Automated and interpretable m-health discrimination of vocal cord pathology enabled by machine learning

Seedat, Nabeel; Aharonson, Vered; Hamzany, Yaniv

doi:10.1109/csde50874.2020.9411529

Cited by 4 publications

(8 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…All of this should increase trust when no bias is found and when explanations are robust across models and make sense to experts. Such a model could fulfill several clinical needs: (1) postoperative screening for thyroid surgery-related UVFP since after thyroid surgery, UVFP is common, occurring in up to 5 to 10% of cases 27 . Furthermore, laryngoscopy is not readily available to all postoperative populations and symptomatic changes are notoriously variable.…”

Section: Discussionmentioning

confidence: 99%

“…Furthermore, laryngoscopy is not readily available to all postoperative populations and symptomatic changes are notoriously variable. An ML-based screening could help identify patients needing further workup and treatment, and earlier diagnosis is essential to optimize long-term outcomes 28,29 . (2) Monitoring voice during speech therapy and after surgical treatment for confirmed UVFP to measure when and if the patient's voice is approximating a healthy voice.…”

Section: Discussionmentioning

confidence: 99%

“…There have been several studies detecting unilateral vocal fold paralysis (UVFP) using machine learning (20)(21)(22)(23)(24)(25)(26)(27)(28); however, most have included the disorder among a set of voice disorders to be predicted. Limitations of these prior studies could be seen to fall into one of following types: not reporting the performance when classifying the subset of participants with UVFP out of the participants with dysphonia they were trying to detect; small sample sizes given most studies contained 10 participants with UVFP or fewer with one study containing 50 participants (29); a lack of algorithmic explanations: they either do not report on the relative importance of each acoustic variable, use hard-tointerpret input data such as a spectrogram, or use a black-box model such as neural network and do not explain it; using a single type of model or just a few features which may impede high predictive performance and/or obscure a more thorough explanation given a single model or few features may capture only certain aspects of the task; not publicly sharing their trained models to test their generalizability to new data.…”

Section: Introductionmentioning

confidence: 99%

“…There have been several studies detecting unilateral vocal fold paralysis (UVFP) using machine learning (22)(23)(24)(25)(26)(27)(28)(29)(30) ; however, most have included the disorder among a set of voice disorders to be predicted. Limitations of these prior studies could be seen to fall into one of following types: not reporting the performance when classifying the subset of participants with UVFP out of the participants with dysphonia they were trying to detect; small sample sizes given most studies contained 10 participants with UVFP or fewer with one study containing 50 participants (31) ; a lack of algorithmic explanations: they either do not report on the relative importance of each acoustic variable; use input data such as a spectrogram in a black-box deep learning model which could make attempts at algorithmic explanations on images such as saliency maps more opaque than results from feature importance of handcrafted features; use a black-box model such as neural network without attempting to explain its predictions with deep learning explainability methods (32) ; use a single type of model which may pick up on certain types of patterns but miss others leading to incomplete conclusions on feature importance; use only a few features which may impede better predictive performance by not capturing certain relevant information; and/or not publicly share models or data to help test their generalizability to new data.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Identifying bias in models that detect vocal fold paralysis from audio recordings using explainable machine learning and clinician ratings

Low

Randolph

Rao

et al. 2020

Preprint

View full text Add to dashboard Cite

ObjectivesTo detect unilateral vocal fold paralysis (UVFP) from voice recordings using an explainable model of machine learning.Study DesignCase series - retrospective with a control group.MethodsPatients with confirmed UVFP through endoscopic examination (N=77) and controls with normal voices matched for age and sex (N=77) were included. Two tasks were used to elicit voice samples: reading the Rainbow Passage and sustaining phonation of the vowel /a/. The eighty-eight extended Geneva Minimalistic Acoustic Parameter Set (eGeMAPS) features were extracted as inputs for four machine learning models of differing complexity. Training and testing were performed using bootstrapped cross-validation. SHAP was used to identify important features.ResultsThe median Area Under the Receiver Operating Characteristic Curve (ROC AUC) score ranged from 0.79 to 0.87 depending on model and task. After removing redundant features for explainability, the highest median ROC AUC score was 0.84 using only 13 features for the vowel task and 0.87 using 39 features for the reading task. The most important features included intensity measures, mean MFCC1, mean F1 amplitude and frequency, and shimmer variability depending on model and task.ConclusionUsing the largest dataset studying UVFP to date, we achieve high performance from just a few seconds of voice recordings while discovering which acoustic features are important across models. Notably, we demonstrate that the models use different combinations of features to achieve similar effect sizes. Overall the categories of features related to vocal fold physiology were conserved across the models. Machine learning thus provides a mechanism to detect UVFP and contextualize the accuracy relative to both model architecture and pathophysiology.Level of EvidenceType 3

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Identifying bias in models that detect vocal fold paralysis from audio recordings using explainable machine learning and clinician ratings

Low

Randolph

Rao

et al. 2020

Preprint

View full text Add to dashboard Cite

show abstract

“…Another automated and interpretable diagnostic application has been proposed by Seedat et al (2020) for voice pathology assessment from smartphone-based microphone recordings. To this aim, they conducted a pilot study to collect and analyse voice recordings obtained from 33 healthy and diseased subjects, then they trained several ML models using a set of handcrafted features extracted through audio signal processing.…”

Section: Explanation Consistency Assessmentmentioning

confidence: 99%

Explainable AI for clinical and remote health applications: a survey on tabular and time series data

Martino

Delmastro

2022

Artif Intell Rev

View full text Add to dashboard Cite

Nowadays Artificial Intelligence (AI) has become a fundamental component of healthcare applications, both clinical and remote, but the best performing AI systems are often too complex to be self-explaining. Explainable AI (XAI) techniques are defined to unveil the reasoning behind the system’s predictions and decisions, and they become even more critical when dealing with sensitive and personal health data. It is worth noting that XAI has not gathered the same attention across different research areas and data types, especially in healthcare. In particular, many clinical and remote health applications are based on tabular and time series data, respectively, and XAI is not commonly analysed on these data types, while computer vision and Natural Language Processing (NLP) are the reference applications. To provide an overview of XAI methods that are most suitable for tabular and time series data in the healthcare domain, this paper provides a review of the literature in the last 5 years, illustrating the type of generated explanations and the efforts provided to evaluate their relevance and quality. Specifically, we identify clinical validation, consistency assessment, objective and standardised quality evaluation, and human-centered quality assessment as key features to ensure effective explanations for the end users. Finally, we highlight the main research challenges in the field as well as the limitations of existing XAI methods.

show abstract

Implementation of an Efficient IoT Enabled Automated Paralysis Healthcare System

Annepu,

Jameel AL-Tamimi,

Aljibori

et al. 2023

2023 IEEE 8th International Conference on Engineering Technologies and Applied Sciences (ICETAS)

View full text Add to dashboard Cite

Automated and interpretable m-health discrimination of vocal cord pathology enabled by machine learning

Cited by 4 publications

References 14 publications

Identifying bias in models that detect vocal fold paralysis from audio recordings using explainable machine learning and clinician ratings

Identifying bias in models that detect vocal fold paralysis from audio recordings using explainable machine learning and clinician ratings

Explainable AI for clinical and remote health applications: a survey on tabular and time series data

Implementation of an Efficient IoT Enabled Automated Paralysis Healthcare System

Contact Info

Product

Resources

About