Speech Intelligibility Scores Research Articles

Chinese speech intelligibility and speech intelligibility index for the elderly

The speech intelligibility index (SII) and speech transmission index (STI) are widely accepted objective metrics for assessing speech intelligibility. In previous work, the relationship between STI and Chinese speech intelligibility (CSI) scores was studied. In this paper, the relationship between SII and CSI scores in rooms for the elderly aged 60–69 and over 70 is investigated by using auralization method under different background noise levels (40dBA and 55dBA) and different reverberation times. The results show that SII has good correlation with CSI score of the elderly. To get the same CSI score as the young adults, the elderly need a larger SII value, and the value increases with the increase of the age for the elderly. Since hearing loss of the elderly is considered in the calculation of SII, the difference in the required SII between the elderly and young is less than that of the required STI under the same CSI score condition. This indicates that SII is a more consistent evaluation criterion for different ages.

Effectiveness of secondary furlow palatoplasty with buccal myomucosal flap in correction of velopharyngeal insufficiency in patients with cleft palate

ObjectivesThe main purpose of this study was evaluation of the effectiveness of secondary furlow palatoplasty with buccal myomucosal flap (FPBF) for the treatment of velopharyngeal insufficiency (VPI) in patients with a cleft palate who were treated with two flap palatoplasty (TFP) in their primary palate repair.Material and methodsTwenty-three medically free children aged 4–8 years with non-syndromic and previously repaired cleft palate via TFP participated in the study. All patients received secondary surgery following the technique of FPBF. Preoperative speech evaluation was done before the secondary repair and 3 months after the surgery using a hypernasal speech scale, speech intelligibility scale, and nasopharyngoscopy.ResultsA statistically significant improvement was observed regarding the degree of hypernasality and speech intelligibility while comparing the preoperative scores after the primary surgery to the postoperative scores after the secondary surgery.In addition, a statistically significant improvement was found in the nasopharyngoscopic assessment.ConclusionsThe incorporation of a buccal myomucosal flap with Furlow palatoplasty was successful in improving hypernasality, speech intelligibility, and nasopharyngoscopic scores in patients with cleft palate.Trial registrationclinicaltrials.gov (NCT05626933).Clinical relevanceThis technique might be the surgical technique of choice while treating patients who are suffering from VPI after cleft palate repair.

A Fused Deep Denoising Sound Coding Strategy for Bilateral Cochlear Implants.

Cochlear implants (CIs) provide a solution for individuals with severe sensorineural hearing loss to regain their hearing abilities. When someone experiences this form of hearing impairment in both ears, they may be equipped with two separate CI devices, which will typically further improve the CI benefits. This spatial hearing is particularly crucial when tackling the challenge of understanding speech in noisy environments, a common issue CI users face. Currently, extensive research is dedicated to developing algorithms that can autonomously filter out undesired background noises from desired speech signals. At present, some research focuses on achieving end-to-end denoising, either as an integral component of the initial CI signal processing or by fully integrating the denoising process into the CI sound coding strategy. This work is presented in the context of bilateral CI (BiCI) systems, where we propose a deep-learning-based bilateral speech enhancement model that shares information between both hearing sides. Specifically, we connect two monaural end-to-end deep denoising sound coding techniques through intermediary latent fusion layers. These layers amalgamate the latent representations generated by these techniques by multiplying them together, resulting in an enhanced ability to reduce noise and improve learning generalization. The objective instrumental results demonstrate that the proposed fused BiCI sound coding strategy achieves higher interaural coherence, superior noise reduction, and enhanced predicted speech intelligibility scores compared to the baseline methods. Furthermore, our speech-in-noise intelligibility results in BiCI users reveal that the deep denoising sound coding strategy can attain scores similar to those achieved in quiet conditions.

Altered Auditory Feedback Improves Speech Intelligibility in Individuals Diagnosed with Parkinson’s Disease: A Pilot Study

Introduction: The purpose of the current study was to evaluate speech intelligibility and speech rate in individuals with Parkinson’s disease using altered auditory feedback (AAF) in comparison to age-matched healthy controls. Based on prior research, it is hypothesized that the use of AAF will improve participants’ speech intelligibility. Methods: Five individuals with Parkinson’s disease (clinical group) and five age-matched healthy controls aged 50 years and above participated in this pre-/post-group experiment. All participants completed reading and monologue tasks with and without AAF. The AAF delayed the participants’ auditory feedback by 150 ms and altered it by 1/20 octave. Thirty-two naïve listeners listened to the recorded participants’ speech samples (with and without AAF) randomly and rated the participants’ speech intelligibility. Results: Results of the mixed model ANOVA revealed that participants in the control group had significantly higher speech intelligibility scores than the participants in the clinical group. Post hoc tests indicated that only participants in the clinical group were more intelligible with AAF than without AAF. Participants in the clinical group had better speech intelligibility when reading the passage than speaking the monologue. With regards to speech rate, participants spoke more slowly with AAF than without. Finally, there was a negative correlation between speech rate and speech intelligibility. Discussion: The current findings support our hypothesis that AAF improves speech intelligibility in people with Parkinson’s disease, though not in healthy controls. Future research should evaluate the long-term effect of AAF use on speech intelligibility in people with Parkinson’s disease.

Relationship Between Chinese Speech Intelligibility of Elderly and Speech Transmission Index

In this paper, the relationship between Chinese speech intelligibility (CSI) scores of the elderly aged 60–69 and over 70 years old, and speech transmission index (STI) were investigated through the auralization method under different reverberation time and background noise levels (BNL, 40 dBA and 55 dBA). The results show that the CSI scores of the elderly are significantly worse than those of young adults. For the elderly over 70, the CSI scores become much lower than those of young adults. To be able to achieve the same CSI, the elderly, especially those over 70, need much higher STI and greater SNR than the young. The elderly aged 60–69 and over 70 need to improve their STI by 0.419 and 0.058 respectively under BNL 40 dBA, as well as 0.282 and 0.072 respectively under BNL 55 dBA, so as to obtain the same CSI scores as the young adults.

Investigation of Development of Phonology using Artificial Intelligence: A Preliminary Study

Background: Investigation of the development of acquisition of phonological processes is a very important phenomenon to predict language development in childhood. Phonological development refers to the gradual acquisition of an adult-like speech system used to convey meaning in a language. Adequate phonological development can lead to higher speech intelligibility, essential for perception and speech production. The test battery approach for phonological assessment includes formal as well as informal testing modalities. Speech intelligibility is a vital parameter for the assessment of phonological development. This is done via scoring collected audio speech samples, followed by IPA transcription and scoring and analysis of data manually. The primary objective of the study was to investigate the speech intelligibility score across real-timed stimuli (rhyming, storytelling, open-ended questions, and picture description tasks) across two age categories, i.e., 3-4 and 4-5 years among 40 Hindi-speaking children. Methods: A total of 40 children (15 girls and 25 boys) in the age range of 3 to 5 years were randomly selected from a junior high school in New Delhi, India. Groupwise speech intelligibility score was calculated for rhyming, storytelling, open-ended questions, and picture description tasks. Results: It was observed that the speech intelligibility score for open-ended questions was significantly higher in the 4-5 years group (94.09 ± 3.7) than the 3-4 years (89.83 ± 8.2), t (1, 38) =2.12, p =0.04). However, no significant improvement in rhyming task [t (1, 38) =1.14, p =0.18), storytelling [t (1, 38) =1.81, p =0.07), and picture description [t (1, 38) =1.89, p =0.06) was observed. Conclusion: Chances of error in conventional method of phonology assessment still persists which may be controlled using Artificial Intelligence (AI). The target phonetic stimulus can be customized based on the linguistic experience and environment.

A Deep Denoising Sound Coding Strategy for Cochlear Implants.

Cochlear implants (CIs) have proven to be successful at restoring the sensation of hearing in people who suffer from profound sensorineural hearing loss. CI users generally achieve good speech understanding in quiet acoustic conditions. However, their ability to understand speech degrades drastically when background interfering noise is present. To address this problem, current CI systems are delivered with front-end speech enhancement modules that can aid the listener in noisy environments. However, these only perform well under certain noisy conditions, leaving quite some room for improvement in more challenging circumstances. In this work, we propose replacing the CI sound coding strategy with a deep neural network (DNN) that performs end-to-end speech denoising by taking the raw audio as input and providing a denoised electrodogram, i.e., the electrical stimulation patterns applied to the electrodes across time. We specifically introduce a DNN that emulates a common CI sound coding strategy, the advanced combination encoder (ACE). We refer to the proposed algorithm as 'Deep ACE'. Deep ACE is designed not only to accurately code the acoustic signals in the same way that ACE would but also to automatically remove unwanted interfering noises, without sacrificing processing latency. The model was optimized using a CI-specific loss function and evaluated using objective measures as well as listening tests in CI participants. Results show that, based on objective measures, the proposed model achieved higher scores when compared to the baseline algorithms. Also, the proposed deep learning-based sound coding strategy gave eight CI users the highest speech intelligibility scores.

No Musician Advantage in the Perception of Degraded-Fundamental Frequency Speech in Noisy Environments.

Pitch variations of the fundamental frequency (fo) contour contribute to speech perception in noisy environments, but whether musicians confer an advantage in speech in noise (SIN) with altered fo information remains unclear. This study investigated the effects of different levels of degraded fo contour (i.e., conveying lexical tone or intonation information) on musician advantage in speech-in-noise perception. A cohort of native Mandarin Chinese speakers, comprising 30 trained musicians and 30 nonmusicians, were tested on the intelligibility of Mandarin Chinese sentences with natural, flattened-tone, flattened-intonation, and flattened-all fo contours embedded in background noise masked under three signal-to-noise ratios (0, -5, and -9 dB). Pitch difference thresholds and innate musical skills associated with speech-in-noise benefits were also assessed. Speech intelligibility score improved with increasing signal-to-noise level for both musicians and nonmusicians. However, no musician advantage was observed for identifying any type of flattened-fo contour SIN. Musicians exhibited smaller fo pitch discrimination limens than nonmusicians, which correlated with benefits for perceiving speech with intact tone-level fo information. Regardless of musician status, performance on the pitch and accent musical-skill subtests correlated with speech intelligibility score. Collectively, these results provide no evidence for a musician advantage for perceiving speech with distorted fo information in noisy environments. Results further show that perceptual musical skills on pitch and accent processing may benefit the perception of SIN, independent of formal musical training. Our findings suggest that the potential application of music training in speech perception in noisy backgrounds is not contingent on the ability to process fo pitch contours, at least for Mandarin Chinese speakers. https://doi.org/10.23641/asha.23706354.

The effect of listener head orientation on front-to-rear speech intelligibility in an automotive cabin

The listener head orientation significantly affects speech intelligibility (SI) in the small acoustic space of an automobile, owing to near-field binaural listening and special acoustic conditions such as early reflections and seat-back occlusions. However, situations involving a back-row listener and a front-row speaker (i.e., front-to-rear scenarios) have not yet been studied. Such scenarios are dominated by non-uniform early reflections rather than direct sounds, making them distinct from cases considered in previous studies. This study investigates the effect of listener head orientation on SI in front-to-rear scenarios in an automobile, using both the speech transmission index (STI) and subjective experiments. A virtual speaker (i.e., a loudspeaker) is located in the driver’s seat and faces forward. Binaural room impulse responses are measured on a dummy head in the three rear-row seats with different head orientations. Using the binaural room impulse responses, the corresponding STIs are calculated, and the Chinese SI scores are measured virtually via headphones. The results show that the SI variation range for various listener head orientations in front-to-rear scenarios is greatly reduced compared with that in the rear-to-front scenarios considered in previous studies. Overall, the effect of listener head orientation on SI is mostly negligible when a listener in the rear row listens to a speaker in the front, with variations in the SI score of no more than 5%. The present study provides a supplement to previous studies and helps to deepen understanding of the effect of multiple factors on SI in an automotive cabin.

Cochlear Implantation in Adults with Post-lingual Hearing Loss: Clinico-Demographical Study and Outcomes in the Current Times.

To describe clinico-demographical profile for post-lingual hearing loss in the adults undergoing cochlear implantation and their outcomes. A retrospective chart review was conducted including adult patients (> 18 years) with bilateral post-lingual severe to profound hearing loss who underwent cochlear implantation in a tertiary care hospital of north India. The clinico-demographical details were collected and the outcomes of the procedure were assessed in terms of speech intelligibility scores, usage and satisfaction scores. Twenty-one patients were included with a mean age of 38.6 years with 15 males and 6 females. The major cause of deafness was infections followed by ototoxicity. The complication rate was 4.8%. Preoperative SDS was not recordable in any of the patients. The mean postoperative SDS was 74%, with no issue of device malfunction in a mean follow-up of 44 months. Cochlear implantation is a safe surgery with good outcomes in post-lingually deafened adults with the major cause for deafness being infections.

Comparison between Predictive and Measurement Methods of Speech Intelligibility for Educational Rooms of Different Sizes with and without HVAC Systems

Heating ventilation and air conditioning (HVAC) systems represent one of the main noise sources inside classrooms. This explain why HVAC systems require careful design, competent installation and balancing, and regular maintenance. Many factors influence the classroom acoustical design, such as air handlers or fans, the velocity of air inside the classroom, as well as the size and acoustical treatment of ducts, returns, and diffusers. Acoustic parameters, including background-noise levels, reverberation time, and intelligibility, were analyzed in 17 classrooms at the Università Politecnica in the Marche region. The study of intelligibility was performed by measuring the objective parameters in situ and using prediction methods to determine the intelligibility score. The relationship between speech intelligibility measurements and speech intelligibility calculation has been studied. The relationship between the STI values with the background-noise levels and the reverberation time was also studied. This research shows that a comparison between predictive methods and measurement methods results in speech intelligibility for classrooms of different sizes with and without HVAC systems. The current method of calculating the voice transmission index (STI), proposed by national and international standards, has been used to determine speech intelligibility scores in classrooms. The results show that the calculation tool has computational robustness allowing its use in preliminary evaluations of speech intelligibility, design of the optimal type of school buildings, and sound amplification systems in classrooms that comply with Italian regulations.

Effects of visual information on the face masked speech communication in Covid-secure classrooms

Acoustical measurements and speech intelligibility tests were carried out to investigate the effects of masks on speech communication experienced in real Covid-secure university classrooms during the pandemic. Face-masked speech levels and noise levels were measured to understand the acoustical effects of masks on speech sounds during 15 multiple lectures in 3 university classrooms. The speech intelligibility scores were also evaluated for lower and higher SNR (signal-to-noise ratio) conditions, and for with and without the presence of visual information conditions to investigate the effects of both the acoustic and visual signals in understanding speech communication in actual classroom situations. In the 3 active university classrooms the students experienced on average: speech levels of 55.1 dBA (σ = 5.5 dBA), noise levels of 42.3 dBA (σ = 3.9 dBA), and a speech-to-noise ratio of 12.8 dBA σ = 5.2 dBA). The mean SNR values at the listener’s position for the 15 lectures varied from 3.6 dBA to 20.0 dBA. The use of a portable sound amplification system increases the face-masked speech levels mostly at mid and high frequencies (500–4 kHz), thus it can be more useful for achieving higher SNR values in classrooms. The presence of visual cues have little effect on achieving more higher speech intelligibility scores in higher SNR conditions. The present results show that visual obstruction of the talker’s mouth decreases speech intelligibility scores by a maximum of 10% in lower SNR conditions, particularly at a SNR of 6 dBA or lower.

Expression-Preserving Face Frontalization Improves Visually Assisted Speech Processing

Face frontalization consists of synthesizing a frontal view from a profile one. This paper proposes a frontalization method that preserves non-rigid facial deformations, i.e. facial expressions. It is shown that expression-preserving frontalization boosts the performance of visually assisted speech processing. The method alternates between the estimation of (i) the rigid transformation (scale, rotation, and translation) and (ii) the non-rigid deformation between an arbitrarily-viewed face and a face model. The method has two important merits: it can deal with non-Gaussian errors in the data and it incorporates a dynamical face deformation model. For that purpose, we use the Student’s t-distribution in combination with a Bayesian filter in order to account for both rigid head motions and time-varying facial deformations, e.g. caused by speech production. The zero-mean normalized cross-correlation score is used to evaluate the ability of the method to preserve facial expressions. The method is thoroughly evaluated and compared with several state of the art methods, either based on traditional geometric models or on deep learning. Moreover, we show that the method, when incorporated into speech processing pipelines, improves word recognition rates and speech intelligibility scores by a considerable margin.

The German hearing in noise test with a female talker: development and comparison with German male speech test

PurposeThe aim of the study was to develop the German Hearing in Noise Test (HINT) with female speaker by fulfilling the recommendations by International Collegium of Rehabilitative Audiology (ICRA) for using a female speaker to create new multilingual speech tests and to determine norms and to compare these norms with German male speech tests—the male speakers HINT and the Oldenburg Sentence Test (OLSA).MethodsThe HINT with a female speaker consists of the same speech material as the male speaking HINT. After recording the speech material, 10 normal hearing subjects were included to determine the performance–intensity function (PI function). 24 subjects were part of the measurements to determine the norms and compare them with the norms of male HINT and OLSA. Comparably, adaptive, open-set methods under headphones (HINT) and sound field (OLSA) were used.ResultsAcoustic phonetic analysis demonstrated significant difference in mean fundamental frequency, its range and mean speaking rate between both HINT speakers. The calculated norms by three of the tested four conditions of the HINT with a female speaker are not significantly different from the norms with a male speaker. No significant effect of the speaker’s gender of the first HINT measurement and no significant correlation between the threshold results of the HINT and the OLSA were determined.ConclusionsThe Norms for German HINT with a female speaker are comparable to the norms of the HINT with a male speaker. The speech intelligibility score of the HINT does not depend on the speakers’ gender despite significant difference of acoustic–phonetic parameters between the female and male HINT speaker’s voice. Instead, the speech intelligibility rating must be seen as a function of the used speech material.

Deep Learning-Based Non-Intrusive Multi-Objective Speech Assessment Model With Cross-Domain Features

This study proposes a cross-domain multi-objective speech assessment model, called MOSA-Net, which can simultaneously estimate the speech quality, intelligibility, and distortion assessment scores of an input speech signal. MOSA-Net comprises a convolutional neural network and bidirectional long short-term memory architecture for representation extraction, and a multiplicative attention layer and a fully connected layer for each assessment metric prediction. Additionally, cross-domain features (spectral and time-domain features) and latent representations from self-supervised learned (SSL) models are used as inputs to combine rich acoustic information to obtain more accurate assessments. Experimental results show that in both seen and unseen noise environments, MOSA-Net can improve the linear correlation coefficient (LCC) scores in perceptual evaluation of speech quality (PESQ) prediction, compared to Quality-Net, an existing single-task model for PESQ prediction, and improve LCC scores in short-time objective intelligibility (STOI) prediction, compared to STOI-Net, an existing single-task model for STOI prediction. Moreover, MOSA-Net can be used as a pre-trained model to be effectively adapted to an assessment model for predicting subjective quality and intelligibility scores with a limited amount of training data. Experimental results show that MOSA-Net can improve LCC scores in mean opinion score (MOS) predictions, compared to MOS-SSL, a strong single-task model for MOS prediction. We further adopt the latent representations of MOSA-Net to guide the speech enhancement (SE) process and derive a quality-intelligibility (QI)-aware SE (QIA-SE) approach. Experimental results show that QIA-SE outperforms the baseline SE system with improved PESQ scores in both seen and unseen noise environments over a baseline SE model.

Assessing the benefits of virtual speaker lateralization for binaural speech intelligibility over the Internet

Binaural speech intelligibility tests using headphones were conducted remotely through the Internet using virtual speaker lateralization at azimuth angles around the front within ±45 degrees, combined with babble noise with signal to noise ratios (SNR) from -12 to -4 dB. Monophonic speech recordings, selected from the publicly available Mozilla Common Voice speech corpus in Spanish, were binaurally processed using a database of generic Head-Related Transfer Functions (HRTF) to produce virtual speaker lateralization. A common signal of babble noise was mixed to both the left and right binaural signals, with different SNR. Speech intelligibility tests were conducted through a Google Forms questionnaire containing Internet links to YouTube videos embedding the corresponding test audio files. Symbolic text difference metrics: Levenshtein distance, and word error rate (WER), commonly used in the field of automatic speech recognition (ASR), were used to automatically calculate estimations of speech intelligibility scores, more conventionally used for (human) speech intelligibility research at the word level, together with microscopic speech intelligibility scores at the phonetic symbol level. Speech reception thresholds (SRT), and intelligibility slopes were determined, showing that the known beneficial effect of spatial release from masking (SRM) in improving speech intelligibility when the speaker is oriented at lateral azimuth angles relative to the listener, is preserved when using virtual speaker lateralization, and by the use of Internet transmitted audio using headphones. Results show that for azimuth angles of 20 and 30 degrees, left or right, speech intelligibility improves with an average unmasking benefit of 3.7±0.7 dB SRM, and an average intelligibility slope of 5.6±0.8%/dB, while also maintaining at these angles, a desirable impression of a virtual speaker with a reasonably frontal orientation. This provides a useful technique for improved Internet speech delivery.

A precise glossectomy for tongue cancer adjacent to or crossing the midline: a novel anatomical unit resection surgery

Previous studies have lacked a clear anatomical and functional definition of glossectomy for tongue cancer adjacent to or crossing the midline (TCML). The aim of this study was to provide a novel surgical approach based on anatomical unit resection surgery to treat TCML. A total of 120 patients with TCML who had undergone radical surgery were recruited retrospectively into the study. The patients who were treated with compartment surgery formed the control group; those treated with anatomical unit resection surgery formed the experimental group. The TCML was classified into cancer adjacent to the midline, cancer invading but not breaching the contralateral musculus verticalis linguae–genioglossus complex (MGC), and cancer breaching the contralateral MGC. No significant difference in the overall survival rate was found between the experimental and control groups overall (P = 0.853) or by TCML classification. In patients with cancer adjacent to the midline, the swallowing score (P = 0.040) and cosmetic outcome (P = 0.015) were significantly better in the experimental group than in the control group. For patients with cancer invading but not breaching the contralateral MGC, the speech intelligibility score (P = 0.001), swallowing score (P = 0.002), and cosmetic outcome (P = 0.037) were significantly better in the experimental group than in the control group. Anatomical unit resection surgery was found to provide a precise surgical treatment to address tongue cancer adjacent to or crossing the midline and maximally maintain tongue tissue and function.

On the Relationship between Speech Intelligibility and Fluency Indicators among English-Speaking Individuals with Parkinson's Diseases.

The purpose of the study is to investigate how much of variance in Parkinson's Disease (PD) individuals' speech intelligibility could be predicted by seven speech fluency indicators (i.e., repetition, omission, distortion, correction, unfilled pauses, filled pauses, and speaking rate). Speech data were retrieved from a database containing a reading task produced by a group of 16 English-speaking individuals with PD (Jaeger, Trivedi & Stadtchnitzer, 2019). The results from a multiple regression indicated that an addition of 54% of variance in the speech intelligibility scores among individuals with PD could be accounted for after the speakers' PD severity level measured based on Hoehn and Yahr's (1967) disease stage was included as a covariate. In addition, omission and correction were the two fluency indicators that contributed to the general intelligibility score in a statistically significant way. Specifically, for every one-unit gain in the number of correction and omission, speech intelligibility scores would decline by 0.687 and 0.131 point (out of a 7-point scale), respectively. The current study hence supported Magee, Copland, and Vogel's (2019) view that the language production abilities and quantified dysarthria measures among individuals with PD should be explored together. Additionally, the clinical implications based on the current findings were discussed.

Use of the double-paddle anterolateral thigh flap for locally advanced tongue carcinoma requiring second site reconstruction

BackgroundAdvanced oral tongue carcinoma can present with extension beyond the oral cavity. Operative defects after resection may involve multiple anatomical sites and significantly impact speech and swallowing. Dependence on long-term enteral feeding is not uncommon for these patients. The anterolateral thigh (ALT) flap is one of the most reliable and flexible flaps used in the reconstruction of total and subtotal tongue defects. The double-paddle flap modification may be a more suitable option for complex oral tongue defects after advanced tumor ablation. MethodsCase series of 31 patients with oral tongue squamous cell carcinoma that were classified as stage IV. The age of patients ranged from 32 to 63 years. We designed the double-paddle ALT flaps to reconstruct the two-site surgical defects (tongue defect and pharynx or neck skin defect). Postoperative viability of the flap was checked by clinical observation. The last examination was performed at 3-months after the completion of adjuvant chemoradiotherapy. The functional capacity of our patients was evaluated by three physicians (Head and Neck Surgeon, Radiation Oncologist, and Physiatrist) using a Speech Intelligibility Score and the Functional Oral Intake Scale. ResultsA total of 31 patients with surgical defects after total or subtotal tongue resection for cancer underwent double-paddle ALT flaps for reconstruction from March 2018 to December 2019. The dimension of flaps from 8 × 12 cm to 10 × 18 cm were divided into double-paddle from 8 × 5 cm to 10 × 10 cm. There was one case of pedicle thrombosis, one case of postoperative bleeding, three cases of neck infection, and six cases of salivary fistula. Our patients were seen in follow up from 6 to 36 months, with median follow-up of 23.5 months. The survival rate of ALT flap was 100%. All of our patients achieved an oral diet by 9 months after surgery. The mean score speech intelligibility was 2.74 ± 0.68 (4-point ordinal scale). The 2-year disease-free survival rate was 61.3%. ConclusionsThe double-paddle ALT flap is a reliable flap suitable for oral defects involving multiple subsites after ablative procedures. The majority of patients demonstrated acceptable functional rehabilitation. Clinical question/ level of evidenceTherapeutic, IV.

Objective speech intelligibility prediction using a deep learning model with continuous speech-evoked cortical auditory responses.

Auditory prostheses provide an opportunity for rehabilitation of hearing-impaired patients. Speech intelligibility can be used to estimate the extent to which the auditory prosthesis improves the user’s speech comprehension. Although behavior-based speech intelligibility is the gold standard, precise evaluation is limited due to its subjectiveness. Here, we used a convolutional neural network to predict speech intelligibility from electroencephalography (EEG). Sixty-four–channel EEGs were recorded from 87 adult participants with normal hearing. Sentences spectrally degraded by a 2-, 3-, 4-, 5-, and 8-channel vocoder were used to set relatively low speech intelligibility conditions. A Korean sentence recognition test was used. The speech intelligibility scores were divided into 41 discrete levels ranging from 0 to 100%, with a step of 2.5%. Three scores, namely 30.0, 37.5, and 40.0%, were not collected. The speech features, i.e., the speech temporal envelope (ENV) and phoneme (PH) onset, were used to extract continuous-speech EEGs for speech intelligibility prediction. The deep learning model was trained by a dataset of event-related potentials (ERP), correlation coefficients between the ERPs and ENVs, between the ERPs and PH onset, or between ERPs and the product of the multiplication of PH and ENV (PHENV). The speech intelligibility prediction accuracies were 97.33% (ERP), 99.42% (ENV), 99.55% (PH), and 99.91% (PHENV). The models were interpreted using the occlusion sensitivity approach. While the ENV models’ informative electrodes were located in the occipital area, the informative electrodes of the phoneme models, i.e., PH and PHENV, were based on the occlusion sensitivity map located in the language processing area. Of the models tested, the PHENV model obtained the best speech intelligibility prediction accuracy. This model may promote clinical prediction of speech intelligibility with a comfort speech intelligibility test.