Robust Vocal Quality Feature Embeddings for Dysphonic Voice Detection

Zhang, Jianwei; Liss, Julie; Jayasuriya, Suren; Berisha, Visar

doi:10.1109/taslp.2023.3261753

Cited by 4 publications

(3 citation statements)

References 52 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It has also been shown that using only the holdout method in small datasets (less than 100 test samples) decreases the generalizability of the data 59 . In the case of our review, 6 total studies (18%) used only record‐wise validation, 24,31,32,34,40,49 while 5 studies (15%) used only the holdout method on test sets of less than 100 samples (Tables 1 and 2). 25,28,33,44,55 While most studies had fairly robust internal validation, lack of external validation was an issue across the investigations included, with only 5 acoustic (26%) and 2 video (13%) studies using an alternate dataset to test their model.…”

Section: Discussionmentioning

confidence: 77%

“…Of these, 142 did not meet our eligibility criteria, leading to 34 total studies being included in our review. This included 18 studies utilizing patient voice, 10,24‐40 15 studies using images from laryngoscopy, 12,41‐54 and 1 study using both as input for their deep learning models 55 . The study selection process is illustrated in the PRISMA flowchart (Figure 1).…”

Section: Resultsmentioning

confidence: 99%

“…To internally validate their models, 14/19 (74%) studies used at least the holdout method, indicated by a discrepancy in training and testing sample sizes (except 2 where the same recordings were spliced 10‐fold only for the training set) 32,40 . Of those, 5 used an additional method of cross‐validation, with k‐fold being the next most popular overall (8/19 studies).…”

Section: Resultsmentioning

confidence: 99%

See 2 more Smart Citations

The Use of Deep Learning Software in the Detection of Voice Disorders: A Systematic Review

Barlow,

Sragi,

Rivera‐Rivera

et al. 2024

Otolaryngol.--head neck surg.

View full text Add to dashboard Cite

ObjectiveTo summarize the use of deep learning in the detection of voice disorders using acoustic and laryngoscopic input, compare specific neural networks in terms of accuracy, and assess their effectiveness compared to expert clinical visual examination.Data SourcesEmbase, MEDLINE, and Cochrane Central.Review MethodsDatabases were screened through November 11, 2023 for relevant studies. The inclusion criteria required studies to utilize a specified deep learning method, use laryngoscopy or acoustic input, and measure accuracy of binary classification between healthy patients and those with voice disorders.ResultsThirty‐four studies met the inclusion criteria, with 18 focusing on voice analysis, 15 on imaging analysis, and 1 both. Across the 18 acoustic studies, 21 programs were used for identification of organic and functional voice disorders. These technologies included 10 convolutional neural networks (CNNs), 6 multilayer perceptrons (MLPs), and 5 other neural networks. The binary classification systems yielded a mean accuracy of 89.0% overall, including 93.7% for MLP programs and 84.5% for CNNs. Among the 15 imaging analysis studies, a total of 23 programs were utilized, resulting in a mean accuracy of 91.3%. Specifically, the twenty CNNs achieved a mean accuracy of 92.6% compared to 83.0% for the 3 MLPs.ConclusionDeep learning models were shown to be highly accurate in the detection of voice pathology, with CNNs most effective for assessing laryngoscopy images and MLPs most effective for assessing acoustic input. While deep learning methods outperformed expert clinical exam in limited comparisons, further studies integrating external validation are necessary.

show abstract

Section: Discussionmentioning

confidence: 77%