Re-Training of Convolutional Neural Networks for Glottis Segmentation in Endoscopic High-Speed Videos.

Endoscopic high-speed video (HSV) systems for visualization and assessment of vocal fold dynamics in the larynx are diverse and technically advancing. To consider resulting "concepts shifts" for neural network (NN)-based image processing, re-training of already trained and used NNs is necessary to allow for sufficiently accurate image processing for new recording modalities. We propose and discuss several re-training approaches for convolutional neural networks (CNN) being used for HSV image segmentation. Our baseline CNN was trained on the BAGLS data set (58,750 images). The new BAGLS-RT data set consists of additional 21,050 images from previously unused HSV systems, light sources, and different spatial resolutions. Results showed that increasing data diversity by means of preprocessing already improves the segmentation accuracy (mIoU + 6.35%). Subsequent re-training further increases segmentation performance (mIoU + 2.81%). For re-training, finetuning with dynamic knowledge distillation showed the most promising results. Data variety for training and additional re-training is a helpful tool to boost HSV image segmentation quality. However, when performing re-training, the phenomenon of catastrophic forgetting should be kept in mind, i.e., adaption to new data while forgetting already learned knowledge.

Validating a psychoacoustic model of voice quality.

No agreed-upon method currently exists for objective measurement of perceived voice quality. This paper describes validation of a psychoacoustic model designed to fill this gap. This model includes parameters to characterize the harmonic and inharmonic voice sources, vocal tract transfer function, fundamental frequency, and amplitude of the voice, which together serve to completely quantify the integral sound of a target voice sample. In experiment 1, 200 voices with and without diagnosed vocal pathology were fit with the model using analysis-by-synthesis. The resulting synthetic voice samples were not distinguishable from the original voice tokens, suggesting that the model has all the parameters it needs to fully quantify voice quality. In experiment 2 parameters that model the harmonic voice source were removed one by one, and the voice tokens were re-synthesized with the reduced model. In every case the lower-dimensional models provided worse perceptual matches to the quality of the natural tokens than did the original set, indicating that the psychoacoustic model cannot be reduced in dimensionality without loss of fit to the data. Results confirm that this model can be validly applied to quantify voice quality in clinical and research applications.

Middle age, a key time point for changes in birdsong and human voice.

Voice changes caused by natural aging and neurodegenerative diseases are prevalent in the aging population and diminish quality of life. Most treatments involve behavioral interventions that target the larynx because of a limited understanding of central brain mechanisms. The songbird offers a unique entry point into studying age-related changes in vocalizations because of a well-characterized neural circuitry for song that shares homology to human vocal control areas. Previously we established a translational dictionary for evaluating acoustic features of birdsong in the context of human voice measurements. In the present study, we conduct extensive analyses of birdsongs from young, middle-aged, and old male zebra finches. Our findings show that birdsongs become louder with age, and changes in periodic energy occur at middle age but are transient; songs appear to stabilize in old birds. Furthermore, faster songs are detected in finches at middle age compared with young and old finches. Vocal disorders in humans emerge at middle age, but the underlying brain pathologies are not well identified. The current findings will motivate future investigations using the songbird model to identify possible brain mechanisms involved in human vocal disorders of aging. (PsycInfo Database Record (c) 2020 APA, all rights reserved).

Frailty Syndrome, Cognition, and Dysphonia in the Elderly.

PURPOSE: The purpose of the current study is to determine the relation of frailty syndrome to acoustic measures of voice quality and voice-related handicap. METHODS: Seventy-three adults (52 community-dwelling participants and 21 assisted living residents) age 60 and older completed frailty screening, acoustic assessment, cognitive screening, and the Voice Handicap Index-10 (VHI-10). Factor analysis was used to consolidate acoustic measures. Statistical analysis included multiple regression, analysis of variance, and Tukey post-hoc tests with alfa of 0.05. RESULTS: Montreal Cognitive Assessment (MoCA) and exhaustion explained 28% of the variance in VHI-10. MoCA and sex explained 27% of the variance in factor 1 (spectral ratio), age and MoCA explained 13% of the variance in factor 2 (cepstral peak prominence for speech), and slowness explained 10% of the variance in factor 3 (cepstral peak prominence for sustained /a/). There were statistically significant differences in two measures across frailty groups: VHI-10 and MoCA. Acoustic factor scores did not differ significantly among frailty groups (P > 0.05). CONCLUSIONS: Voice-related handicap and cognitive status differed among robust and frail older adults, yet vocal function measures did not. The components of frailty most related to VHI-10 were exhaustion and weight loss rather than slowness, weakness, or inactivity. Based on these findings, routine screening of physical frailty and cognition are recommended as part of a complete voice evaluation for older adults.

Common Terminology and Acoustic Measures for Human Voice and Birdsong.

Purpose The zebra finch is used as a model to study the neural circuitry of auditory-guided human vocal production. The terminology of birdsong production and acoustic analysis, however, differs from human voice production, making it difficult for voice researchers of either species to navigate the literature from the other. The purpose of this research note is to identify common terminology and measures to better compare information across species. Method Terminology used in the birdsong literature will be mapped onto terminology used in the human voice production literature. Measures typically used to quantify the percepts of pitch, loudness, and quality will be described. Measures common to the literature in both species will be made from the songs of 3 middle-age birds using Praat and Song Analysis Pro. Two measures, cepstral peak prominence (CPP) and Wiener entropy (WE), will be compared to determine if they provide similar information. Results Similarities and differences in terminology and acoustic analyses are presented. A core set of measures including frequency, frequency variability within a syllable, intensity, CPP, and WE are proposed for future studies. CPP and WE are related yet provide unique information about the syllable structure. Conclusions Using a core set of measures familiar to both human voice and birdsong researchers, along with both CPP and WE, will allow characterization of similarities and differences among birds. Standard terminology and measures will improve accessibility of the birdsong literature to human voice researchers and vice versa. Supplemental Material https://doi.org/10.23641/asha.7438964.

Vocal Fold Vibration in Older Adults With and Without Age-Related Dysphonia.

Purpose: The purpose of this study was to identify the extent to which 7 measures of glottal area timing and regularity differ between older adults with and without age-related dysphonia (ARD). Method: Laryngeal high-speed videoendoscopy was completed at 4,000 frames per second for 42 adults aged 70 years and older (ARD: 9 female, 5 male; control group: 15 female, 13 male). Relative glottal gap, open quotient, speed index, maximum area declination rate, harmonics-to-noise ratio, harmonic richness factor, and standard deviation of fundamental frequency were measured from a 0.5-s segment of the glottal area waveform. Eta squared (η2) was computed to estimate group effect. Results: Small effect sizes (η2 = .18-.35) were present for relative glottal gap, open quotient, maximum area declination rate, harmonic richness factor, and standard deviation of fundamental frequency. Speed index and glottal harmonics-to-noise ratio did not explain group membership (η2 = .001 and .05, respectively). Conclusion: These findings provide evidence that vocal fold vibration in ARD is different than in normal aging, whereas the overlap in values for every measure is consistent with the concept that normal aging and ARD exist as a continuum of health and disease.

Influence of Left-Right Asymmetries on Voice Quality in Simulated Paramedian Vocal Fold Paralysis.

Purpose: The purpose of this study was to determine the vocal fold structural and vibratory symmetries that are important to vocal function and voice quality in a simulated paramedian vocal fold paralysis. Method: A computational kinematic speech production model was used to simulate an exemplar "voice" on the basis of asymmetric settings of parameters controlling glottal configuration. These parameters were then altered individually to determine their effect on maximum flow declination rate, spectral slope, cepstral peak prominence, harmonics-to-noise ratio, and perceived voice quality. Results: Asymmetry of each of the 5 vocal fold parameters influenced vocal function and voice quality; measured change was greatest for adduction and bulging. Increasing the symmetry of all parameters improved voice, and the best voice occurred with overcorrection of adduction, followed by bulging, nodal point ratio, starting phase, and amplitude of vibration. Conclusions: Although vocal process adduction and edge bulging asymmetries are most influential in voice quality for simulated vocal fold motion impairment, amplitude of vibration and starting phase asymmetries are also perceptually important. These findings are consistent with the current surgical approach to vocal fold motion impairment, where goals include medializing the vocal process and straightening concave edges. The results also explain many of the residual postoperative voice limitations.

Modeling the voice source in terms of spectral slopes.

A psychoacoustic model of the voice source spectrum is proposed. The model is characterized by four spectral slope parameters: the difference in amplitude between the first two harmonics (H1-H2), the second and fourth harmonics (H2-H4), the fourth harmonic and the harmonic nearest 2 kHz in frequency (H4-2 kHz), and the harmonic nearest 2 kHz and that nearest 5 kHz (2 kHz-5 kHz). As a step toward model validation, experiments were conducted to establish the acoustic and perceptual independence of these parameters. In experiment 1, the model was fit to a large number of voice sources. Results showed that parameters are predictable from one another, but that these relationships are due to overall spectral roll-off. Two additional experiments addressed the perceptual independence of the source parameters. Listener sensitivity to H1-H2, H2-H4, and H4-2 kHz did not change as a function of the slope of an adjacent component, suggesting that sensitivity to these components is robust. Listener sensitivity to changes in spectral slope from 2 kHz to 5 kHz depended on complex interactions between spectral slope, spectral noise levels, and H4-2 kHz. It is concluded that the four parameters represent non-redundant acoustic and perceptual aspects of voice quality.

Perceptual consequences of changes in epilaryngeal area and shape.

The influence of epilaryngeal area on glottal flow and the acoustic signal has been described [Titze, J. Acoust. Soc. Am. 123, 2733-2749 (2008)], but it is not known how (or whether) changes in epilaryngeal area influence perceived voice quality. This study examined these relationships in a kinematic vocal tract model. Epilaryngeal constrictions and expansions were simulated at the levels of the aryepiglottic folds and the ventricular folds in the context of four glottal configurations representing normal vibration to severe vocal fold paralysis, for the three corner vowels /a/, /i/, and /u/. Minimum and maximum glottal flow, maximum flow declination rate, spectral slope, cepstral peak prominence, and the harmonics-to-noise ratio were measured, and listeners completed a perceptual sort-and-rate task for all samples. Epilaryngeal constriction and expansion caused salient differences in voice quality. The location of constriction was also perceivable. Vowels simulated with aryepiglottic constriction demonstrated lower maximum airflow and less noise than the other epilaryngeal shapes, and listeners consistently perceived them as distinct from other stimuli. Acoustic differences decreased with increasing severity of simulated paralysis. Results of epilaryngeal constriction and expansion were similar for /a/ and /i/, and produced slightly different patterns for /u/.

Acoustic and perceptual effects of left-right laryngeal asymmetries based on computational modeling.

PURPOSE: Computational modeling was used to examine the consequences of 5 different laryngeal asymmetries on acoustic and perceptual measures of vocal function. METHOD: A kinematic vocal fold model was used to impose 5 laryngeal asymmetries: adduction, edge bulging, nodal point ratio, amplitude of vibration, and starting phase. Thirty /a/ and /ɪ/ vowels were generated for each asymmetry and analyzed acoustically using cepstral peak prominence (CPP), harmonics-to-noise ratio (HNR), and 3 measures of spectral slope (H1*-H2*, B0-B1, and B0-B2). Twenty listeners rated voice quality for a subset of the productions. RESULTS: Increasingly asymmetric adduction, bulging, and nodal point ratio explained significant variance in perceptual rating (R2 = .05, p < .001). The same factors resulted in generally decreasing CPP, HNR, and B0-B2 and in increasing B0-B1. Of the acoustic measures, only CPP explained significant variance in perceived quality (R2 = .14, p < .001). Increasingly asymmetric amplitude of vibration or starting phase minimally altered vocal function or voice quality. CONCLUSION: Asymmetries of adduction, bulging, and nodal point ratio drove acoustic measures and perception in the current study, whereas asymmetric amplitude of vibration and starting phase demonstrated minimal influence on the acoustic signal or voice quality.

Toward a unified theory of voice production and perception.

At present, two important questions about voice remain unanswered: When voice quality changes, what physiological alteration caused this change, and if a change to the voice production system occurs, what change in perceived quality can be expected? We argue that these questions can only be answered by an integrated model of voice linking production and perception, and we describe steps towards the development of such a model. Preliminary evidence in support of this approach is also presented. We conclude that development of such a model should be a priority for scientists interested in voice, to explain what physical condition(s) might underlie a given voice quality, or what voice quality might result from a specific physical configuration.

Relation of perceived breathiness to laryngeal kinematics and acoustic measures based on computational modeling.

PURPOSE: In this study, the authors sought to determine (a) how specific vocal fold structural and vibratory features relate to breathy voice quality and (b) the relation of perceived breathiness to 4 acoustic correlates of breathiness. METHOD: A computational, kinematic model of the vocal fold medial surfaces was used to specify features of vocal fold structure and vibration in a manner consistent with breathy voice. Four model parameters were altered: vocal process separation, surface bulging, vibratory nodal point, and epilaryngeal constriction. Twelve naïve listeners rated breathiness of 364 samples relative to a reference. The degree of breathiness was then compared to (a) the underlying kinematic profile and (b) 4 acoustic measures: cepstral peak prominence (CPP), harmonics-to-noise ratio, and two measures of spectral slope. RESULTS: Vocal process separation alone accounted for 61.4% of the variance in perceptual rating. Adding nodal point ratio and bulging to the equation increased the explained variance to 88.7%. The acoustic measure CPP accounted for 86.7% of the variance in perceived breathiness, and explained variance increased to 92.6% with the addition of one spectral slope measure. CONCLUSION: Breathiness ratings were best explained kinematically by the degree of vocal process separation and acoustically by CPP.

Developmental changes in the connective tissues of the porcine recurrent laryngeal nerve.

The recurrent laryngeal nerve (RLN) branches from the vagus cranial nerve to innervate structures important for voicing and swallowing. Damage to this nerve, commonly associated with surgery or idiopathic etiologies that largely occur with aging, results in impaired voicing and swallowing (Myssiorek, 2004). Sunderland proposed a model of peripheral nerve damage whereby a nerve's ability to resist damage from stretch and compression is determined by the quantity and composition of its epineurial connective tissues (Sunderland, 1951). Thus, it would be expected that epineurium differs depending upon the forces imposed on a nerve within its anatomical setting. The purpose of this study was to investigate RLN epineurium quantity and composition with development. A porcine model (piglet vs. juvenile) was used because of the similarity between porcine and human laryngeal innervation, anatomy and function. The entire RLN was excised bilaterally, and stereological methods were used to quantify the composition of epineurial connective tissues. Compared with the piglet, the juvenile pig RLN was double the diameter. While the piglet had no differences in the percentage of epineurial collagen and adipose between proximal and distal segments of both sides of the RLN, the juvenile pig had a greater percentage of collagen in the proximal segment of both sides of the RLN and a greater percentage of adipose in the distal segment of the left RLN compared with the proximal segment. In addition, unlike the piglet, the juvenile pig had a greater number of fascicles in the proximal than distal segment of the RLN, regardless of nerve side. These findings are consistent with predicted patterns associated with the different anatomical settings of the left and right RLN, show that the RLN changes with age, and support Sunderland's model.

Relation of structural and vibratory kinematics of the vocal folds to two acoustic measures of breathy voice based on computational modeling.

PURPOSE: To relate vocal fold structure and kinematics to 2 acoustic measures: cepstral peak prominence (CPP) and the amplitude of the first harmonic relative to the second (H1-H2). METHOD: The authors used a computational, kinematic model of the medial surfaces of the vocal folds to specify features of vocal fold structure and vibration in a manner consistent with breathy voice. Four model parameters were altered: degree of vocal fold adduction, surface bulging, vibratory nodal point, and supraglottal constriction. CPP and H1-H2 were measured from simulated glottal area, glottal flow, and acoustic waveforms and were related to the underlying vocal fold kinematics. RESULTS: CPP decreased with increased separation of the vocal processes, whereas the nodal point location had little effect. H1-H2 increased as a function of separation of the vocal processes in the range of 1.0 mm to 1.5 mm and decreased with separation > 1.5 mm. CONCLUSIONS: CPP is generally a function of vocal process separation. H1*-H2* (see paragraph 6 of article text for an explanation of the asterisks) will increase or decrease with vocal process separation on the basis of vocal fold shape, pivot point for the rotational mode, and supraglottal vocal tract shape, limiting its utility as an indicator of breathy voice. Future work will relate the perception of breathiness to vocal fold kinematics and acoustic measures.

Inhaled triamcinolone with proton pump inhibitor for treatment of vocal process granulomas: a series of 67 granulomas.

OBJECTIVES: We sought to analyze the outcomes of vocal process granulomas treated with proton pump inhibitors and inhaled triamcinolone acetonide. METHODS: We reviewed the medical records of patients with a diagnosis of contact granuloma or vocal process granuloma between 1995 and 2008. Data included age, gender, intubation history, reflux history, lesion location, previous treatment methods, treatment course, and recurrence. All patients were treated with daily or twice-daily protein pump inhibitors and inhaled triamcinolone acetonide (300 microg 3 times a day). RESULTS: Sixty-seven granulomas were diagnosed in 54 patients: 13 bilateral and 41 unilateral. Twenty patients, including all 11 women, had a recent history of intubation. Sixty-two granulomas in 50 patients were treated with triamcinolone and a proton pump inhibitor. Of the 57 granulomas that completed treatment, 5 (9%) did not respond (mean follow-up, 50 weeks; range, 30.3 to 78.3 weeks), 13 (22%) partially responded (mean follow-up, 11 weeks; range, 3 to 30 weeks), and 40 (69%) completely responded (mean follow-up, 21 weeks; range, 5.9 to 84.6 weeks). Three cases had recurrence: 2 nonresponders and 1 complete responder. One patient developed oral thrush. CONCLUSIONS: In this study, vocal process granulomas occurred more frequently in men, whereas women developed granulomas only after intubation. The anti-inflammatory action of inhaled triamcinolone combined with antireflux proton pump inhibitors successfully treats most vocal process granulomas with low rates of side effects and recurrence.

Supracricoid partial laryngectomy: swallowing, voice, and speech outcomes.

OBJECTIVES: The purpose of the study was to describe the swallowing and vocal function of patients after supracricoid partial laryngectomy (SCPL) as they changed over the first postoperative year. METHODS: Ten patients with laryngeal carcinoma underwent SCPL at Johns Hopkins Hospital between August 2003 and May 2005. Clinical and videofluoroscopic swallowing examinations and perceptual, acoustic, aerodynamic, and video-stroboscopic voice evaluations were completed before operation and at 3 weeks (swallowing only) and 2 (voice only), 6, and 12 months after operation. RESULTS: The mean time to gastrostomy tube removal was 82 days. The patients tolerated an increased variety of foods over the first postoperative year. All patients initially used therapeutic strategies to swallow safely, and some still required them at 1 postoperative year. Over the year, the perceptual ratings of voice quality improved significantly. There were no consistent changes in acoustic or aerodynamic measures. The number of patients who used multiple vibratory sources to phonate increased over the year. CONCLUSIONS: The patients tolerated regular diets, yet continued to exhibit silent aspiration and a variety of decompensations. Their voices were breathy, rough, and strained. Their voice quality ratings improved over the year. Group changes were not captured, and it appears that the changes in speech and voice 2 months after surgery were subtle.

High efficiency gene delivery into laryngeal muscle with bidirectional electroporation.

OBJECTIVE: The impact of polarity change on the efficiency of in vivo electroporative (EP) gene transfection was assessed in rat laryngeal muscle. STUDY DESIGN AND SETTING: High (HV) and low field voltage (LV) were combined with polarity changes to determine transfection in 5 different conditions: 1) without EP (EP[-]), 2) HV+LV (HL), 3) HV+LV followed by HV+LV with no change in polarity (HLHL unidirectional), 4) HV+LV followed by HV+LV with opposite polarity (HLHL bidirectional), 5) HV+LV followed by LV with opposite polarity (HLL bidirectional). RESULTS: HLL bidirectional sequence showed the best result with less interindividual variability and extended expression period. With the exception of repeated high voltage sequences, EP parameters were not likely to induce cell injury or inflammation. CONCLUSION: HLL bidirectional electroporative gene delivery produces high transfection rates with limited tissue trauma. SIGNIFICANCE: Bidirectional EP provides a safe and highly efficient method for therapeutic gene delivery into skeletal muscle.

The utility of evaluating true vocal fold motion before thyroid surgery.

OBJECTIVE: The objective of this study was to evaluate the utility of screening laryngoscopic examination in evaluating vocal fold (VF) mobility before thyroid surgery. METHODS: The authors conducted a retrospective chart review of 340 patients who have undergone thyroid surgery from January 1998 to June 2005 and had preoperative laryngoscopy by mirror, fiberoptic, or videostroboscopic examination. Reports of preoperative voice change or complaint and reports of preoperative VF examination, including the method of examination, were recorded. For patients with VF motion impairment, reports of the intraoperative condition of the recurrent laryngeal nerve (RLN), preoperative diagnosis based on fine needle aspiration, and final postoperative histopathologic examination results were recorded. RESULTS: Twenty-two patients were found to have preoperative VF motion impairment, of which seven (32%) patients were asymptomatic with no detectable subjective or objective voice problems. This differs significantly from the hypothesis that patients with VF motion impairment are always symptomatic (P=.009). Using voice symptoms as a screening test to predict VF motion impairment in 340 patients reveals that the sensitivity was 68%, specificity was 91%, positive predictive value (PPV) was 31%, and negative predictive value (NPV) was 98%. Among the 22 patients with preoperative VF motion impairment, five (72%) of the seven asymptomatic patients had benign, slowly progressive disease on their final histopathology reports. Six of these asymptomatic patients had their preoperative VF evaluation by fiberoptic examination, whereas one patient had indirect mirror laryngoscopy. Of 22 patients with preoperative VF motion impairment, five (22.5%) patients had abnormal VF mobility contralateral to the thyroid lesion on their preoperative evaluation, and only two of them had nerve injury reported after a previous thyroid surgery. This result differs significantly from the hypothesis that impaired mobility is ipsilateral to the side of the lesion (P=.05). CONCLUSIONS: Patients without voice complaints can have VF motion impairment. Patients can also have VF motion impairment contralateral to the thyroid lesion. Preoperative VF examination helps counsel patients appropriately about the risks of surgery and helps outline a plan for the extent of surgery while minimizing the medicolegal ramifications of iatrogenic RLN injury.

Proteomic analysis of rat laryngeal muscle following denervation.

Laryngeal muscle atrophy induced by nerve injury is a major factor contributing to the disabling symptoms associated with laryngeal paralysis. Alterations of global proteins in rat laryngeal muscle following denervation were, therefore, studied using proteomic techniques. Twenty-eight adult Sprague-Dawley rats were divided into normal control and denervated groups. The thyroarytenoid (TA) muscle was excised 60 days after right recurrent laryngeal nerve was resected. Protein separation and identification were preformed using 2-DE and MALDI-MS with database search. Forty-four proteins were found to have significant alteration in expression level after denervation. The majority of these proteins (57%), most of them associated with energy metabolism, cellular proliferation and differentiation, signal transduction and stress reaction, were decreased levels of expression in denervated TA muscle. The remaining 43% of the proteins, most of them involved with protein degradation, immunoreactivity, injury repair, contraction, and microtubular formation, were found to have increased levels of expression. The protein modification sites by phosphorylation were detected in 22% of the identified proteins that presented multiple-spot patterns on 2-D gel. Significant changes in protein expression in denervated laryngeal muscle may provide potential therapeutic strategies for the treatment of laryngeal paralysis.

Swallowing and speech therapy after definitive treatment for laryngeal cancer.

As the trend in laryngeal cancer treatment shifts towards organ-conservation surgeries and organ-preservation protocols, patients will more often retain anatomy vital to communication and swallowing. Despite a conservative approach, results of treatment may have debilitating effects. Rehabilitation efforts are directed towards a return to functional, if not normal, status. Although there are predictable trends in voice and swallowing disorders of patients with laryngeal cancer, posttreatment dysphonia and dysphagia are diverse in presentation. Considering the significant diversity of this population, speech pathologists should work closely with otolaryngologists to determine the most appropriate treatment for each patient. As this article demonstrates, voice and swallowing therapy are necessary components of the rehabilitation process following treatment for head and neck cancers. As always, treatment is tailored to the specific individual and based on information obtained during a thorough evaluation by a speech pathologist. Fortunately, with the help of voice and swallowing therapy, many patients return to functional communication and oral feeding.