NeuroRehabilitation, 44(2), 231–238. https://doi.org/10.3233/NRE-182569
Kwon, S., Cha, S., Kim, J., Han, K., Paik, N.-J., & Kim, W.-S. (2023). Trends in the incidence and prevalence of dysphagia requiring medical attention among adults in South Korea, 2006-2016: A nationwide population study. PloS One, 18(6), e0287512. https://doi.org/10.1371/journal.pone.0287512
Leonard, R. J., Kendall, K. A., McKenzie, S., Gonçalves, M. I., & Walker, A. (2000). Structural Displacements in Normal Swallowing: A Videofluoroscopic Study. Dysphagia, 15(3), 146–152. https://doi.org/10.1007/s004550010017
Li, W., Mao, S., Mahoney, A. S., Coyle, J. L., & Sejdić, E. (2024). Automatic Tracking of Hyoid Bone Displacement and Rotation Relative to Cervical Vertebrae in Videofluoroscopic Swallow Studies Using Deep Learning. Journal of Imaging Informatics in Medicine, 37(4), 1922–1932. https://doi.org/10.1007/s10278-024- 01039-4
Logemann, J. A., Pauloski, B. R., Rademaker, A. W., Colangelo, L. A., Kahrilas, P. J., & Smith, C. H. (2000). Temporal and Biomechanical Characteristics of Oropharyngeal Swallow in Younger and Older Men. Journal of Speech, Language, and Hearing Research, 43(5), 1264–1274. https://doi.org/10.1044/jslhr.4305.1264
Mandell, L. A., & Niederman, M. S. (2019). Aspiration Pneumonia. New England Journal of Medicine, 380(7), 651–663. https://doi.org/10.1056/NEJMra1714562
Medda, B. K., Kern, M., Ren, J., Xie, P., Ulualp, S. O., Lang, I. M., & Shaker, R. (2003). Relative contribution of various airway protective mechanisms to prevention of aspiration during swallowing. American Journal of Physiology-Gastrointestinal and Liver Physiology, 284(6), G933–G939. https://doi.org/10.1152/ajpgi.00395.2002
Molfenter, S. M., & Steele, C. M. (2011). Physiological Variability in the Deglutition Literature: Hyoid and Laryngeal Kinematics. Dysphagia, 26(1), 67–74. https://doi.org/10.1007/s00455-010-9309-x
Molfenter, S. M., & Steele, C. M. (2014). Use of an Anatomical Scalar to Control for Sex- Based Size Differences in Measures of Hyoid Excursion During Swallowing. Journal of Speech, Language, and Hearing Research, 57(3), 768–778. https://doi.org/10.1044/2014_JSLHR-S-13-0152
Nibali, A., He, Z., Morgan, S., & Prendergast, L. (2018). Numerical Coordinate Regression with Convolutional Neural Networks (arXiv:1801.07372). arXiv. https://doi.org/10.48550/arXiv.1801.07372
Pongpipatpaiboon, K., Inamoto, Y., Saitoh, E., Kagaya, H., Shibata, S., Aoyagi, Y., Fujii, N., Palmer, J. B., & Fernández, M. G. (2018). Pharyngeal swallowing in older adults: Kinematic analysis using three-dimensional dynamic computed tomography. Journal of Oral Rehabilitation, 45(12), 959–966. https://doi.org/10.1111/joor.12703
Romano, J., Kromrey, J. D., Coraggio, J., & Skowronek, J. (2006, February 1). Appropriate statistics for ordinal level data: Should we really be using t-test and Cohen's d for evaluating group differences on the NSSE and other surveys? Ronneberger, O., Fischer, P., & Brox, T. (2015). U-Net: Convolutional Networks for Biomedical Image Segmentation (arXiv:1505.04597). arXiv. https://doi.org/10.48550/arXiv.1505.04597
Rosenbek, J. C., Robbins, J. A., Roecker, E. B., Coyle, J. L., & Wood, J. L. (1996). A penetration-aspiration scale. Dysphagia, 11(2), 93–98. https://doi.org/10.1007/BF00417897
Ryu, Y. H., Kim, J. H., Kim, D., Kim, S. Y., & Lee, S. J. (2024). Diagnostic value of a deep learning-based hyoid bone tracking model for aspiration in patients with post-stroke dysphagia. Digital Health, 10, 20552076241271778. https://doi.org/10.1177/20552076241271778
Smaoui, S., Peladeau-Pigeon, M., & Steele, C. M. (2021). Variations in Hyoid Kinematics Across Liquid Consistencies in Healthy Swallowing. Journal of Speech, Language, and Hearing Research : JSLHR, 64(1), 51–58. https://doi.org/10.1044/2020_JSLHR-20-00508
Smaoui, S., Peladeau-Pigeon, M., & Steele, C. M. (2022). Determining the Relationship Between Hyoid Bone Kinematics and Airway Protection in Swallowing. Journal of Speech, Language, and Hearing Research : JSLHR, 65(2), 419–430. https://doi.org/10.1044/2021_JSLHR-21-00238
Steele, C. M., Bailey, G. L., Chau, T., Molfenter, S. M., Oshalla, M., Waito, A. A., & Zoratto, D. C. B. H. (2011). The relationship between hyoid and laryngeal displacement and swallowing impairment. Clinical Otolaryngology, 36(1), 30–36. https://doi.org/10.1111/j.1749-4486.2010.02219.x
Steele, C. M., Bayley, M. T., Bohn, M. K., Higgins, V., Peladeau-Pigeon, M., & Kulasingam, V. (2023). Reference Values for Videofluoroscopic Measures of Swallowing: An Update. Journal of Speech, Language, and Hearing Research, 66(10), 3804–3824. https://doi.org/10.1044/2023_JSLHR-23-00246
Steele, C. M., & Cichero, J. A. Y. (2014). Physiological Factors Related to Aspiration Risk: A Systematic Review. Dysphagia, 29(3), 295–304. https://doi.org/10.1007/s00455-014-9516-y
Steele, C. M., Peladeau-Pigeon, M., Barbon, C. A. E., Guida, B. T., Namasivayam- MacDonald, A. M., Nascimento, W. V., Smaoui, S., Tapson, M. S., Valenzano, T. J., Waito, A. A., & Wolkin, T. S. (2019). Reference Values for Healthy Swallowing Across the Range From Thin to Extremely Thick Liquids. Journal of Speech, Language, and Hearing Research, 62(5), 1338–1363. https://doi.org/10.1044/2019_JSLHR-S-18-0448
Tanrıverdi, M., Heybeli, C., Çalım, Ö. F., Durna, M., Özturan, O., & Soysal, P. (2024). The relationship between oropharyngeal dysphagia and dehydration in older adults. BMC Geriatrics, 24(1), 885. https://doi.org/10.1186/s12877-024-05492-2
Van Der Kruis, J. G. J., Baijens, L. W. J., Speyer, R., & Zwijnenberg, I. (2011). Biomechanical Analysis of Hyoid Bone Displacement in Videofluoroscopy: A Systematic Review of Intervention Effects. Dysphagia, 26(2), 171–182. https://doi.org/10.1007/s00455-010-9318-9
Wei, K.-C., Cheng, S.-H., Hsiao, M.-Y., Wang, Y.-C., Weng, C.-H., Chen, J.-Y., & Wang, T.-G. (2022). Swallowing kinematic analysis might be helpful in predicting aspiration and pyriform sinus stasis. Scientific Reports, 12(1), 1354. https://doi.org/10.1038/s41598-022-05441-2
Wei, K.-C., Hsiao, M.-Y., & Wang, T.-G. (2022). The kinematic features of hyoid bone movement during swallowing in different disease populations: A narrative review. Journal of the Formosan Medical Association, 121(10), 1892–1899. https://doi.org/10.1016/j.jfma.2022.04.007
Yagis, E., Atnafu, S. W., García Seco de Herrera, A., Marzi, C., Scheda, R., Giannelli, M., Tessa, C., Citi, L., & Diciotti, S. (2021). Effect of data leakage in brain MRI classification using 2D convolutional neural networks. Scientific Reports, 11(1),22544. https://doi.org/10.1038/s41598-021-01681-w
Youden, W. J. (1950). Index for rating diagnostic tests. Cancer, 3(1), 32–35. https://doi.org/10.1002/1097-0142(1950)3:1<32::aid- cncr2820030106>3.0.co;2-3
Zhang, Z., Coyle, J. L., & Sejdić, E. (2018). Automatic hyoid bone detection in fluoroscopic images using deep learning. Scientific Reports, 8(1), 12310. https://doi.org/10.1038/s41598-018-30182-6
Zhao, Y., Lv, W., Xu, S., Wei, J., Wang, G., Dang, Q., Liu, Y., & Chen, J. (2024). DETRs Beat YOLOs on Real-time Object Detection. 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 16965–16974. https://doi.org/10.1109/CVPR52733.2024.01605
Zuiderveld, K. (1994). VIII.5. - Contrast Limited Adaptive Histogram Equalization. In P. S. Heckbert (Ed.), Graphics Gems (pp. 474–485). Academic Press. https://doi.org/10.1016/B978-0-12-336156-1.50061-6