Machine Learning for Automatic Detection of Velopharyngeal Dysfunction: A Preliminary Report.
Journal
The Journal of craniofacial surgery
ISSN: 1536-3732
Titre abrégé: J Craniofac Surg
Pays: United States
ID NLM: 9010410
Informations de publication
Date de publication:
06 May 2024
06 May 2024
Historique:
received:
18
12
2023
accepted:
16
02
2024
medline:
6
5
2024
pubmed:
6
5
2024
entrez:
6
5
2024
Statut:
aheadofprint
Résumé
Even after palatoplasty, the incidence of velopharyngeal dysfunction (VPD) can reach 30%; however, these estimates arise from high-income countries (HICs) where speech-language pathologists (SLP) are part of standardized cleft teams. The VPD burden in low- and middle-income countries (LMICs) is unknown. This study aims to develop a machine-learning model that can detect the presence of VPD using audio samples alone. Case and control audio samples were obtained from institutional and publicly available sources. A machine-learning model was built using Python software. The initial 110 audio samples used to test and train the model were retested after format conversion and file deidentification. Each sample was tested 5 times yielding a precision of 100%. Sensitivity was 92.73% (95% CI: 82.41%-97.98%) and specificity was 98.18% (95% CI: 90.28%-99.95%). One hundred thirteen prospective samples, which had not yet interacted with the model, were then tested. Precision was again 100% with a sensitivity of 88.89% (95% CI: 78.44%-95.41%) and a specificity of 66% (95% CI: 51.23%-78.79%). VPD affects nearly 100% of patients with unrepaired overt soft palatal clefts and up to 30% of patients who have undergone palatoplasty. VPD can render patients unintelligible, thereby accruing significant psychosocial morbidity. The true burden of VPD in LMICs is unknown, and likely exceeds estimates from HICs. The ability to access a phone-based screening machine-learning model could expand access to diagnostic, and potentially therapeutic modalities for an innumerable amount of patients worldwide who suffer from VPD.
Sections du résumé
BACKGROUND
BACKGROUND
Even after palatoplasty, the incidence of velopharyngeal dysfunction (VPD) can reach 30%; however, these estimates arise from high-income countries (HICs) where speech-language pathologists (SLP) are part of standardized cleft teams. The VPD burden in low- and middle-income countries (LMICs) is unknown. This study aims to develop a machine-learning model that can detect the presence of VPD using audio samples alone.
METHODS
METHODS
Case and control audio samples were obtained from institutional and publicly available sources. A machine-learning model was built using Python software.
RESULTS
RESULTS
The initial 110 audio samples used to test and train the model were retested after format conversion and file deidentification. Each sample was tested 5 times yielding a precision of 100%. Sensitivity was 92.73% (95% CI: 82.41%-97.98%) and specificity was 98.18% (95% CI: 90.28%-99.95%). One hundred thirteen prospective samples, which had not yet interacted with the model, were then tested. Precision was again 100% with a sensitivity of 88.89% (95% CI: 78.44%-95.41%) and a specificity of 66% (95% CI: 51.23%-78.79%).
DISCUSSION
CONCLUSIONS
VPD affects nearly 100% of patients with unrepaired overt soft palatal clefts and up to 30% of patients who have undergone palatoplasty. VPD can render patients unintelligible, thereby accruing significant psychosocial morbidity. The true burden of VPD in LMICs is unknown, and likely exceeds estimates from HICs. The ability to access a phone-based screening machine-learning model could expand access to diagnostic, and potentially therapeutic modalities for an innumerable amount of patients worldwide who suffer from VPD.
Identifiants
pubmed: 38709082
doi: 10.1097/SCS.0000000000010147
pii: 00001665-990000000-01509
doi:
Types de publication
Journal Article
Langues
eng
Sous-ensembles de citation
IM
Informations de copyright
Copyright © 2024 by Mutaz B. Habal, MD.
Déclaration de conflit d'intérêts
The authors report no conflicts of interest
Références
Iglesias A, Kuehn DP, Morris HL. Simultaneous assessment of pharyngeal wall and velar displacement for selected speech sounds. J Speech Hear Res 1980;23:429–446
Chiang SN, Fotouhi AR, Grames LM, et al. Buccal myomucosal flap repair for velopharyngeal dysfunction. Plast Reconstr Surg 2023;152:842–850
Mann RJ, Neaman KC, Armstrong SD, et al. The double-opposing buccal flap procedure for palatal lengthening. Plast Reconstr Surg 2011;127:2413–2418
Ysunza A, Pamplona C, Ramirez E, et al. Velopharyngeal surgery: a prospective randomized study of pharyngeal flaps and sphincter pharyngoplasties. Plast Reconstr Surg 2002;110:1401–1407
Naran S, Ford M, Losee JE. What’s new in cleft palate and velopharyngeal dysfunction management? Plast Reconstr Surg 2017;139:1343e–1355e
Jackson O, Stransky CA, Jawad AF, et al. The Children’s Hospital of Philadelphia modification of the Furlow double-opposing Z-palatoplasty: 30-year experience and long-term speech outcomes. Plast Reconstr Surg 2013;132:613–622
D’Antonio LL, Scherer NJ. Communication disorderes associated with cleft palate. In: Losee JE, Kirschner RE, eds. Comprehensive Cleft Care. McGraw-Hill Medical; 2009:569–588
Fisher DM, Sommerlad BC. Cleft lip, cleft palate, and velopharyngeal insufficiency. Plast Reconstr Surg 2011;128:342e–360e
de Blacam C, Smith S, Orr D. Surgery for velopharyngeal dysfunction: a systematic review of interventions and outcomes. Cleft Palate Craniofac J 2018;55:405–422
Ha JH, Lee H, Kwon SM, et al. Deep learning-based diagnostic system for velopharyngeal insufficiency based on videofluoroscopy in patients with repaired cleft palates. J Craniofac Surg 2023;34:2369–2375
He L, Zhang J, Liu Q, et al. Automatic evaluation of hypernasality based on a cleft palate speech database. J Med Syst 2015;39:61
Dhillon H, Chaudhari PK, Dhingra K, et al. Current applications of artificial intelligence in cleft care: a scoping review. Front Med (Lausanne) 2021;8:676490
Golabbakhsh M, Abnavi F, Kadkhodaei Elyaderani M, et al. Automatic identification of hypernasality in normal and cleft lip and palate patients with acoustic analysis of speech. J Acoust Soc Am 2017;141:929
Wang X, Yang S, Tang M, et al. HypernasalityNet: deep recurrent neural network for automatic hypernasality detection. Int J Med Inform 2019;129:1–12
Prevention CfDCa. CDC’s developmental milestones. 2023. Accessed December 1, 2023. https://www.cdc.gov/ncbddd/actearly/milestones/index.html
Unit EOH. Let’s talk: tips for building your child’s speech and language skills. 2017. Accessed December 1, 2023. https://www.youtube.com/watch?v=K0aHjxzDb7I
ENT F. What is VPI (velopharyngeal insufficiency)? 2018. Accessed December 1, 2023. https://www.youtube.com/watch?v=WM5fVCdBPHs
Learning JB Hypernasality. 2018. Accessed December 1, 2023. https://www.youtube.com/watch?v=KWz5_fpnZYc
LEADERSproject. Cleft palate speech therapy using books for phrases and sentences. 2019. Accessed December 1, 2023. https://www.youtube.com/watch?v=1nHhqdCnwBI
Chicago SCs. New app may help kids with cleft palate speak easier. 2016. Accessed December 1, 2023. https://www.youtube.com/watch?v=5fubZitvY-Q
Therapy BPaH. Case study: Pediatric speech therapy for cleft palate. 2016. Accessed December 1, 2023. https://www.youtube.com/watch?v=noUGRjClUg4
LEADERSproject. Cleft palate speech and feeding: addressing speech and language before surgery. 2016. Accessed December 1, 2023. https://www.youtube.com/watch?v=-sEt3i0sHr4
Association TAS-L-H. Evaluation and treatment of resonance disrders and velopharyngeal insufficiency. 2018. Accessed December 1, 2023. https://fb.watch/nBEd3Y93AQ/
Audiomass. 2019. Accessed December 1, 2023. https://audiomass.co/about.html
TubeRipper. 2014. Accessed December 1, 2023. https://tuberipper.com/16/
Bhuskute A, Skirko JR, Roth C, et al. Association of velopharyngeal insufficiency with quality of life and patient-reported outcomes after speech surgery. JAMA Facial Plast Surg 2017;19:406–412
Carlson LC, Stewart BT, Hatcher KW, et al. A model of the unmet need for cleft lip and palate surgery in low- and middle-income countries. World J Surg 2016;40:2857–2867
Moreno-Torres I, Lozano A, Nava E, et al. Which utterance types are most suitable to detect hypernasality automatically? Appl Sci 2021;11:1–16
Maier A, Honig F, Bocklet T, et al. Automatic detection of articulation disorders in children with cleft lip and palate. J Acoust Soc Am 2009;126:2589–2602
Vikram CM, Tripathi A, Kalita S, et al. Estimation of hypernasality scores from cleft lip and palate speech. Interspeech 2018:1701–1705
Brian B, Raffel C, Liang D, et al. librosa: Audio and music signal analysis in python. In Proceedings of the 14th python in science conference. 2015:18–25