Digital assessment of speech in Huntington disease

Background

Speech changes are an early symptom of Huntington disease (HD) and may occur prior to other motor and cognitive symptoms. Assessment of HD commonly uses clinician-rated outcome measures, which can be limited by observer variability and episodic administration. Speech symptoms are well suited for evaluation by digital measures which can enable sensitive, frequent, passive, and remote administration.

Methods

We collected audio recordings using an external microphone of 36 (18 HD, 7 prodromal HD, and 11 control) participants completing passage reading, counting forward, and counting backwards speech tasks. Motor and cognitive assessments were also administered. Features including pausing, pitch, and accuracy were automatically extracted from recordings using the BioDigit Speech software and compared between the three groups. Speech features were also analyzed by the Unified Huntington Disease Rating Scale (UHDRS) dysarthria score. Random forest machine learning models were implemented to predict clinical status and clinical scores from speech features.

Results

Significant differences in pausing, intelligibility, and accuracy features were observed between HD, prodromal HD, and control groups for the passage reading task (e.g., p < 0.001 with Cohen'd = -2 between HD and control groups for pause ratio). A few parameters were significantly different between the HD and control groups for the counting forward and backwards speech tasks. A random forest classifier predicted clinical status from speech tasks with a balanced accuracy of 73% and an AUC of 0.92. Random forest regressors predicted clinical outcomes from speech features with mean absolute error ranging from 2.43-9.64 for UHDRS total functional capacity, motor and dysarthria scores, and explained variance ranging from 14 to 65%. Montreal Cognitive Assessment scores were predicted with mean absolute error of 2.3 and explained variance of 30%.

Conclusion

Speech data have the potential to be a valuable digital measure of HD progression, and can also enable remote, frequent disease assessment in prodromal HD and HD. Clinical status and disease severity were predicted from extracted speech features using random forest machine learning models. Speech measurements could be leveraged as sensitive marker of clinical onset and disease progression in future clinical trials.

Impact of Neighborhood Deprivation and Social Vulnerability on Outcomes and Interventions in Patients with Cleft Palate

OBJECTIVE

To examine whether neighborhood disadvantage impacts length of follow-up, interventions, and outcomes for patients with cleft palate.

DESIGN

Retrospective cohort.

SETTING

Cleft Palate Craniofacial Institute Database at St. Louis Children's Hospital.

PATIENTS/PARTICIPANTS

Patients with cleft palate following in St. Louis Children's Hospital Cleft Palate Multidisciplinary Team Clinic.

INTERVENTIONS

Primary palatoplasty between 2012-2017. Patients were divided into quartiles across area deprivation index (ADI) and social vulnerability index (SVI), two validated, composite metrics of neighborhood disadvantage, to examine whether living in neighborhoods from different deprivation quartiles impacts outcomes of interest.

MAIN OUTCOME MEASURE

Follow-up through age 5, surgeries and surgical complications, speech, developmental, and behavioral outcomes.

RESULTS

205 patients were included. 39% of patients belonged to the most deprived ADI quartile, while 15% belonged to the most vulnerable SVI quartile. There were no differences between ADI or SVI quartiles in number of operations received (p ≥ 0.40). Patients in the most deprived ADI quartile were significantly more likely to have speech/language concerns (OR 2.32, 95% CI [1.20-4.89], p = 0.01). Being in a more vulnerable SVI quartile was associated with developmental delay (OR 2.29, 95% CI [1.04-5.15], p = 0.04). ADI and SVI quartile did not impact risk of loss to follow-up in the isolated and combined cleft lip and palate subgroups (p ≥ 0.21).

CONCLUSIONS

Neighborhood disadvantage impacts speech and developmental outcomes in patients with cleft palate despite comparable length of follow-up in multidisciplinary team clinic.

Standardization of Videorecorded Speech Samples for Children with Cleft Palate: Methods to Facilitate High-Quality Speech Outcomes Data Collection

BACKGROUND

Collection of high-quality videorecorded speech samples is essential for speech outcomes research.

SOLUTION

Cleft palate team SLPs were trained to collect standard videorecorded speech samples in the clinic setting across 20 sites in North America. Standard training and equipment was provided.

WHAT WE DID THAT IS NEW

Quality management procedures were developed and utilized to verify video quality and protocol adherence. Over 97% of speech samples collected by trained SLPs met defined quality standards.

Speech Outcomes of Buccal Myomucosal Flap Palatal Lengthening for Treatment of Velopharyngeal Insufficiency: Systematic Literature Review and Meta-Analysis

OBJECTIVE

Buccal myomucosal flaps (BMF) anatomically lengthen the palate in the treatment of velopharyngeal insufficiency (VPI). We systematically reviewed the existing literature on speech outcome of BMF palatal lengthening.

DESIGN

Three databases were used to identify studies of interest published in English. Studies that did not use standardized speech assessments were excluded. PRISMA checklist was followed, and the risk of bias in the included studies was assessed.

SETTING

Long-term follow up of patients.

PATIENTS

With history of cleft palate presenting with VPI.

INTERVENTION

BMF palatal lengthening.

MAIN OUTCOME MEASURE

Random-effects model meta-analyses were performed for hypernasality, intelligibility, and nasal air emission score improvements, which were derived from reported preoperative and postoperative scores, and controlled for variability of scales and timing of postoperative assessment.

RESULTS

From the initial 7115 articles, 13 were included in this review. Two of these had a significant patient overlap and a study with a smaller patient population was excluded. All 12 included articles met the National Institutes of Health Quality Assessment Tool criteria. Six retrospective studies evaluated 230 patients and six prospective studies evaluated 181 patients. The most common indications for BMF were large size of the velopharyngeal gap and prior surgery for VPI. Meta-analyses demonstrated effect sizes below zero, confirming the improvement of standardized assessment scores in patients with VPI after BMF palatal lengthening. Egger regressions revealed low risk of publication bias.

CONCLUSIONS

BMF palatal lengthening provides adequate treatment for VPI in patients with large velopharyngeal gap size and a history of prior unsuccessful surgery.

Using "Real-World Data" to Study Cleft Lip/Palate Care: An Exploration of Speech Outcomes from a Multi-Center US Learning Health Network

OBJECTIVE

To assess the ability of a cleft-specific multi-site learning health network registry to describe variations in cleft outcomes by cleft phenotypes, ages, and treatment centers. Observed variations were assessed for coherence with prior study findings.

DESIGN

Cross-sectional analysis of prospectively collected data from 2019-2022.

SETTING

Six cleft treatment centers collected data systematically during routine clinic appointments according to a standardized protocol.

PARTICIPANTS

714 English-speaking children and adolescents with non-syndromic cleft lip/palate.

INTERVENTION

Routine multidisciplinary care and systematic outcomes measurement by cleft teams.

OUTCOME MEASURES

Speech outcomes included articulatory accuracy measured by Percent Consonants Correct (PCC), velopharyngeal function measured by Velopharyngeal Competence (VPC) Rating Scale (VPC-R), intelligibility measured by caregiver-reported Intelligibility in Context Scale (ICS), and two CLEFT-Q™ surveys, in which patients rate their own speech function and level of speech distress.

RESULTS

12year-olds exhibited high median PCC scores (91-100%), high frequency of velopharyngeal competency (62.50-100%), and high median Speech Function (80-91) relative to younger peers parsed by phenotype. Patients with bilateral cleft lip, alveolus, and palate reported low PCC scores (51-91%) relative to peers at some ages and low frequency of velopharyngeal competency (26.67%) at 5 years. ICS scores ranged from 3.93-5.0 for all ages and phenotypes. Speech Function and Speech Distress were similar across phenotypes.

CONCLUSIONS

This exploration of speech outcomes demonstrates the current ability of the cleft-specific registry to support cleft research efforts as a source of "real-world" data. Further work is focused on developing robust methodology for hypothesis-driven research and causal inference.

Assessing the Agreement of Hypernasality and Audible Nasal Emission Ratings Between Audio-Recordings and a Clinic Setting

Assess agreement of hypernasality and audible nasal emission (ANE) ratings between audio-recordings and a clinic setting.

Cross-sectional study using retrospective clinical recordings.

Audio-recording ratings by two trained speech language pathologists.

Percent agreement and intra- and inter-rater reliability of perceptual ratings.

Intra-rater reliability (AC2) of 167 audio-recorded speech samples for the primary and secondary raters, respectively, was 0.82 and 0.79 for hypernasality; for ANE, it was 0.57 and 0.75. Inter-rater reliability was 0.77 for hypernasality and 0.63 for ANE. When comparing ratings made from audio-recording versus the original clinical ratings, intra-rater reliability was 0.85 and 0.61 (primary and secondary rater, respectively) for hypernasality and 0.21 and 0.34 for ANE.

Ratings for hypernasality made from audio recordings were consistent with clinical evaluation, while ratings of ANE were not. ANE ratings made from audio recordings may not be a valid measure of velopharyngeal insufficiency speech characteristics.

CNN-Based Identification of Parkinson's Disease from Continuous Speech in Noisy Environments

Parkinson's disease is a progressive neurodegenerative disorder caused by dopaminergic neuron degeneration. Parkinsonian speech impairment is one of the earliest presentations of the disease and, along with tremor, is suitable for pre-diagnosis. It is defined by hypokinetic dysarthria and accounts for respiratory, phonatory, articulatory, and prosodic manifestations. The topic of this article targets artificial-intelligence-based identification of Parkinson's disease from continuous speech recorded in a noisy environment. The novelty of this work is twofold. First, the proposed assessment workflow performed speech analysis on samples of continuous speech. Second, we analyzed and quantified Wiener filter applicability for speech denoising in the context of Parkinsonian speech identification. We argue that the Parkinsonian features of loudness, intonation, phonation, prosody, and articulation are contained in the speech, speech energy, and Mel spectrograms. Thus, the proposed workflow follows a feature-based speech assessment to determine the feature variation ranges, followed by speech classification using convolutional neural networks. We report the best classification accuracies of 96% on speech energy, 93% on speech, and 92% on Mel spectrograms. We conclude that the Wiener filter improves both feature-based analysis and convolutional-neural-network-based classification performances.

Microscopic and Blind Prediction of Speech Intelligibility: Theory and Practice

Being able to estimate speech intelligibility without the need for listening tests would confer great benefits for a wide range of speech processing applications. Many attempts have therefore been made to introduce an objective, and ideally referencefree measure for this purpose. Most works analyze speech intelligibility prediction (SIP) methods from a macroscopic point of view, averaging over longer time spans. This paper, in contrast, presents a theoretical framework for the microscopic evaluation of SIP methods. Within our framework, a Statistically estimated Accuracy based on Theory (StAT) is derived, which numerically quantifies the statistical limitations inherent in microscopic SIP. A state-of-the-art approach to microscopic SIP, namely, the use of automatic speech recognition (ASR) to directly predict listening test results, is evaluated within this framework. The practical results are in good agreement with the theory. As the final contribution, a fully blind DIscriminative Speech intelligibility Predictor (DISP) is introduced and is also evaluated within the StAT framework. It is shown that this novel, blind estimator can predict intelligibility as well as-and often even with better accuracy than-the non-blind ASR-based approach, and that its results are again in good agreement with its theoretically derived performance potential.

Palatopharyngeal obturator prosthesis - A substitute for a dynamic separator: A technique

The dynamic of velopharyngeal sphincter mechanism is a complex motor skill involving coordination of soft palate and posterior and lateral pharyngeal walls. At rest, the soft palate drapes downward so that the oropharynx and the nasopharynx open allowing for normal breathing. However, during deglutition and certain speech, sounds such as plosives require complete or nearly complete velopharyngeal closure, whereas during utterance of vowels, the port needs to be open at varying degrees. Defects in velopharyngeal mechanism lead to hypernasality and decreased intelligibility of speech. The aim of this article is to understand the technique used to rehabilitate a patient with velopharyngeal insufficiency using a palatopharyngeal obturator prosthesis connected via a metal velar connector to a maxillary complete denture, with nasal endoscopic and lateral cephalometric examinations done to evaluate the outcome.

Effect of linguistic context on the perception of consonants in Parkinsonian Read French speech

We examined the effect of linguistic factors on the perceptual identification of intervocalic consonants produced by speakers with Parkinson's Diseases (PD). To neutralize the effect of preceding and following contexts, all the intervocalic consonants were excised with half the preceding and following vowels.We recorded 10 PD and 10 healthy speakers reading a text. An average of 114 VCV sequences per speaker was obtained. In total, our corpus consisted of 2280 stimuli. For the perception test, 20 adults native speakers of French were instructed that they would be presented utterances produced by different speakers and that they were to identify the sequences and write what they heard. No information was given on the sequence type (VCV).The reported consonant was examined in relation to the intended consonant; the score of distorsion was the number of phonetic features differing from the prototypical consonant. The results were examined as a function of the following/or preceding linguistic factors: consonant nature, oral/nasal vocalic context, class of word (function or content) and prosodic position within sentences.Consonant imprecision was confirmed in the speech of PD speakers. Two groups of patients were observed: the former with a low degree of dysarthria severity and scores of consonant identification close to that of healthy speakers; the latter with a high degree of dysarthria severity and a low identification score.Linguistic factors were shown to affect consonant production and perception. In both normal and PD speech, consonants had more features identified when they belonged to content words, word-initial syllables or final-phrase syllables. This suggests that in Parkinson's disease speech disorders relate to motor control and not to a loss of the linguistic knowledge.

A Deep Learning Algorithm for Objective Assessment of Hypernasality in Children With Cleft Palate

OBJECTIVES

Evaluation of hypernasality requires extensive perceptual training by clinicians and extending this training on a large scale internationally is untenable; this compounds the health disparities that already exist among children with cleft. In this work, we present the objective hypernasality measure (OHM), a speech-based algorithm that automatically measures hypernasality in speech, and validate it relative to a group of trained clinicians.

METHODS

We trained a deep neural network (DNN) on approximately 100 hours of a publicly-available healthy speech corpus to detect the presence of nasal acoustic cues generated through the production of nasal consonants and nasalized phonemes in speech. Importantly, this model does not require any clinical data for training. The posterior probabilities of the deep learning model were aggregated at the sentence and speaker-levels to compute the OHM.

RESULTS

The results showed that the OHM was significantly correlated with perceptual hypernasality ratings from the Americleft database (r = 0.797, p < 0.001) and the New Mexico Cleft Palate Center (NMCPC) database (r = 0.713, p < 0.001). In addition, we evaluated the relationship between the OHM and articulation errors; the sensitivity of the OHM in detecting the presence of very mild hypernasality; and established the internal reliability of the metric. Further, the performance of the OHM was compared with a DNN regression algorithm directly trained on the hypernasal speech samples.

SIGNIFICANCE

The results indicate that the OHM is able to measure the severity of hypernasality on par with Americleft-trained clinicians on thisdataset.

Exploring Japanese EFL Learners' Attitudes Toward English Pronunciation and its Relationship to Perceived Accentedness

This study investigates what individual differences may play a role in second language (L2) learners' pronunciation, exploring whether English as a Foreign Language (EFL) learners' attitudes toward English is linked to their perceived accentedness. Japanese EFL secondary school students (N = 62) carried out a 69-word read-aloud task and their speech samples were evaluated by 16 raters for accentedness. A ten-item questionnaire examined the attitudes toward L2 pronunciation of Japanese EFL learners. From the questionnaire, an exploratory factor analysis revealed three dimensions: pronunciation significance, interest in English sounds, and confidence in pronunciation. However, only confidence in pronunciation was significantly correlated with accentedness scores. Results are discussed in terms of the relationship between affective factors and L2 pronunciation attainment.

Analysis of velopharyngeal function and speech outcomes of Sommerlad palatoplasty combined with sphincter pharyngoplasty in surgical repair of older patients with cleft palate：experience from a major craniofacial surgery centre in eastern China

We aimed to evaluate velopharyngeal function and speech outcomes of Sommerlad palatoplasty combined with sphincter pharyngoplasty in surgical repair of cleft palate in patients over five years old. Fifty-eight patients were reviewed between the years 2013 and 2017, 31 of whom were treated with Sommerlad palatoplasty combined with sphincter pharyngoplasty, (mean age 15 (range 9 - 22) years), and 27 were treated with Sommerlad palatoplasty alone (mean age 18 (range 10-25) years). Velopharyngeal function was evaluated by radiographic lateral cephalometry and nasoendoscopy. Hypernasality, nasal emissions, and intelligibility were used to assess speech. The rate of velopharyngeal competence was 20/31 in the palatoplasty plus pharyngoplasty group and 7/27 in the palatoplasty alone group after surgical treatment (p=0.003). The improvements in hypernasality (p=0.024), air emission (p=0.004), and speech intelligibility (p=0.004) in the palatoplasty plus pharyngoplasty group was better than that in the palatoplasty alone group. It has been suggested that the surgical approach with the palatoplasty together with the sphincter pharyngoplasty has a higher rate of success in surgical repair of older patients with cleft palate.

Assessing speech at three years of age in the cleft palate population: a scoping review of assessment practices

BACKGROUND

There is no consensus in the UK regarding the types of speech samples or parameters of speech that should be assessed at 3 years of age in children with cleft palate ± cleft lip (CP±L), despite cleft units routinely assessing speech at this age. The standardization of assessment practices would facilitate comparisons of outcomes across UK cleft units; earlier identification of speech impairments-which could support more timely treatments; and more reliable recording of therapy impacts and surgical interventions.

AIMS

To explore assessment practices used to assess speech in 3-year-old children with CP±L, including speech parameters, methods of assessment and the nature of the speech sample used.

METHODS & PROCEDURES

A broad examination of the literature was undertaken through the use of a scoping review conducted in accordance with Joanna Briggs Institute guidelines. Search terms were generated from a preliminary search and then used in the main search (Medline, CINAHL, Embase, AMED and PsycINFO).

MAIN CONTRIBUTION

A combination of approaches (medical, linguistic, developmental and functional) is required to assess CP±L speech at age 3. A developmental approach is recommended at this age, considering the complexity of speech profiles at age 3, in which typically developing speech processes may occur alongside cleft speech characteristics. A combined measure for both nasal emission and turbulence, and an overall measure for velopharyngeal function for speech, show potential for assessment at this age. Categorical ordinal scales are frequently used; the use of continuous scales has yet to be fully explored at age 3. Although single-word assessments, including a subset of words developed for cross-linguistic comparisons, are frequently used, more than one type of speech sample may be needed to assess speech at this age validly. The lack of consensus regarding speech samples highlights a need for further research into the types of speech samples 3-year-olds can complete; the impact of incomplete speech samples on outcome measures (particularly relevant at this age when children may be less able to complete a full sample); the impact of different speech samples on the validity of assessments; and the reliability of listener judgements.

CONCLUSIONS & IMPLICATIONS

Whilst a medical model and linguistic approaches are often central in assessments of age-3 cleft speech, this review highlights the importance of developmental and functional approaches to assessment. Cross-linguistic single-word assessments show potential, and would facilitate the comparison of UK speech outcomes with other countries. Further research should explore the impact of different speech samples and rating scales on assessment validity and listener reliability.

Palatal augmentation prostheses in individuals treated for head and neck cancer: Effects on speech and oral transport

BACKGROUND

Negative side effects affecting articulation and oral transport are common after treatment for head and neck cancer (HNC). The aim of this study was to assess the effects on speech and patient-reported experience of palatal augmentation prostheses (PAP) in this patient group.

METHODS

Twenty consecutive patients who had received a PAP were included. We performed a blinded randomized listener assessment of recordings of the patients' speech with and without the PAP. Patient-reported experiences were collected from medical records.

RESULTS

The listener assessment showed that production of velar sounds improved significantly for 12 of 19 patients; 12 of 19 patients also reported easier or better speech with the PAP. All six patients treated for tongue cancer reported speech improvement. Seven of 20 patients reported better transport of food or saliva.

CONCLUSION

A PAP can help patients treated for HNC to more precisely articulated velar consonants and can benefit oral transport.

Introducing WCM-SE: The word complexity measure phonetically justified and adapted to Swedish

This paper presents the Word Complexity Measure for Swedish (WCM-SE), an adaptation of the original WCM developed for English by Stoel-Gammon. These measures are used to calculate the phonological complexity of words or vocalizations, based on a number of phonological complexity parameters. Each production receives a complexity score based on how many of the parameters are present in the production. Using phonological complexity scores to measure expressive phonology is suitable for assessing very young children, children with early phonology and children with phonological deficits. It is useful for both relational and independent analyses and enables comparisons between children and across development. The original WCM uses eight phonological complexity parameters in three domains: word patterns, syllable structures and sound classes. The parameters selected are phonological characteristics that are acquired late in development among English-speaking children. In the WCM-SE, complexity parameters in the domain sound classes were modified or added according to Swedish or universal patterns of phonology development. The parameters' complexity is accounted for in terms of language-general phonetic characteristics.

Assessment of Speech in Primary Cleft Palate by Two-layer Closure (Conservative Management)

Treatment of the cleft palate has evolved over a long period of time. Various techniques of cleft palate repair that are practiced today are the results of principles learned through many years of modifications. The challenge in the art of modern palatoplasty is no longer successful closure of the cleft palate but an optimal speech outcome without compromising maxillofacial growth. Throughout these periods of evolution in the treatment of cleft palate, the effectiveness of various treatment protocols has been challenged by controversies concerning speech and maxillofacial growth. In this article we have evaluated the results of Pinto's modification of Wardill-Kilner palatoplasty without radical dissection of the levator veli palitini muscle on speech and post-op fistula in two different age groups in 20 patients. Preoperative and 6-month postoperative speech assessment values indicated that two-layer palatoplasty (modified Wardill-Kilner V-Y pushback technique) without an intravelar veloplasty technique was good for speech.

Assessing vocal development in infants and toddlers

The purpose of this study was to examine changes in prelinguistic vocal production during the first 20 months of life. Vocalizations were classified into 23 mutually exclusive and exhaustive types, and grouped into five ascending levels using the Stark Assessment of Early Vocal Development-Revised (SAEVD-R). Data from 30 typically developing infants, aged 0-20 months, show that older infants attained higher developmental levels on the SAEVD-R than younger infants. Infants 0-2, 3-5, and 6-8 months of age primarily produced vocalizations from Levels 1 (Reflexive), 2 (Control of Phonation), and 3 (Expansion). Infants 9-20 months of age also produced vocalizations from Level 4 (Basic Canonical Syllables). Only infants from 16-20 months of age produced Level 5 (Advanced Forms) vocalizations in significant quantities. The outcomes indicate that the SAEVD-R is a valuable instrument for evaluating prelinguistic vocal development.