Automatic Segmentation of Membranous Glottal Gap Area with U-Net-Based Architecture.

BACKGROUND: While videostroboscopy is recognized as the most popular approach for investigating vocal fold function, evaluating the numerical values, such as the membranous glottal gap area, remains too time consuming for clinical applications. METHODS: We used a total of 2507 videostroboscopy images from 137 patients and developed five U-Net-based deep-learning image segmentation models for automatic masking of the membranous glottal gap area. To further validate the models, we used another 410 images from 41 different patients. RESULTS: During development, all five models exhibited acceptable and similar metrics. While the VGG19 U-Net had a long inference time of 1654 ms, the other four models had more practical inference times, ranging from 16 to 138 ms. During further validation, Efficient U-Net demonstrated the highest intersection over union of 0.8455, the highest Dice coefficient of 0.9163, and the lowest Hausdorff distance of 1.5626. The normalized membranous glottal gap area index was also calculated and validated. Efficient U-Net and VGG19 U-Net exhibited the lowest mean squared errors (3.5476 and 3.3842) and the lowest mean absolute errors (1.8835 and 1.8396). CONCLUSIONS: Automatic segmentation of the membranous glottal gap area can be achieved through U-net-based architecture. Considering the segmentation quality and speed, Efficient U-Net is a reasonable choice for this task, while the other four models remain valuable competitors. The models' masked area enables possible calculation of the normalized membranous glottal gap area and analysis of the glottal area waveform, revealing promising clinical applications for this model. LEVEL OF EVIDENCE: NA Laryngoscope, 134:2835-2843, 2024.

Homology and the evolution of vocal folds in the novel avian voice box.

The origin of novel traits, those that are not direct modifications of a pre-existing ancestral structure, remains a fundamental problem in evolutionary biology. For example, little is known about the evolutionary and developmental origins of the novel avian vocal organ, the syrinx. Located at the tracheobronchial junction, the syrinx is responsible for avian vocalization, but it is unclear whether avian vocal folds are homologous to the laryngeal vocal folds in other tetrapods or convergently evolved. Here, we identify a core developmental program involved in avian vocal fold formation and infer the morphology of the syrinx of the ancestor of modern birds. We find that this ancestral syrinx had paired sound sources induced by a conserved developmental pathway and show that shifts in these signals correlate with syringeal diversification. We show that, despite being derived from different developmental tissues, vocal folds in the syrinx and larynx have similar tissue composition and are established through a strikingly similar developmental program, indicating that co-option of an ancestral developmental program facilitated the origin of vocal folds in the avian syrinx.

Novel and pragmatic exploration of variation in glottic parameters in non-parallel versus parallel vocal cord CT planes with potential reporting pitfalls.

Oblique orientation of vocal cord demands strict compliance, by technicians and clinicians, to the recommended parallel plane CT scan of larynx. Repercussions of non-compliance has never been investigated before. We aimed to observe influence of non-parallel vocal cord plane CT scan on qualitative and quantitative glottic parameters, keeping parallel plane CT as a standard for comparison. Simultaneous identification of potential suboptimal imaging sequelae as a result of unformatted CT plane was also identified. In this study we included 95 normal adult glottides and retrospectively analyzed their anatomy in two axial planes, non-parallel plane ① and parallel to vocal cord plane ②. Qualitative (shape, structures at glottic level) and quantitative (anterior commissure ACom, vocal cord width VCw, anteroposterior AP, transverse Tr, cross-sectional area CSA) glottic variables were recorded. Multivariate statistical analysis was used to predict pattern and their impact on glottic anatomy. Plane ① displayed supraglottic features in glottis; adipose (90.5%) and split thyroid laminae (70.6%). Other categorical variables: atypical shape, submental structures and multilevel vertebral crossing were also in majority. All glottic dimensions varied significantly between two planes with most in ACom (-5.8mm) and CSA (-15.0 mm2). In contrast, plane ② manifested higher VCw (>73%), Tr (66.3%), CSA (64.2%) and AP (44.2%) measurements. On correlation analysis, variation in ACom, CSA, Tr was positively associated with VC or plane obliquity (p<0.05). This variability was more in obese and short necked subjects. Change in one parameter also modified other significantly i.e., ACom versus AP and CSA versus Tr. Results indicated statistically significant change in subjective and objective anatomical parameters of glottis on non-application of appropriate CT larynx protocol for image analysis hence highlighting importance of image reformation.

Amount and spatial arrangement of muscle fibers in the human laryngeal Musculus ventricularis.

Textbooks and atlases of human macroscopic and microscopic anatomy of the larynx generally provide, if at all, only sparse information on the laryngeal Musculus ventricularis. However, several studies indicate that this muscle takes over the function of vestibular (ventricular) fold phonation after denervation of the Musculus vocalis. In the present study, 29 laryngeal specimens were coronally dissected at different levels, i.e. the anterior (L1), middle (L2), and posterior third of the vestibular fold (L3), and they underwent histological analysis. In all specimens the vestibular folds of both hemi-larynxes contained striated muscle bundles in variable amounts, representing a ventricularis muscle. These muscle bundles obviously originated from the lateral (external) and thyroepiglottic part of the thyroarytenoid muscle and the aryepiglottic part of the oblique arytenoid muscle, as has been described by other authors. The areas of vestibular folds and their amounts of ventricularis muscle bundles were measured using image analysis software (imageJ) by manual tracing. The mean area of the vestibular folds of both hemi-larynxes was 27.9 mm2 (SD [standard deviation] ± 9.17), and the area occupied by fibers of the ventricularis muscle was 1.5 mm2 (SD ± 1.78). Statistical analysis comparing the areas of both hemi-larynxes and levels resulted in no significant differences, except for the levels 2 and 3. In level 2, significantly more muscle fibers (2.0 mm2 ; SD ± 2.21) were detectable within the vestibular fold than in level 3 (0.9 mm2 ; SD ± 1.43). Level 1 also contained more muscle fibers (1.1 mm2 ; SD ± 1.06) than level 3, however, without significance. In conclusion, the laryngeal ventricularis muscle is present in the majority of reported cases. Since the muscle is of clinical relevance, it should be included in anatomical textbooks by default.

Métodos utilizados na análise morfométrica do trato vocal em cantores: revisão integrativa / Methods used in morphometric analysis of singers' vocal tracts: an integrative review

RESUMO Objetivo descrever os principais métodos quantitativos utilizados para análise morfométrica do trato vocal em cantores, suas aplicações e os principais segmentos estudados. Estratégia de pesquisa trata-se de uma revisão integrativa guiada pela pergunta condutora: "Quais os principais métodos quantitativos utilizados para análise morfométrica do trato vocal, suas aplicações e os principais segmentos estudados em cantores?". Foram utilizadas as bases eletrônicas PubMed, Scopus e BVS, por meio da chave de busca Vocal tract OR Oropharynx AND Morphology OR Geometry AND Evaluation OR Diagnosis AND voice, sem restrições de ano de publicação, sendo incluídos artigos em três idiomas: português, inglês e espanhol. Critérios de seleção: a seleção se deu de forma independente, por meio da leitura por pares e posterior aplicação dos critérios de exclusão e inclusão. Resultados dos 380 estudos, foram excluídos 30 duplicados. A partir da leitura dos resumos, aplicando-se os critérios de seleção, foram selecionados 18 estudos para leitura na íntegra, dos quais, 12 foram incluídos nesta revisão. A ressonância magnética foi o principal instrumento utilizado e os segmentos analisados incluíram desde o comprimento e volume do trato vocal, como segmentos isolados e suas particularidades morfológicas. Conclusão os métodos de quantificação morfológica do trato vocal integram instrumentos importantes para a avaliação instrumental do trato vocal e de seus segmentos, colaborando na atualização tecnológica em voz para melhor compreensão e intervenções fonoaudiológicas na voz cantada

ABSTRACT Purpose To describe the main quantitative methods used for morphometric analysis of the vocal tract in singers, their applications and the main segments studied. Research strategy This is an integrative review guided by the guiding question "What are the main quantitative methods used for morphometric analysis of the vocal tract, their applications and the main segments studied in singers?". The electronic databases PubMed, Scopus and VHL were used through the search key (Vocal tract OR Oropharynx) AND (Morphology OR Geometry) AND (Evaluation OR Diagnosis) AND (voice), without restriction of years of publication, including articles in three languages: Portuguese, English and Spanish. Selection criteria The selection took place independently through reading by pairs and subsequent application of exclusion and inclusion criteria. Results Of the 380 studies, 30 duplicates were excluded. After reading the abstracts, applying the selection criteria, 18 studies were selected for full reading, of which 12 were included in this review. Magnetic resonance imaging was the main instrument used and the analyzed segments ranged from the length and volume of the vocal tract to isolated segments and their morphological particularities. Conclusion Vocal tract morphological quantification methods are important for the instrumental evaluation of the vocal tract and its segments, a technological update leading to better understanding of singers' voice and therapeutical intervention.

Evolutionary loss of complexity in human vocal anatomy as an adaptation for speech.

Human speech production obeys the same acoustic principles as vocal production in other animals but has distinctive features: A stable vocal source is filtered by rapidly changing formant frequencies. To understand speech evolution, we examined a wide range of primates, combining observations of phonation with mathematical modeling. We found that source stability relies upon simplifications in laryngeal anatomy, specifically the loss of air sacs and vocal membranes. We conclude that the evolutionary loss of vocal membranes allows human speech to mostly avoid the spontaneous nonlinear phenomena and acoustic chaos common in other primate vocalizations. This loss allows our larynx to produce stable, harmonic-rich phonation, ideally highlighting formant changes that convey most phonetic information. Paradoxically, the increased complexity of human spoken language thus followed simplification of our laryngeal anatomy.

Mechanisms of sound production in deer mice (Peromyscus spp.).

Rodent diversification is associated with a large diversity of species-specific social vocalizations generated by two distinct laryngeal sound production mechanisms: whistling and airflow-induced vocal fold vibration. Understanding the relative importance of each modality to context-dependent acoustic interactions requires comparative analyses among closely related species. In this study, we used light gas experiments, acoustic analyses and laryngeal morphometrics to identify the distribution of the two mechanisms among six species of deer mice (Peromyscus spp.). We found that high frequency vocalizations (simple and complex sweeps) produced in close-distance contexts were generated by a whistle mechanism. In contrast, lower frequency sustained vocalizations (SVs) used in longer distance communication were produced by airflow-induced vocal fold vibrations. Pup isolation calls, which resemble adult SVs, were also produced by airflow-induced vocal fold vibrations. Nonlinear phenomena (NLP) were common in adult SVs and pup isolation calls, suggesting irregular vocal fold vibration characteristics. Both vocal production mechanisms were facilitated by a characteristic laryngeal morphology, including a two-layered vocal fold lamina propria, small vocal membrane-like extensions on the free edge of the vocal fold, and a singular ventral laryngeal air pocket known as the ventral pouch. The size and composition of vocal folds (rather than total laryngeal size) appears to contribute to species-specific acoustic properties. Our findings suggest that dual modes of sound production are more widespread among rodents than previously appreciated. Additionally, the common occurrence of NLP highlights the nonlinearity of the vocal apparatus, whereby small changes in anatomy or physiology trigger large changes in behavior. Finally, consistency in mechanisms of sound production used by neonates and adults underscores the importance of considering vocal ontogeny in the diversification of species-specific acoustic signals.

Vocal tract allometry in a mammalian vocal learner.

Acoustic allometry occurs when features of animal vocalisations can be predicted from body size measurements. Despite this being considered the norm, allometry sometimes breaks, resulting in species sounding smaller or larger than expected for their size. A recent hypothesis suggests that allometry-breaking mammals cluster into two groups: those with anatomical adaptations to their vocal tracts and those capable of learning new sounds (vocal learners). Here, we tested which mechanism is used to escape from acoustic allometry by probing vocal tract allometry in a proven mammalian vocal learner, the harbour seal (Phoca vitulina). We tested whether vocal tract structures and body size scale allometrically in 68 young individuals. We found that both body length and body mass accurately predict vocal tract length and one tracheal dimension. Independently, body length predicts vocal fold length while body mass predicts a second tracheal dimension. All vocal tract measures are larger in weaners than in pups and some structures are sexually dimorphic within age classes. We conclude that harbour seals do comply with anatomical allometric constraints. However, allometry between body size and vocal fold length seems to emerge after puppyhood, suggesting that ontogeny may modulate the anatomy-learning distinction previously hypothesised as clear cut. We suggest that seals, and perhaps other species producing signals that deviate from those expected from their vocal tract dimensions, may break allometry without morphological adaptations. In seals, and potentially other vocal learning mammals, advanced neural control over vocal organs may be the main mechanism for breaking acoustic allometry.

Individual differences in vocal size exaggeration.

The human voice carries socially relevant information such as how authoritative, dominant, and attractive the speaker sounds. However, some speakers may be able to manipulate listeners by modulating the shape and size of their vocal tract to exaggerate certain characteristics of their voice. We analysed the veridical size of speakers' vocal tracts using real-time magnetic resonance imaging as they volitionally modulated their voice to sound larger or smaller, corresponding changes to the size implied by the acoustics of their voice, and their influence over the perceptions of listeners. Individual differences in this ability were marked, spanning from nearly incapable to nearly perfect vocal modulation, and was consistent across modalities of measurement. Further research is needed to determine whether speakers who are effective at vocal size exaggeration are better able to manipulate their social environment, and whether this variation is an inherited quality of the individual, or the result of life experiences such as vocal training.

Length of the Cricoid and Trachea in Children: Predicting Intubation Depth to Prevent Subglottic Stenosis.

OBJECTIVE: Define the length of the subglottis and trachea in children to predict a safe intubation depth. METHODS: Patients <18 years undergoing rigid bronchoscopy from 2013 to 2020 were included. The carina and inferior borders of the cricoid and true vocal folds were marked on a bronchoscope and distances were measured. Patient age, weight, height, and chest height were recorded. Four styles of cuffed pediatric endotracheal tubes (ETT) were measured and potential positions of each cuff and tip were calculated within each trachea using five depth of intubation scenarios. Multivariate linear regression was performed to identify predictors of subglottic and tracheal length. RESULTS: Measurements were obtained from 210 children (141 male, 69 female), mean (SD) age 3.21 (3.66) years. Patient height was the best predictor of subglottic length (R2 : 0.418): Lengthsg (mm) = 0.058 * height (cm) + 2.8, and tracheal length (R2 : 0.733): Lengtht (mm) = 0.485 * height (cm) + 21.3. None of the depth of intubation scenarios maintained a cuff-free subglottis for all ETT styles investigated. A formula for depth of intubation: Lengthdi (mm) = 0.06 * height (cm) + 8.8 found that no ETT cuffs would be in the subglottis and all tips would be above the carina. CONCLUSION: Current strategies for determining appropriate depth of intubation pose a high risk of subglottic ETT cuff placement. Placing the inferior border of the vocal cords 0.06 * height (cm) + 8.8 from the superior border of the inflated ETT cuff may prevent subglottic cuff placement and endobronchial intubation. LEVEL OF EVIDENCE: 4 Laryngoscope, 132:S1-S10, 2022.

Selection on vocal output affects laryngeal morphology in rats.

Although laryngeal morphology often reflects adaptations for vocalization, the structural consequences of selection for particular aspects of vocal behavior remain poorly understood. In this study, we investigated the effects of increased ultrasonic calling in pups on the adult larynx morphology in selectively bred rat lines. Laryngeal morphology was assessed using multiple techniques: mineralized cartilage volumes were compared in 3D-models derived from microCT scans, internal structure was compared using clearing and staining procedures combined with microscopy, cellular structure was compared using histology and microscopy, and element composition was assessed with scanning energy dispersive X-ray spectroscopy. Our results show that adult rats from lines bred to produce ultrasonic calls at higher rates as pups have shorter vocal folds and a more mineralized thyroid cartilage compared to rats bred to produce ultrasonic calls at lower rates. The change in vocal fold length appears to account for differences in low-frequency calls in these two rat lines. We suggest that the observed increases in mineralization of the thyroid cartilage in the high-ultrasound lineage provide increased reinforcement of the laryngeal structure during ultrasonic call production. Our findings therefore demonstrate an effect of selection for vocal behavior on laryngeal morphology, with acoustic consequences.

A Dissectional Study of the Level of Anterior Commissure of the Larynx.

PURPOSE: Precise knowledge of the level of the vocal fold as projected on the external thyroid cartilage is of critical importance for the performance of many surgical approaches. This study aims to identify the level of the anterior commissure, as well as the lengths of the vocal muscle and arytenoid cartilage in Turkish population. MATERIALS AND METHOD: Specimens were collected after autopsy from the Council of Forensic Medicine. One hundred human larynges (52 men, 48 women; age range: 25-80 years) were dissected under a stereomicroscope. Projection of the vocal fold was analyzed in relation to the superior thyroid (A) and the inferior border of the thyroid cartilage (B). Then, the larynx was dissected parallel to the level of the vocal fold to measure the length of the vocal muscle (C) and the length of the interarytenoid space (D). RESULTS: The mean value of the "a" was 9.15 ± 1.99 mm in male and 9.38 ± 3.43 mm in female. Mean value of the "b" was 10.54 ± 1.73 mm and 8.88 ± 1.81 mm in male and female, respectively. The mean value of the parameter corresponding the length of vocal muscle which was "c" was found 15.00 ± 3.18 mm in male and 12.88 ± 4.12 mm in female. The mean value of the interarytenoid space "d" was 8.31 ± 1.76 mm in male and 8.13 ± 1.90 mm in female. Comparing between genders, no statistical differences were observed in parameters of a, c, d, a + b, a + b/2 (P > .05). However, the difference with female and male for the parameters of b and c + d was statistically significant (P < .05). CONCLUSION: Our results indicate that the anterior commissure projects slightly above the midline height for male and at the level to slightly below in female subjects in Turkish population.

Volumetric Analysis of the Vocal Folds Using Computed Tomography: Effects of Age, Height, and Gender.

OBJECTIVE: The goal of this study was to use computed tomography (CT) volumetric analysis to assess the effect of age, gender, height, body mass index (BMI), and ethnicity on vocal fold volume in patients with normal larynges. STUDY DESIGN: Retrospective cross-sectional study. METHODS: Vocal fold length, width, and height were measured in a total of 105 patients without a history of laryngeal or thyroid pathology on thin-section soft-tissue neck CTs. The product of the three dimensions was used to calculate vocal fold volume. Simple and multiple linear regression analyses were used to assess for an association between vocal fold volume and age, gender, height, BMI, and ethnicity. Intraclass correlation coefficients (ICCs) were estimated to evaluate the degree interobserver and intraobserver agreement. RESULTS: Vocal fold volume was not associated with age, BMI, or ethnicity. Gender-adjusted height (P = .002) and height-adjusted gender (P = .016) were significantly associated with volume. Height remained significantly associated with volume after stratifying by gender (P < 0.001). There was moderate-to-good correlation in both interobserver (ICC = 0.690 to 0.761) and intraobserver (ICC = 0.733 to 0.873) agreement. CONCLUSION: Age was not associated with vocal fold volume, which is in accordance with several prior negative studies. Age-related vocal fold atrophy may not substantially contribute to presbyphonia symptoms, but other processes such as changes in the extracellular matrix may play a larger role. However, both gender and height were independently associated with vocal fold volume. LEVEL OF EVIDENCE: 4 Laryngoscope, 131:E240-E247, 2021.

Structure, Function and Insertion of the Human Vocal Folds.

We here summarize the structures of the laryngeal vocal fold as well as its insertion structures at the anterior commissure and at the area of the vocal process and place these findings within the context of biomechanical, functional, and clinical implications.

Preparation of the Rat Vocal Fold for Neuromuscular Analyses.

The purpose of this tutorial is to describe the preparation of the rat vocal fold for histochemical neuromuscular study. This protocol outlines procedures for rat laryngeal dissection, flash-freezing, and cryosectioning of the vocal folds. This study describes how to cryosection vocal folds in both longitudinal and cross-sectional planes. A novelty of this protocol is the laryngeal tracking during cryosectioning that ensures accurate identification of the intrinsic laryngeal muscles and reduces the chance of tissue loss. Figures demonstrate the progressive cryosectioning in both planes. Twenty-nine rat hemi-larynges were cryosectioned and tracked from the emergence of the thyroid cartilage to the appearance of the first section that included the full vocal fold. The full vocal fold was visualized for all animals in both planes. There was high variability in the distance from the appearance of the thyroid cartilage to the appearance of the full vocal fold in both planes. Weight was not correlated to depth of laryngeal landmarks, suggesting individual variability and other factors related to tissue preparation may be responsible for the high variability in the appearance of landmarks during sectioning. This study details a methodology and presents morphological data for preparing the rat vocal fold for histochemical neuromuscular investigation. Due to high individual variability, laryngeal landmarks should be closely tracked during cryosectioning to prevent oversectioning tissue and tissue loss. The use of a consistent methodology, including adequate tissue preparation and awareness of landmarks within the rat larynx, will assist with consistent results across studies and aid new researchers interested in using the rat vocal fold as a model to investigate laryngeal neuromuscular mechanisms.

Differential DNA methylation of vocal and facial anatomy genes in modern humans.

Changes in potential regulatory elements are thought to be key drivers of phenotypic divergence. However, identifying changes to regulatory elements that underlie human-specific traits has proven very challenging. Here, we use 63 reconstructed and experimentally measured DNA methylation maps of ancient and present-day humans, as well as of six chimpanzees, to detect differentially methylated regions that likely emerged in modern humans after the split from Neanderthals and Denisovans. We show that genes associated with face and vocal tract anatomy went through particularly extensive methylation changes. Specifically, we identify widespread hypermethylation in a network of face- and voice-associated genes (SOX9, ACAN, COL2A1, NFIX and XYLT1). We propose that these repression patterns appeared after the split from Neanderthals and Denisovans, and that they might have played a key role in shaping the modern human face and vocal tract.

Automatic vocal tract landmark localization from midsagittal MRI data.

The various speech sounds of a language are obtained by varying the shape and position of the articulators surrounding the vocal tract. Analyzing their variations is crucial for understanding speech production, diagnosing speech disorders and planning therapy. Identifying key anatomical landmarks of these structures on medical images is a pre-requisite for any quantitative analysis and the rising amount of data generated in the field calls for an automatic solution. The challenge lies in the high inter- and intra-speaker variability, the mutual interaction between the articulators and the moderate quality of the images. This study addresses this issue for the first time and tackles it by means of Deep Learning. It proposes a dedicated network architecture named Flat-net and its performance are evaluated and compared with eleven state-of-the-art methods from the literature. The dataset contains midsagittal anatomical Magnetic Resonance Images for 9 speakers sustaining 62 articulations with 21 annotated anatomical landmarks per image. Results show that the Flat-net approach outperforms the former methods, leading to an overall Root Mean Square Error of 3.6 pixels/0.36 cm obtained in a leave-one-out procedure over the speakers. The implementation codes are also shared publicly on GitHub.

Prediction of anterior commissure exposure of vocal folds prior to laryngeal microscopic surgery: Parameters reflecting occipito-atlanto-axial complex extension capacity.

OBJECTIVE: The purpose of this prospective study was to identify preoperative predictors of unstable exposure of vocal folds with focus on the anterior commissure (AC) prior to Laryngeal Microscopic Surgery. METHODS AND MATERIALS: Patients were classified into four groups based on the degree of AC exposure during Laryngeal Microscopic Surgery, and for the analysis, these groups were subdivided into unstable exposure and stable exposure subgroups depending on whether external manipulation was required to achieve AC exposure. Correlation of the degree of AC exposure with demographics, physical measurements, and anatomical measurements taken using landmarks in simple radiograph were evaluated. Receiver operating characteristic curve analysis was used to determine optimal cutoff values to predict unstable AC exposure. RESULTS: Fifty-nine patients were included in the analyses. Thyroid-mandible angle (TMA) in the extended position, thyroid-mental distance (TMD) ratio, and TMA difference in the neutral and extended positions were significantly correlated with the degree of AC exposure. However, only a TMD ratio of <1.25 reliably predicted unstable AC exposure. CONCLUSIONS: TMD ratio of <1.25 reliably predicted unstable AC exposure. If there is no increase of the distance between the thyroid notch and the mental prominence (TMD) more than 25% on neck extension, the probability of getting stable exposure of the anterior commissure is low.

A Convolutional Neural Network for Real Time Classification, Identification, and Labelling of Vocal Cord and Tracheal Using Laryngoscopy and Bronchoscopy Video.

BACKGROUND: The use of artificial intelligence, including machine learning, is increasing in medicine. Use of machine learning is rising in the prediction of patient outcomes. Machine learning may also be able to enhance and augment anesthesia clinical procedures such as airway management. In this study, we sought to develop a machine learning algorithm that could classify vocal cords and tracheal airway anatomy real-time during video laryngoscopy or bronchoscopy as well as compare the performance of three novel convolutional networks for detecting vocal cords and tracheal rings. METHODS: Following institutional approval, a clinical dataset of 775 video laryngoscopy and bronchoscopy videos was used. The dataset was divided into two categories for use for training and testing. We used three convolutional neural networks (CNNs): ResNet, Inception and MobileNet. Backpropagation and a mean squared error loss function were used to assess accuracy as well as minimize bias and variance. Following training, we assessed transferability using the generalization error of the CNN, sensitivity and specificity, average confidence error, outliers, overall confidence percentage, and frames per second for live video feeds. After the training was complete, 22 models using 0 to 25,000 steps were generated and compared. RESULTS: The overall confidence of classification for the vocal cords and tracheal rings for ResNet, Inception and MobileNet CNNs were as follows: 0.84, 0.78, and 0.64 for vocal cords, respectively, and 0.69, 0.72, 0.54 for tracheal rings, respectively. Transfer learning following additional training resulted in improved accuracy of ResNet and Inception for identifying the vocal cords (with a confidence of 0.96 and 0.93 respectively). The two best performing CNNs, ResNet and Inception, achieved a specificity of 0.985 and 0.971, respectively, and a sensitivity of 0.865 and 0.892, respectively. Inception was able to process the live video feeds at 10 FPS while ResNet processed at 5 FPS. Both were able to pass a feasibility test of identifying vocal cords and tracheal rings in a video feed. CONCLUSIONS: We report the development and evaluation of a CNN that can identify and classify airway anatomy in real time. This neural network demonstrates high performance. The availability of artificial intelligence may improve airway management and bronchoscopy by helping to identify key anatomy real time. Thus, potentially improving performance and outcomes during these procedures. Further, this technology may theoretically be extended to the settings of airway pathology or airway management in the hands of experienced providers. The researchers in this study are exploring the performance of this neural network in clinical trials.

Characterization of the Continuous Elastic Parameters of Porcine Vocal Folds.

This paper presents an evaluation of the elastic properties of porcine vocal folds through uniaxial tensile tests. Inferior vocal fold tissue samples were subjected to tension in the longitudinal direction while digital image correlation techniques were employed to determine strain values throughout the tests. The stress-strain results showed a low-strain linear region, followed by both a nonlinear exponential and then a higher strain linear region. Data from 16 porcine vocal fold samples were analyzed following a similar optimization method as proposed in prior studies [1] to yield continuous model parameters which describe the elastic properties of the tissue. The average low and high strain linear modulus values were found to be 17.86 kPa and 609.27 kPa, respectively. The model also identified the location of two transition points: p1, describing the transition from the low-strain linear region to an exponential region at 0.122 ±â€¯0.058 mm/mm and p2, describing the transition from the exponential to the high strain linear region at 0.308 ±â€¯0.069 mm/mm. The exponential region of the averaged data set was found to be described by the relationship [Formula: see text] kPa. In addition to locating transition points, the optimization method maintained modulus continuity across all strain values. Averaged elastic modulus values across strain from 0 to 0.40 mm/mm were compared to representative low and high strain linear modulus which were measured at 0.05 and 0.35 mm/mm, respectively. Statistically significant differences were found among all strain intervals between the two transition points and the linear modulus values. These results indicate the need to consider the location of transition points and further highlight the nonlinearity and changes in elastic modulus which are especially important when using excised porcine vocal folds as a model for phonation. The results quantify continuous linear and nonlinear parameters describing the elastic properties which can be used as a framework for future excised larynx tests and while evaluating the dynamics of sound production, which rely heavily on the elastic properties of the tissue.