Pith. sign in

REVIEW 2 major objections 6 minor 57 references

Open-Source Manually Annotated Vocal Tract Database for Automatic Segmentation from 3D MRI Using Deep Learning: Benchmarking 2D and 3D Convolutional and Transformer Networks

T0 review · 2 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A new open-source dataset of 53 manually annotated 3D vocal tract MRI volumes from 10 French speakers, benchmarked across four deep learning architectures, shows that 3D U-Nets—including a transfer-learned model trained on only 20…

desk verdict A genuinely useful open-source vocal tract MRI dataset with a benchmark that needs a corrected evaluation before the numbers can be trusted. read the letter →

arxiv 2501.06229 v2 pith:KQ3QJJS3 submitted 2025-01-08 cs.CV cs.SDeess.AS

Subin Erattakulangara , Karthika Kelat , Katie Burnham , Rachel Balbi , Sarah E. Gerard , David Meyer , Sajan Goud Lingala This is my paper · ORCID
classification cs.CVcs.SDeess.AS
keywords vocaltractsegmentation3DMRIdeeplearningU-NettransferSTAPLEopen-sourcedatasetspeech
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper establishes that a small, open-source set of manually annotated 3D vocal tract MRI volumes is enough to train automatic segmentation models for speech research. The authors annotated 53 volumes from 10 French speakers covering 21 phonemes and 3 voiceless tasks, and benchmarked four deep learning architectures against a STAPLE consensus of three human annotators. The 3D U-Net and a transfer-learned 3D U-Net both reached an average Dice coefficient of 0.896, with the transfer-learned version using fewer than half the training volumes (20 vs 45). The 2D slice-by-slice U-Net scored lower (0.823), and the transformer-based 3D U-NetR produced frequent non-anatomical segmentations. If these results hold, researchers can automatically extract vocal tract geometry from MRI at scale, enabling larger studies of speech, singing, and voice disorders.

What carries the argument

The load-bearing mechanism is the transfer-learned 3D U-Net: a three-dimensional convolutional encoder–decoder architecture pre-trained to segment lungs in chest CT, then fine-tuned on 20 annotated vocal tract volumes with the early layers frozen. This setup lets the model reuse low-level biomedical image features from CT while learning upper-airway anatomy from a small target dataset. The evaluation machinery is the STAPLE algorithm, which fuses three independent manual segmentations into a probabilistic reference that mitigates inter-annotator variability when scoring test volumes.

What would settle it

Measure the pairwise Dice agreement among the three annotators on the eight test volumes; if the annotators' mutual agreement is comparable to or lower than the models' 0.896 Dice against the STAPLE consensus, the reported scores reflect the consensus construction rather than true anatomical accuracy. A second check: re-annotate the test volumes with an independent team and compare the resulting STAPLE consensus segmentations; large disagreement would invalidate the reference.

Watch

Extended reading notes

Core claim

The central discovery is that 3D convolutional U-Nets, one trained from scratch on 45 annotated volumes and one pre-trained on lung CT and fine-tuned on 20 volumes, produce vocal tract segmentations that agree with a multi-annotator STAPLE reference at an average Dice coefficient of 0.896 (±0.05 and ±0.04). Transfer learning from an unrelated medical imaging domain (lung CT) gives the same accuracy with less than half the training data, while the 2D slice-by-slice U-Net and the transformer-based U-NetR lag behind in accuracy and anatomical plausibility. The paper also releases the dataset and code, with eight test volumes accompanied by three independent manual segmentations and their STAPLE consensus.

Load-bearing premise

The results stand on the assumption that the STAPLE consensus of three annotators is an accurate reference for the vocal tract airspace, even though the annotation task is genuinely ambiguous near teeth and glottal landmarks and one of the annotators belongs to the same team that produced the training labels.

Editorial extensions

If this is right

  • Automatic segmentation of 3D vocal tract MRI is achievable with only about 20 manually annotated volumes when starting from a lung CT pre-trained model.
  • The released dataset provides the first open-source labeled 3D vocal tract volumes from French speakers, extending speech MRI resources beyond English.
  • Researchers can use the trained models to compute 3D vocal tract area functions and mid-sagittal shape changes from new MRI data without manual tracing.
  • The comparison suggests 3D convolutional networks are better suited than transformer U-Nets for this segmentation task with limited training data.
  • The models' consistent errors near teeth and the glottis define the specific anatomical regions where further methodological work is needed.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the transfer-learning result generalizes across scanners and languages, then new vocal tract segmentation studies could be launched with just a handful of annotated volumes, making large-cohort articulatory research far cheaper.
  • The frequent mis-segmentation near teeth and the hard palate points toward a concrete testable extension: combining these models with zero-echo-time MRI or bone priors should reduce the 'islands of airspace' errors the authors describe.
  • The dataset's French phoneme inventory, paired with the released code, could support cross-linguistic comparisons of vocal tract geometry that English-only datasets cannot, and could serve as a pretraining resource for other upper-airway segmentation tasks.
  • A direct follow-up measurement the paper does not report: the pairwise Dice agreement among the three annotators on the test volumes. That number would show how much of the models' 0.896 score is limited by human rater variability.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

2 major / 6 minor

Summary. The paper introduces an open-source dataset of 53 manually annotated 3D vocal tract MRI volumes from 10 French speakers and benchmarks four deep learning segmentation architectures (2D slice-by-slice U-Net, 3D U-Net, 3D U-Net with transfer learning from lung CT, and 3D transformer U-NetR). The central empirical claim is that the 3D U-Net and the transfer-learned 3D U-Net achieve the highest Dice coefficients (0.896 ± 0.05 and 0.896 ± 0.04, respectively) on an eight-volume STAPLE-based test set, with transfer learning matching the 3D U-Net while using fewer than half the training volumes. The authors release the manual annotations and the model code, which supports reproducibility.

Significance. If the findings hold after correcting the evaluation issues, the paper provides a valuable community resource: 53 manually annotated 3D vocal tract volumes with STAPLE consensus and public training code. The empirical comparison of 2D versus 3D convolution versus transformer architectures is directly relevant to the speech-MRI community, and the transfer-learning result is practically meaningful for small-data settings. The authors are to be credited for releasing the data and code. However, the benchmark's numerical claims are currently compromised by the STAPLE reference dependence and by contradictory Hausdorff distance reporting; the dataset contribution is independent of these issues, but the accuracy claims need revision.

major comments (2)
  1. [Methods, Data Annotation and Evaluation & Metrics; Discussion] The test reference used for benchmarking is not independent of the training-label style. In the Methods section, the STAPLE reference for the eight test volumes is built from three annotators, and 'segmentations from a third annotator were those provided by the voice-science team as described above'—the same team (two voice-science graduate students plus D. Meyer) that produced the training labels. Because the models are trained to reproduce that team's segmentation conventions, including a member of that team in the test consensus can inflate apparent accuracy even if the two BME graduate-student annotators are independent. The Discussion itself concedes that 'STAPLE can be sensitive to systematic errors; if annotators make consistent mistakes, the algorithm may reinforce rather than correct them.' Given the known ambiguity near teeth (MRI signal voids) and glottal landmarks, this is not a hypothetical concern. Please re-evaluate the test set against a STAPLE consensus of only the two independent annotators, or report pairwise inter-annotator Dice and Hausdorff values to demonstrate that the shared team's segmentations do not dominate the consensus.
  2. [Results, Table 2 paragraph] The Hausdorff distance results are internally contradictory. The text states that the 2D slice-by-slice U-Net achieved an average HD of 11.3 ± 5.4, 'lower than the 3D U-Net (14 ± 28) and the transfer learning 3D U-Net (15 ± 24.9)', and then immediately states that 'The 3D U-Net transfer learning model showed the best HD distance with an average of (3.95 + 5.2)'. Both statements cannot be correct, and the discrepancy directly affects the conclusion about which models have lower boundary variability and which is 'best' in HD. Please correct the numbers in Table 2 and the text, and re-derive any conclusions about model ranking and variability after the correction.
minor comments (6)
  1. [Results, Figure 4 caption] The caption states that 'The four sounds are labeled' but then lists five labels (/f/, /l/, 'UP', /k/, /a/); please correct the count, for example to 'four sounds and one voiceless posture'.
  2. [Abstract and Table 1A] The phoneme /kõn/ in the Abstract is written as 'kon' in Table 1A and as /k/ in Figure 4; please use a consistent phonetic transcription throughout.
  3. [Methods, Data Annotation] Please clarify whether the 20 volumes used for the transfer-learning model are a subset of the 45 training volumes and how they were selected; the current text only states that 'only 20 samples from the French speaker dataset were utilized.'
  4. [Methods, Implementation and hyper-parameter tuning] Please report the final selected hyperparameters for each architecture (number of epochs, steps per epoch, learning rate, dropout, and frozen layers for the transfer-learning model) in the text or a supplementary table; listing only the search ranges is insufficient for reproducibility.
  5. [Results, Table 2] The large HD standard deviations (e.g., 14 ± 28 and 15 ± 24.9) suggest the presence of strong outliers; consider reporting the median and interquartile range in addition to mean ± SD, or identify the outlier volumes, to give a more robust summary of boundary errors.
  6. [References] References 40 and 45 cite the same STAPLE paper, and references 35 and 53 are also duplicates; please consolidate duplicate references.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the paper is an empirical benchmark and dataset release; reported numbers are measured on held-out test volumes, not derived from fitted inputs or defined in terms of the target result.

full rationale

This paper makes no first-principles derivation claim. Its central contributions are a newly released set of 53 manually annotated 3D vocal tract MRI volumes and an empirical comparison of four deep-learning architectures on eight held-out test volumes. The reported Dice, Hausdorff, and SSIM values are actual outputs of trained networks evaluated against STAPLE reference segmentations; no fitted parameter is relabeled as a prediction, and no equation in the paper defines a predicted quantity in terms of the input that generated it. Self-citations appear in the introduction and in the transfer-learning design (e.g., reference [54] is cited for the choice of 20 retraining samples), but these citations motivate experimental choices; they do not serve as proof of the benchmark results, which are measured independently in this study. The STAPLE test reference includes one annotator team whose segmentation conventions also produced the training labels, and the Discussion concedes that STAPLE can reinforce systematic errors; this is a real evaluation-validity limitation and should be addressed by reporting annotator-specific agreement, but it is not circular in the defined sense because the model predictions are not equal to the reference by construction and the dataset contribution stands independently. The inconsistent Hausdorff values reported for the transfer-learning model (15 ± 24.9 versus 3.95 ± 5.2 in the same Results section) are likewise a reporting/correctness concern, not evidence of circularity. Under the rule that circularity requires a specific reduction to inputs by definition or by self-citation chain, no such step is present.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

The central results are empirical. The main input is manual annotation quality, with no derived constants and no new physical entities. The benchmark rests on the assumption that the manual labels define the true vocal tract airspace despite known MRI ambiguities.

assumptions (3)
  • domain assumption Manual segmentations by the voice-science team are anatomically correct enough to serve as ground truth.
    The training labels and one of the three test annotators come from this team, and the paper acknowledges ambiguity near teeth and glottis without quantitative validation of label accuracy.
  • domain assumption The STAPLE consensus of three annotators approximates the true vocal tract segmentation.
    STAPLE is used to construct the evaluation reference; its validity depends on annotator independence and correctness, which is partially weakened by shared annotation team style.
  • domain assumption Preprocessing (cropping, resampling to 256x256x32, gradient anisotropic diffusion) preserves vocal tract airspace sufficiently.
    Volumes are resampled from 320x290x36 to 256x256x32, altering resolution and potentially affecting boundary accuracy, but the paper does not quantify this effect.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Open-Source Manually Annotated Vocal Tract Database for Automatic Segmentation from 3D MRI Using Deep Learning: Benchmarking 2D and 3D Convolutional and Transformer Networks." pith.science (2026). https://pith.science/paper/KQ3QJJS3

@misc{pith2026250106229,
  author       = {Pith},
  title        = {Pith review of: Open-Source Manually Annotated Vocal Tract Database for Automatic Segmentation from 3D MRI Using Deep Learning: Benchmarking 2D and 3D Convolutional and Transformer Networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KQ3QJJS3}},
  note         = {Machine review of arXiv:2501.06229}
}
read the original abstract

Accurate segmentation of the vocal tract from magnetic resonance imaging (MRI) data is essential for various voice and speech applications. Manual segmentation is time intensive and susceptible to errors. This study aimed to evaluate the efficacy of deep learning algorithms for automatic vocal tract segmentation from 3D MRI.

Figures

Figures reproduced from arXiv: 2501.06229 by the authors.

Figure 1
Figure 1. Sample images from the training and testing datasets. The first column displays a [PITH_FULL_IMAGE:figures/full_fig_p023_1.png] view at source ↗
Figure 3
Figure 3. showcases the diverse neural network architectures employed in this study. Subfigure 3(a) illustrates the 2D slice-by-slice U-Net utilized for segmentation generation from 2D slices, featuring 2D convolution operations. Subfigure 3(b) displays the 3D U-Net architecture incorporating 3D convolutions, operating on 3D volumes as input. Subfigure 3(c) presents a customized version of the standard 3D U-Net, where initial… view at source ↗
Figure 4
Figure 4. Mid-sagittal representations of the five test volumes used in the experiment are depicted in the first row. The second row presents the reference segmentation overlaid on a mid-sagittal slice for each volume. Subsequent rows display network segmentations overlaid on mid-sagittal slices, highlighting disjoint masks in columns two, three, and four. The specific sounds sustained during MRI scans (/f/, /l/, “UP”, /k/, /… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

57 extracted references · 46 canonical work pages

  1. [1]

    We develop an open-source manually labelled 3D vocal tract database from the French speaker database13. We provide annotations of 53 vocal tract volumes (or 1696 2D slices with 32 slices per volume) manually segmented from 10 speakers producing different French phonemes and voiceless tasks. A total of 21 unique phonemes and 3 unique voiceless tasks were i...

    work page

  2. [2]

    We evaluate the performance of four state-of-the-art deep learning segmentation algorithms, including: a) the slice-by-slice 2D U-Net method employing 2D convolutions; b) the 3D U-Net method employing 3D convolutions; c) the 3D U-Net method leveraging transfer learning by pre-training with open-source annotated samples from another lung CT imaging applica...

    work page

  3. [3]

    We quantitatively evaluate segmentations from the above algorithms against ground truth segmentations created by the simultaneous truth and performance level estimation (STAPLE) method40, which mitigates inter-user human variability while creating ground truth segmentations. In this work, we used manual segmentations by 3 human experts who created them in...

    work page

  4. [4]

    METHODS: Datasets and Pre-processing: Datasets used in this work were divided into two parts: the pre- training dataset, and the training dataset

    We provide open-source code of the above four models to facilitate ease of reproducibility by other researchers. METHODS: Datasets and Pre-processing: Datasets used in this work were divided into two parts: the pre- training dataset, and the training dataset. The pre-training dataset contains data that were used for pre-training the 3D U-NET transfer lear...

    work page

  5. [5]

    Fabre, D., Hueber, T., Girin, L., Alameda-Pineda, X., & Badin, P. (2017). Automatic animation of an articulatory tongue model from ultrasound images of the vocal tract. Speech Communication, 93, 63–75. https://doi.org/10.1016/j.specom.2017.08.002

    work page doi:10.1016/j.specom.2017.08.002 2017

  6. [6]

    F., Mehta, S., Wood, M., & Wang, J

    Katz, W. F., Mehta, S., Wood, M., & Wang, J. (2017). Using electromagnetic articulography with a tongue lateral sensor to discriminate manner of articulation. The Journal of the Acoustical Society of America, 141(1), EL57. https://doi.org/10.1121/1.4973907

    work page doi:10.1121/1.4973907 2017

  7. [7]

    Döllinger, M., Kunduk, M., Kaltenbacher, M., Vondenhoff, S., Ziethe, A., Eysholdt, U., & Bohr, C. (2012). Analysis of Vocal Fold Function From Acoustic Data Simultaneously Recorded With High- Speed Endoscopy. Journal of Voice, 26(6), 726–733. https://doi.org/10.1016/j.jvoice.2012.02.001

    work page doi:10.1016/j.jvoice.2012.02.001 2012

  8. [8]

    Badin, P. (1991). Fricative consonants: acoustic and X-ray measurements. Journal of Phonetics, 19(3– 4), 397–408. https://doi.org/10.1016/S0095-4470(19)30331-6

    work page doi:10.1016/s0095-4470(19)30331-6 1991

Show all 57 references
  1. [9]

    E., & Jermy, M

    Kabaliuk, N., Nejati, A., Loch, C., Schwass, D., Cater, J. E., & Jermy, M. C. (2017). Strategies for Segmenting the Upper Airway in Cone-Beam Computed Tomography (CBCT) Data. Open Journal of Medical Imaging, 07(04), 196–219. https://doi.org/10.4236/ojmi.2017.74019

    2017

  2. [10]

    G., Toutios, A., Toger, J., Lim, Y., Zhu, Y., Kim, Y

    Lingala, S. G., Toutios, A., Toger, J., Lim, Y., Zhu, Y., Kim, Y. C., Vaz, C., Narayanan, S., & Nayak, K. (2016). State-of-the-art MRI protocol for comprehensive assessment of vocal tract structure and function. Proceedings of the Annual Conference of the International Speech ...

    2016 doi

  3. [12]

    H., Titze, I

    Story, B. H., Titze, I. R., & Hoffman, E. A. (1998). Vocal tract area functions for an adult female speaker based on volumetric imaging. The Journal of the Acoustical Society of America, 104(1), 471–487. https://doi.org/10.1121/1.423298

    1998 doi

  4. [13]

    G., Sutton, B

    Lingala, S. G., Sutton, B. P., Miquel, M. E., & Nayak, K. S. (2016). Recommendations for real-time speech MRI. Journal of Magnetic Resonance Imaging, 43(1), 28–44. https://doi.org/10.1002/jmri.24997

    2016 doi

  5. [14]

    Bresch, E., Yoon-Chul Kim, Nayak, K., Byrd, D., & Narayanan, S. (2008). Seeing speech: Capturing vocal tract shaping using real-time magnetic resonance imaging [Exploratory DSP]. IEEE Signal Processing Magazine, 25(3), 123–132. https://doi.org/10.1109/MSP.2008.918034

    2008

  6. [15]

    H., Titze, I

    Story, B. H., Titze, I. R., & Hoffman, E. A. (1998). Vocal tract area functions from magnetic resonance imaging. The Journal of the Acoustical Society of America, 100(1), 537. https://doi.org/10.1121/1.415960

    1998 doi

  7. [16]

    G., Vaz, C., Sorensen, T., Oh, M., Harper, S., Chen, W., Lee, Y., Töger, J., Monteserin, M

    Lim, Y., Toutios, A., Bliesener, Y., Tian, Y., Lingala, S. G., Vaz, C., Sorensen, T., Oh, M., Harper, S., Chen, W., Lee, Y., Töger, J., Monteserin, M. L., Smith, C., Godinez, B., Goldstein, L., Byrd, D., Nayak, K. S., & Narayanan, S. S. (2021). A multispeaker dataset of raw an...

    2021 doi

  8. [17]

    G., Hennig, J., & Zaitsev, M

    Burdumy, M., Traser, L., Burk, F., Richter, B., Echternach, M., Korvink, J. G., Hennig, J., & Zaitsev, M. (2017). One-second MRI of a three-dimensional vocal tract to measure dynamic articulator modifications. Journal of Magnetic Resonance Imaging : JMRI, 46(1), 94–101. https:...

    2017 doi

  9. [18]

    K., Felblinger, J., & Vuissoz, P

    Isaieva, K., Laprie, Y., Leclère, J., Douros, I. K., Felblinger, J., & Vuissoz, P. A. (2021). Multimodal dataset of real-time 2D and static 3D MRI of healthy French speakers. Scientific Data 2021 8:1, 8(1), 1–9. https://doi.org/10.1038/s41597-021-01041-3

    2021 doi

  10. [19]

    E., & McGettigan, C

    Belyk, M., Waters, S., Kanber, E., Miquel, M. E., & McGettigan, C. (2022). Individual differences in vocal size exaggeration. Scientific Reports 2022 12:1, 12(1), 1–12. https://doi.org/10.1038/s41598- 022-05170-6

    2022 doi

  11. [20]

    Feng, M., & Howard, D. M. (2023). The Dynamic Effect of the Valleculae on Singing Voice – An Exploratory Study Using 3D Printed Vocal Tracts. Journal of Voice, 37(2), 178–186. https://doi.org/10.1016/J.JVOICE.2020.12.012

    2023 doi

  12. [21]

    H., Titze, I

    Story, B. H., Titze, I. R., & Hoffman, E. A. (2001). The relationship of vocal tract shape to three voice qualities. The Journal of the Acoustical Society of America, 109(4), 1651. https://doi.org/10.1121/1.1352085

    2001 doi

  13. [22]

    Story, B. H. (2013). Phrase-level speech simulation with an airway modulation model of speech production. Computer Speech & Language, 27(4), 989–1010. https://doi.org/10.1016/J.CSL.2012.10.005

    2013 doi

  14. [23]

    Z., Meyer, D., Jacob, M., Story, B., & Lingala, S

    Rusho, R. Z., Meyer, D., Jacob, M., Story, B., & Lingala, S. G. (2023). Synthesizing speech through a tube talker model informed by dynamic MRI-derived vocal tract area functions. Proc Intl Soc Mag Reson Med

    2023

  15. [24]

    Birkholz, P., Kürbis, S., Stone, S., Häsner, P., Blandin, R., & Fleischer, M. (2020). Printable 3D vocal tract shapes from MRI data and their acoustic and aerodynamic properties. Scientific Data 2020 7:1, 7(1), 1–16. https://doi.org/10.1038/s41597-020-00597-w

    2020 doi

  16. [25]

    A., Frahm, J., Lamalle, L., Vilain, C., & Boë, L

    Labrunie, M., Badin, P., Voit, D., Joseph, A. A., Frahm, J., Lamalle, L., Vilain, C., & Boë, L. J. (2018). Automatic segmentation of speech articulators from real-time midsagittal MRI based on supervised learning. Speech Communication, 99, 27–46. https://doi.org/10.1016/J.SPEC...

    2018 doi

  17. [26]

    Howard, D. M. (2018). The Vocal Tract Organ: A New Musical Instrument Using 3-D Printed Vocal Tracts. Journal of Voice, 32(6), 660–667. https://doi.org/10.1016/J.JVOICE.2017.09.014

    2018 doi

  18. [27]

    C., Khoo, M

    Javed, A., Kim, Y. C., Khoo, M. C. K., Ward, S. L. D., & Nayak, K. S. (2016). Dynamic 3-D MR visualization and detection of upper airway obstruction during sleep using region-growing segmentation. IEEE Transactions on Biomedical Engineering. https://doi.org/10.1109/TBME.2015.2462750

    2016

  19. [28]

    I., Toutios, A., Toger, J., & Narayanan, S

    Skordilis, Z. I., Toutios, A., Toger, J., & Narayanan, S. (2017). Estimation of vocal tract area function from volumetric Magnetic Resonance Imaging. ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. https://doi.org/10.1109/ICASSP....

    2017

  20. [29]

    I., Ramanarayanan, V., Goldstein, L., & Narayanan, S

    Skordilis, Z. I., Ramanarayanan, V., Goldstein, L., & Narayanan, S. S. (2015). Experimental assessment of the tongue incompressibility hypothesis during speech production. Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH

    2015

  21. [30]

    E., & King, A

    Ruthven, M., Miquel, M. E., & King, A. P. (2021). Deep-learning-based segmentation of the vocal tract and articulators in real-time magnetic resonance images of speech. Computer Methods and Programs in Biomedicine. https://doi.org/10.1016/j.cmpb.2020.105814

    2021

  22. [31]

    Silva, S., & Teixeira, A. (2015). Unsupervised segmentation of the vocal tract from real-time MRI sequences. Computer Speech & Language, 33(1), 25–46. https://doi.org/10.1016/J.CSL.2014.12.003

    2015 doi

  23. [32]

    Sampaio, R. D. A., & Jackowski, M. P. (2017). Vocal Tract Morphology Using Real-Time Magnetic Resonance Imaging. Proceedings - 30th Conference on Graphics, Patterns and Images, SIBGRAPI 2017, 359–366. https://doi.org/10.1109/SIBGRAPI.2017.54

    2017 doi

  24. [33]

    S., & Narayanan, S

    Ramanarayanan, V., Tilsen, S., Proctor, M., Töger, J., Goldstein, L., Nayak, K. S., & Narayanan, S. (2018). Analysis of speech production real-time MRI. Computer Speech & Language, 52, 1–22. https://doi.org/10.1016/J.CSL.2018.04.002

    2018 doi

  25. [34]

    Bresch, E., & Narayanan, S. (2009). Region segmentation in the frequency domain applied to upper airway real-time magnetic resonance images. IEEE Transactions on Medical Imaging. https://doi.org/10.1109/TMI.2008.928920

    2009

  26. [36]

    Erattakulangara, S., & Lingala, S. G. (2020). Airway segmentation in speech MRI using the U-net architecture. IEEE International Symposium on Biomedical Imaging (ISBI), to appear

    2020

  27. [38]

    A., Kumar, A., Mannem, R., Karthik, G

    Valliappan, C. A., Kumar, A., Mannem, R., Karthik, G. R., & Ghosh, P. K. (2019). An Improved Air Tissue Boundary Segmentation Technique for Real Time Magnetic Resonance Imaging Video Using Segnet. ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing...

    2019

  28. [39]

    A., Mannem, R., & Kumar Ghosh, P

    Valliappan, C. A., Mannem, R., & Kumar Ghosh, P. (2018). Air-tissue boundary segmentation in real- time magnetic resonance imaging video using semantic segmentation with fully convolutional networks. Proceedings of the Annual Conference of the International Speech Communicatio...

    2018 doi

  29. [41]

    K., Tong, Y., Torigian, D

    Xie, L., Udupa, J. K., Tong, Y., Torigian, D. A., Huang, G., Kogan, R. M., Nathan, J. Ben, Wootton, D. M., Choy, K. R., Sin, S., Wagshul, M., & Arens, R. (2021). Automatic upper airway segmentation in static and dynamic MRI via deep convolutional neural networks. Https://Doi.O...

    2021 doi

  30. [42]

    L., Duerk, J

    Ma, D., Gulani, V., Seiberlich, N., Liu, K., Sunshine, J. L., Duerk, J. L., & Griswold, M. a. (2013). Magnetic resonance fingerprinting. Nature, 495(7440), 187–192. https://doi.org/10.1038/nature11971

    2013 doi

  31. [43]

    L., Erus, G., Doshi, J., Singh, A., Keenan, B

    Bommineni, V. L., Erus, G., Doshi, J., Singh, A., Keenan, B. T., Schwab, R. J., Wiemken, A., & Davatzikos, C. (2023). Automatic Segmentation and Quantification of Upper Airway Anatomic Risk Factors for Obstructive Sleep Apnea on Unprocessed Magnetic Resonance Images. Academic ...

    2023 doi

  32. [44]

    M., Adams, D

    Ruthven, M., Peplinski, A. M., Adams, D. M., King, A. P., & Miquel, M. E. (2023). Real-time speech MRI datasets with corresponding articulator ground-truth segmentations. Scientific Data 2023 10:1, 10(1), 1–10. https://doi.org/10.1038/s41597-023-02766-z

    2023 doi

  33. [45]

    K., Zou, K

    Warfield, S. K., Zou, K. H., & Wells, W. M. (2004). Simultaneous truth and performance level estimation (STAPLE): An algorithm for the validation of image segmentation. IEEE Transactions on Medical Imaging, 23(7), 903–921. https://doi.org/10.1109/TMI.2004.828354

    2004

  34. [46]

    OSIC Pulmonary Fibrosis Progression | Kaggle. (n.d.). Retrieved December 6, 2024, from https://www.kaggle.com/competitions/osic-pulmonary-fibrosis-progression

    2024

  35. [47]

    CT Lung & Heart & Trachea segmentation. (n.d.). Retrieved December 6, 2024, from https://www.kaggle.com/datasets/sandorkonya/ct-lung-heart-trachea-segmentation

    2024

  36. [48]

    Perona, P., & Malik, J. (1990). Scale-space and edge detection using anisotropic diffusion. IEEE Transactions on Pattern Analysis and Machine Intelligence, 12(7), 629–639. https://doi.org/10.1109/34.56205

    1990 doi

  37. [49]

    Pieper, S., Halle, M., & Kikinis, R. (n.d.). 3D Slicer. 2004 2nd IEEE International Symposium on Biomedical Imaging: Macro to Nano (IEEE Cat No. 04EX821), 2, 632–635. https://doi.org/10.1109/ISBI.2004.1398617

    2004 arXiv

  38. [50]

    S., Brox, T., & Ronneberger, O

    Çiçek, Ö., Abdulkadir, A., Lienkamp, S. S., Brox, T., & Ronneberger, O. (2016). 3D U-Net: Learning Dense Volumetric Segmentation from Sparse Annotation (pp. 424–432). https://doi.org/10.1007/978-3-319-46723-8_49

    2016 doi

  39. [51]

    Bertels, J., Eelbode, T., Berman, M., Vandermeulen, D., Maes, F., Bisschops, R., & Blaschko, M. (2019). Optimizing the Dice Score and Jaccard Index for Medical Image Segmentation: Theory & Practice. https://doi.org/10.1007/978-3-030-32245-8_11

    2019 doi

  40. [52]

    C., Sheikh, H

    Wang, Z., Bovik, A. C., Sheikh, H. R., & Simoncelli, E. P. (2004). Image Quality Assessment: From Error Visibility to Structural Similarity. IEEE Transactions on Image Processing, 13(4), 600–612. https://doi.org/10.1109/TIP.2003.819861

    2004

  41. [53]

    Hausdorff, F. (1914). Grundzüge der Mengenlehre. Leipzig Viet

    1914

  42. [54]

    Ronneberger, O., Fischer, P., & Brox, T. (2015). U-Net: Convolutional Networks for Biomedical Image Segmentation (pp. 234–241). https://doi.org/10.1007/978-3-319-24574-4_28

    2015 doi

  43. [55]

    Cardoso, M. J., Li, W., Brown, R., Ma, N., Kerfoot, E., Wang, Y., Murrey, B., Myronenko, A., Zhao, C., Yang, D., Nath, V., He, Y., Xu, Z., Hatamizadeh, A., Myronenko, A., Zhu, W., Liu, Y., Zheng, M., Tang, Y., … Feng, A. (2022). MONAI: An open-source framework for deep learnin...

    2022

  44. [56]

    Hatamizadeh, A., Tang, Y., Nath, V., Yang, D., Myronenko, A., Landman, B., Roth, H., & Xu, D. (2021). UNETR: Transformers for 3D Medical Image Segmentation. http://arxiv.org/abs/2103.10504

    2021 arXiv

  45. [57]

    T., Berger, C., Ha, S

    Bakas, S., Reyes, M., Jakab, A., Bauer, S., Rempfler, M., Crimi, A., Shinohara, R. T., Berger, C., Ha, S. M., Rozycki, M., Prastawa, M., Alberts, E., Lipkova, J., Freymann, J., Kirby, J., Bilello, M., Fathallah- Shaykh, H., Wiest, R., Kirschke, J., … Menze, B. (2018). Identify...

    2018

  46. [58]

    K., Tong, Y., Torigian, D

    Xie, L., Udupa, J. K., Tong, Y., Torigian, D. A., Huang, Z., Kogan, R. M., Wootton, D., Choy, K. R., Sin, S., Wagshul, M. E., & Arens, R. (2022). Automatic upper airway segmentation in static and dynamic MRI via anatomy‐guided convolutional neural networks. Medical Physics, 49...

    2022 doi

  47. [59]

    Erattakulangara, S., Kelat, K., Meyer, D., Priya, S., & Lingala, S. G. (2023). Automatic Multiple Articulator Segmentation in Dynamic Speech MRI Using a Protocol Adaptive Stacked Transfer Learning U-NET Model. Bioengineering, 10(5). https://doi.org/10.3390/bioengineering10050623

    2023 doi

  48. [61]

    UP”, /k/, /a/) are labeled in the first row. Non- /f/ /l/ “UP

    Aydıngöz, Ü., Yıldız, A. E., & Ergen, F. B. (2022). Zero Echo Time Musculoskeletal MRI: Technique, Optimization, Applications, and Pitfalls. Radiographics, 42(5), 1398–1414. https://doi.org/10.1148/RG.220029/ASSET/IMAGES/LARGE/RG.220029.FIG18.JPEG FIGURES Figure 1: Sample imag...

    2022 doi

  49. [255]

    closeness

    Furthermore, a cropping of 70 percent was applied specifically for MRI datasets to focus on the upper airway. Finally, both CT and MRI datasets were sampled to a size of 256 x 256 x 32 voxels. For deep learning models, maintaining fixed input dimensions across training and tes...

    2015

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.