Classification of Voice Disorders Using a One-Dimensional Convolutional Neural Network.

OBJECTIVES: Auditory-perceptual voice analysis is a standard method for quantifying pathological voice quality, but perceptual ratings are based on subjective evaluations and therefore may vary among examiners. Although many acoustic metrics have been studied for potential use in the objective evaluation of pathological voices, the interpretation of acoustic metrics in individual cases is difficult and the technique is not widely used by clinicians. The aim of this study was to establish standardized methods to discriminate grade, roughness, breathiness, asthenia, strain (GRBAS) scale scores of pathological voices directly using one-dimensional convolutional neural network (1D-CNN) models. METHODS: We constructed an original dataset utilizing 1,377 voice samples of sustained phonation of the vowel /a/. Each voice sample was rated by three experts according to the GRBAS scale and the median values were used as the correct answer label. We designed an end-to-end 1D-CNN model with a raw voice waveform input having a frame width of 9,600 samples. The models were trained with our original dataset for each GRBAS category individually and the model performance was tested by the five-fold cross validation method. RESULTS: The accuracy, F1 score, and quadratic weighted Cohen's kappa for the testing dataset were determined. The metrics for the G scale showed the most balanced model performance, with high accuracy (0.771) and substantial agreement (kappa = 0.710). The model for the R scale had relatively high accuracy (0.765) and F1 score (0.743) with moderate agreement (kappa = 0.536). The accuracy (0.883) and the F1 score (0.865) for the S scale were the highest among the five categories, whereas the Cohen's kappa was the lowest (0.190). CONCLUSIONS: The end-to-end 1D-CNN models can evaluate overall pathological voice quality with a reliability comparable to human evaluations. The efficiency with which the machine learning models can be trained and evaluated is closely related to the dataset quality.

Objective Assessment of Pathological Voice Using Artificial Intelligence Based on the GRBAS Scale.

OBJECTIVES: The validity and reliability of the psychological assessment of auditory perceptions, as typified by the grade, roughness, breathiness, asthenia, and strain (GRBAS) scale, have been widely recognized. However, due to their subjective nature, inter- and intra-examiner reliability are unavoidable. In this study, we aimed to add objectivity to the GRBAS scale using artificial intelligence and to compare the accuracy of two methods-one based on Google's TensorFlow and another based on Apple's Core ML. METHODS: The GRBAS scale of 1,377 vowel samples was evaluated and used as training data to create a machine learning model. We used TensorFlow and Apple's Create ML to create two machine learning models and examined the difference in their accuracies for classifying the severity of pathological Voice data based on the GRBAS scale. RESULTS: Absolute comparisons are difficult to make because of the difference in methods; however, both training models could objectively evaluate GRBAS scales and were statistically correlated in G and B. CONCLUSION: While TensorFlow requires creation of a training model from scratch, Create ML is a relatively easy way to create a training model for voice by adding training data for GRBAS scales to an existing training model for sounds. Although the data handling and learning methods are different, both models performed well. Findings from this study could be used for medical screening purposes, and there is the potential to change the clinical approach to voice diagnostics in the future.

Successful treatment of eosinophilic chronic rhinosinusitis and eosinophilic otitis media using the anti-IL-5 receptor monoclonal antibody benralizumab: A case report.

Eosinophilic chronic rhinosinusitis (ECRS) is characterized by the presence of nasal polyps, dominant ethmoid shadows in computed tomography (CT) scans, and elevated levels of eosinophil infiltration into the nasal polyps and peripheral blood. ECRS is often accompanied by severe asthma. The recent development of monoclonal antibody-based biologics, including benralizumab, has offered new therapeutic approaches for the treatment of asthma and allergic diseases. Asthma and ECRS are closely related; hence, benralizumab could provide clinical benefit in ECRS patients with severe asthma. Herein, we report a case of a 47-year-old female patient with severe asthma that presented with nasal obstruction and hearing impairment. Nasal endoscopic and otoscopic examinations indicated the presence of bilateral nasal polyps in the middle nasal meatus, as well as a bilateral effusion in the tympanic cavity. Sinus and temporal CT images showed dominant ethmoid sinus and tympanic cavity shadows. Biopsy of nasal polyps revealed high numbers of eosinophils, which led to the diagnosis of ECRS; eosinophilic otitis media (EOM) with hypereosinophilia was also suspected. Treatment with benralizumab reduced the number of peripheral blood eosinophils and improved asthma symptoms. Prolonged benralizumab administration also resulted in a remarkable size reduction in bilateral middle nasal polyps and aeration of the tympanic cavity. In conclusion, benralizumab treatment improved the symptoms of severe asthma, ECRS, and EOM. Eosinophil depletion could be an important mechanism by which benralizumab improves ECRS and EOM. The use of benralizumab for the treatment of ECRS and EOM patients with severe asthma merits further investigation in large-cohort studies.

Real-Time Acoustic Voice Analysis Using a Handheld Device Running Android Operating System.

OBJECTIVES: We developed a highly accessible acoustic voice analysis system (VArt) using a handheld device running Android operating system. To provide stable and reliable analysis using readily obtainable equipment under unfavorable conditions, we modified the fundamental frequency (F0) extraction algorithm and designed an intuitive user interface representing a new hoarseness index (real-time Ra: Rart), which is a derivative of harmonics-to-noise ratio developed by Kojima and Shoji (Ra2). Since Rart continues to display analysis results in real time, unlike conventional acoustic analysis, it can be used for evaluation such as during phonosurgery and speech therapy. We evaluated the agreement between the earlier version of acoustic voice analysis software (VA) running on a Windows personal computer and VArt running on a handheld device. METHODS: F0, Ra2, and Rart were measured in voice samples of sustained vowel phonation /a/ from 10 healthy volunteers and 22 patients with voice disorders using VA running on a Windows personal computer and VArt running on two types of handheld devices in a sound-treated room or in a medical examination room. Intraclass correlation coefficients were calculated for both systems under both conditions. RESULTS: All of the comparisons were highly correlated. CONCLUSIONS: Measurements obtained using our newly developed VArt were highly consistent with those using VA, indicating high reliability. Moreover, the new system increases the clinical feasibility of acoustic voice analysis.

Adrenal insufficiency following nivolumab therapy in patients with recurrent or metastatic head and neck cancer.

Nivolumab, an anti-programmed cell death-1 monoclonal antibody, is currently used to treat many types of advanced cancers including recurrent and metastatic head and neck cancer. However, there are increasing reports concerning immune-related adverse events related to nivolumab therapy. Here, we report three patients who presented with adrenal insufficiency following nivolumab therapy. Two were diagnosed as having isolated adrenocorticotropic hormone (ACTH) deficiency and one was diagnosed as having primary adrenal insufficiency. All three patients complained of progressive fatigue and appetite loss, so we measured their blood cortisol and ACTH levels and diagnosed them as having adrenal deficiency. Treatment with nivolumab was discontinued for all three patients, and replacement therapy using hydrocortisone was successful after a few days in all cases. Two patients subsequently resumed nivolumab therapy because their general condition had improved. Complaints of fatigue and appetite loss during cancer treatment are common and tend to be regarded as unimportant. Although adrenal insufficiency due to nivolumab is relatively rare, complaints of these symptoms could lead to the detection of adrenal insufficiency at an early stage. The present report highlights the importance of the early recognition of adrenal insufficiency.

An Innovative Voice Analyzer "VA" Smart Phone Program for Quantitative Analysis of Voice Quality.

OBJECTIVE: The 'VA' Windows program that we developed in 2011 for analyzing voice quality quantitatively uses zerocross picking to find individual basic pitch periods. It has a simple and user-friendly user interface and high accuracy. This program determines the fundamental frequency, jitter, shimmer, PPQ, APQ, and signal-to-noise ratio (Ra). It needs only a general-purpose Windows PC, USB audio interface and a microphone. The aim of this study is to improve the version of the VA Windows program in English and to develop a VA smart phone program to allow wider use of objective acoustic analysis. STUDY DESIGN: Cross-sectional study. METHODS: Sustained vowel /a/ sounds from 40 subjects without evident vocal problems, and 40 subjects with slight hoarseness, were examined. We compared the analyzed data with data from other software (MDVP and Praat). For a comparison between VA for Windows and VA for a smart phone, sustained vowel /a/ sounds from six subjects without hoarseness were recorded with each system simultaneously. RESULTS: The normal voice and slightly hoarse voice data analyzed with VA showed a high correlation with most parameters from both MDVP and Praat. There was a strong correlation between the Windows and smart phone versions of VA in terms of the fundamental frequency and Ra. CONCLUSIONS: The results showed that the VA software was not inferior to the other acoustic analysis software tested. The simple and easy to use smart phone version may facilitate our goal of creating an objective, widely available method to evaluate hoarseness.

Discrimination of "hot potato voice" caused by upper airway obstruction utilizing a support vector machine.

OBJECTIVES/HYPOTHESIS: "Hot potato voice" (HPV) is a thick, muffled voice caused by pharyngeal or laryngeal diseases characterized by severe upper airway obstruction, including acute epiglottitis and peritonsillitis. To develop a method for determining upper-airway emergency based on this important vocal feature, we investigated the acoustic characteristics of HPV using a physical, articulatory speech synthesis model. The results of the simulation were then applied to design a computerized recognition framework using a mel-frequency cepstral coefficient domain support vector machine (SVM). STUDY DESIGN: Quasi-experimental research design. METHODS: Changes in the voice spectral envelope caused by upper airway obstructions were analyzed using a hybrid time-frequency model of articulatory speech synthesis. We evaluated variations in the formant structure and thresholds of critical vocal tract area functions that triggered HPV. The SVMs were trained using a dataset of 2,200 synthetic voice samples generated by an articulatory synthesizer. Voice classification experiments on test datasets of real patient voices were then performed. RESULTS: On phonation of the Japanese vowel /e/, the frequency of the second formant fell and coalesced with that of the first formant as the area function of the oropharynx decreased. Changes in higher-order formants varied according to constriction location. The highest accuracy afforded by the SVM classifier trained with synthetic data was 88.3%. CONCLUSIONS: HPV caused by upper airway obstruction has a highly characteristic spectral envelope. Based on this distinctive voice feature, our SVM classifier, who was trained using synthetic data, was able to diagnose upper-airway obstructions with a high degree of accuracy. LEVEL OF EVIDENCE: 2c Laryngoscope, 129:1301-1307, 2019.

[Parotidectomy for parotid gland tumor using a retrograde approach from the peripheral branches of the facial nerve].

There are two techniques for excising a parotid gland tumor. The first technique is the anterograde approach, and the second is the retrograde approach. Although the anterograde approach has generally been adopted, the retrograde approach may be useful depending on the identification and dissection of the facial nerve and the localization of the tumor. We reviewed 35 patients who were surgically treated with a parotidectomy using the retrograde approach between 2002 and 2012 at Tenri Hospital. We also reviewed the clinical records for evidence of temporary or definitive injury to the facial nerve. In cases where the identification of the main nerve trunk was expected to be difficult, the parotidectomy was performed using the retrograde approach guided by the mandibular branch. In cases where a tumor was located in the marginal region of the parotid gland, the retrograde approach was guided by the mandibular branch or the zygomatic branch. Of the 35 patients, 17 underwent the retrograde approach guided by the mandibular branch and 18 underwent this approach guided by the zygomatic branch. The mean surgical time and amount of bleeding using the retrograde approach were 70.7 minutes and 33.9 g, respectively. Two of the 35 patients in the retrograde group exhibited a temporary facial deficit.

Four-dimensional computed tomographic angiography of an arteriovenous fistula in the neck.

Arteriovenous fistula (AVF), a type of vascular anomaly, is a relatively uncommon disease characterized by abnormal connections between the arterial and venous systems. Recently, remarkable advances in four-dimensional computed tomographic angiography (4D-CTA) have enabled detailed analysis of the hemodynamic features of vascular anomalies, which cannot be achieved by conventional imaging modalities. We report a case of AVF in a 59-year-old female presenting as a right neck mass. Using reconstructed animations from various viewing angles, 4D-CTA could clearly visualize not only a mass but the feeding artery and draining vein from the mass. 4D-CTA was useful for the diagnosis of AVF and to determine the surgical strategy preoperatively.

Hyperfractionated radiotherapy for T2 glottic cancer for preservation of the larynx.

To evaluate the effectiveness of hyperfractionation for T2 glottic cancer from a viewpoint of laryngeal preservation, we analyzed 21 patients (twice-a-day group) who were treated with hyperfractionation between 1992 and 1998 and compared the results with those of 27 patients (once-a-day group) treated with conventional once-a-day radiation between 1987 and 1992. In the twice-a-day group, radiation was performed with two fractions of 1.2 Gy/day up to a total dose of 72-74.4 Gy. In the once-a-day group, radiation was performed with a fraction of 2 Gy/day up to a total dose of 66 Gy. If radiation was ineffective at 40 Gy, it was stopped, and surgical treatment was carried out. Kaplan-Meier estimates were used for the analysis of the survival rate and laryngeal preservation rate, and the results were compared. In the once-a-day group, the 5-year survival rate was 92.3%. The 5-year laryngeal preservation rate was 51.8%, and it was 60.3% in 20 patients who had undergone full-dose radiation (once-a-day full-dose group). In the twice-a-day group, no major complication, such as laryngeal necrosis, was seen in any case, and the 5-year survival rate was 95.3%. The 5-year laryngeal preservation rate was 95.3%, and it was significantly better than that of both the once-a-day group and the once-a-day full-dose group. Hyperfractionation is considered to be useful for preserving the larynx for the treatment of T2 glottic cancer.

Concurrent chemotherapy and radiotherapy with low-dose Cisplatin for nasopharyngeal carcinoma.

Eighty-one patients with nondisseminated nasopharyngeal carcinoma consecutively treated between January 1977 and December 1998 were analyzed to evaluate whether a concurrent adjunction of low-dose cisplatin enhances the outcome of definitive radiotherapy. Ninety-eight percent (n = 79) of the cases were ranked as stage III/IV according to the 1987 Union International Contre le Cancer staging criteria. Patients treated before 1987 and treated after 1988 were mainly managed by radiotherapy alone (historical group: n = 48) and concurrent chemoradiotherapy with relatively low-dose cisplatin (CCRT group: n = 33), respectively. The locoregional failure-free survival rate of the CCRT group was significantly better than that of the historical group (72.8% vs. 35.9% at 5 years, p = 0.0041). However, multivariate analysis identified only the total dose and the T-stage as significant independent factors for locoregional control. No difference was observed on overall, disease-specific, and distant failure-free survival between the two groups. The results of the present study suggest that concurrent adjunction of low-dose cisplatin will not improve the outcome of definitive radiotherapy for nasopharyngeal carcinoma. Full-dose concurrent chemoradiotherapy, as well as the appropriate dose escalation for better locoregional control, will be mandatory to achieve better survival.