Trends in Distributions of Hearing Threshold Levels by Ages: A Comparison of the ISO 7029 and Newly Available Country-Specific Data.

Hearing thresholds provide essential information and references about the human auditory system. This study aimed to identify changing trends in distributions of hearing threshold levels across ages by comparing the International Organization for Standardization (ISO) 7029 and newly available data after publishing ISO 7029. To compare ISO 7029 and newly available hearing threshold data after publishing ISO 7029, four country-specific datasets that presented average hearing threshold levels under conditions similar to ISO 7029 were utilized. For frequencies between 125 Hz and 8,000 Hz, the deviations of hearing threshold values by ages from the hearing threshold of the youngest age group for each data point were utilized. For frequencies from 9,000 Hz to 12,500 Hz, the median threshold information was utilized. Hearing threshold data reported after publishing ISO 7029 from the four countries were mostly similar to the ISO 7029 data but tended to deviate in some age groups and sexes. As national hearing threshold trends change, the following ISO 7029 revision suggests the need to integrate hearing threshold data from different countries.

Improving Accuracy and Reliability of Hearing Tests: An Exploration of International Standards.

This study explores the internal standards for hearing tests and benefits of implementing international standard protocols, including the International Organization for Standardization (ISO) and International Electrotechnical Commission (IEC), and discusses how ISO and IEC standards provide a framework for designing, calibrating, assessing hearing test instruments and methods, and exchanging and comparing data globally. ISO and IEC standards for hearing tests improve accuracy, reliability, and consistency of test results by applying standardized methods and environments. Moreover, they promote international harmonization and data interoperability, enabling information exchange and research collaboration. Those standards for hearing tests are beneficial but have challenges and limitations, such as variation in equipment and calibration, lag in updating standards, variation in implementation and compliance, and lack of coverage of clinical aspects, cultural diversity, and linguistic diversity. These affect the quality and interpretation of test results. Adapting ISO or IEC standards locally would improve their applicability and acceptability, while balancing customization and compatibility with global standards.

Human-machine cooperation meta-model for clinical diagnosis by adaptation to human expert's diagnostic characteristics.

Artificial intelligence (AI) using deep learning approaches the capabilities of human experts in medical image diagnosis. However, due to liability issues in medical decisions, AI is often relegated to an assistant role. Based on this responsibility constraint, the effective use of AI to assist human intelligence in real-world clinics remains a challenge. Given the significant inter-individual variations in clinical decisions among physicians based on their expertise, AI needs to adapt to individual experts, complementing weaknesses and enhancing strengths. For this adaptation, AI should not only acquire domain knowledge but also understand the specific human experts it assists. This study introduces a meta-model for human-machine cooperation that first evaluates each expert's class-specific diagnostic tendencies using conditional probability, based on which the meta-model adjusts the AI's predictions. This meta-model was applied to ear disease diagnosis using otoendoscopy, highlighting improved performance when incorporating individual diagnostic characteristics, even with limited evaluation data. The highest accuracy was achieved by combining each expert's conditional probabilities with machine classification probability, using optimal weights specific to each individual's overall classification accuracy. This tailored model aims to mitigate potential misjudgments due to psychological effects caused by machine suggestions and to capitalize on the unique expertise of individual clinicians.

Differential Biases and Variabilities of Deep Learning-Based Artificial Intelligence and Human Experts in Clinical Diagnosis: Retrospective Cohort and Survey Study.

BACKGROUND: Deep learning (DL)-based artificial intelligence may have different diagnostic characteristics than human experts in medical diagnosis. As a data-driven knowledge system, heterogeneous population incidence in the clinical world is considered to cause more bias to DL than clinicians. Conversely, by experiencing limited numbers of cases, human experts may exhibit large interindividual variability. Thus, understanding how the 2 groups classify given data differently is an essential step for the cooperative usage of DL in clinical application. OBJECTIVE: This study aimed to evaluate and compare the differential effects of clinical experience in otoendoscopic image diagnosis in both computers and physicians exemplified by the class imbalance problem and guide clinicians when utilizing decision support systems. METHODS: We used digital otoendoscopic images of patients who visited the outpatient clinic in the Department of Otorhinolaryngology at Severance Hospital, Seoul, South Korea, from January 2013 to June 2019, for a total of 22,707 otoendoscopic images. We excluded similar images, and 7500 otoendoscopic images were selected for labeling. We built a DL-based image classification model to classify the given image into 6 disease categories. Two test sets of 300 images were populated: balanced and imbalanced test sets. We included 14 clinicians (otolaryngologists and nonotolaryngology specialists including general practitioners) and 13 DL-based models. We used accuracy (overall and per-class) and kappa statistics to compare the results of individual physicians and the ML models. RESULTS: Our ML models had consistently high accuracies (balanced test set: mean 77.14%, SD 1.83%; imbalanced test set: mean 82.03%, SD 3.06%), equivalent to those of otolaryngologists (balanced: mean 71.17%, SD 3.37%; imbalanced: mean 72.84%, SD 6.41%) and far better than those of nonotolaryngologists (balanced: mean 45.63%, SD 7.89%; imbalanced: mean 44.08%, SD 15.83%). However, ML models suffered from class imbalance problems (balanced test set: mean 77.14%, SD 1.83%; imbalanced test set: mean 82.03%, SD 3.06%). This was mitigated by data augmentation, particularly for low incidence classes, but rare disease classes still had low per-class accuracies. Human physicians, despite being less affected by prevalence, showed high interphysician variability (ML models: kappa=0.83, SD 0.02; otolaryngologists: kappa=0.60, SD 0.07). CONCLUSIONS: Even though ML models deliver excellent performance in classifying ear disease, physicians and ML models have their own strengths. ML models have consistent and high accuracy while considering only the given image and show bias toward prevalence, whereas human physicians have varying performance but do not show bias toward prevalence and may also consider extra information that is not images. To deliver the best patient care in the shortage of otolaryngologists, our ML model can serve a cooperative role for clinicians with diverse expertise, as long as it is kept in mind that models consider only images and could be biased toward prevalent diseases even after data augmentation.

Local Differential Privacy in the Medical Domain to Protect Sensitive Information: Algorithm Development and Real-World Validation.

BACKGROUND: Privacy is of increasing interest in the present big data era, particularly the privacy of medical data. Specifically, differential privacy has emerged as the standard method for preservation of privacy during data analysis and publishing. OBJECTIVE: Using machine learning techniques, we applied differential privacy to medical data with diverse parameters and checked the feasibility of our algorithms with synthetic data as well as the balance between data privacy and utility. METHODS: All data were normalized to a range between -1 and 1, and the bounded Laplacian method was applied to prevent the generation of out-of-bound values after applying the differential privacy algorithm. To preserve the cardinality of the categorical variables, we performed postprocessing via discretization. The algorithm was evaluated using both synthetic and real-world data (from the eICU Collaborative Research Database). We evaluated the difference between the original data and the perturbated data using misclassification rates and the mean squared error for categorical data and continuous data, respectively. Further, we compared the performance of classification models that predict in-hospital mortality using real-world data. RESULTS: The misclassification rate of categorical variables ranged between 0.49 and 0.85 when the value of ε was 0.1, and it converged to 0 as ε increased. When ε was between 102 and 103, the misclassification rate rapidly dropped to 0. Similarly, the mean squared error of the continuous variables decreased as ε increased. The performance of the model developed from perturbed data converged to that of the model developed from original data as ε increased. In particular, the accuracy of a random forest model developed from the original data was 0.801, and this value ranged from 0.757 to 0.81 when ε was 10-1 and 104, respectively. CONCLUSIONS: We applied local differential privacy to medical domain data, which are diverse and high dimensional. Higher noise may offer enhanced privacy, but it simultaneously hinders utility. We should choose an appropriate degree of noise for data perturbation to balance privacy and utility depending on specific situations.

Implementing Vertical Federated Learning Using Autoencoders: Practical Application, Generalizability, and Utility Study.

BACKGROUND: Machine learning (ML) is now widely deployed in our everyday lives. Building robust ML models requires a massive amount of data for training. Traditional ML algorithms require training data centralization, which raises privacy and data governance issues. Federated learning (FL) is an approach to overcome this issue. We focused on applying FL on vertically partitioned data, in which an individual's record is scattered among different sites. OBJECTIVE: The aim of this study was to perform FL on vertically partitioned data to achieve performance comparable to that of centralized models without exposing the raw data. METHODS: We used three different datasets (Adult income, Schwannoma, and eICU datasets) and vertically divided each dataset into different pieces. Following the vertical division of data, overcomplete autoencoder-based model training was performed for each site. Following training, each site's data were transformed into latent data, which were aggregated for training. A tabular neural network model with categorical embedding was used for training. A centrally based model was used as a baseline model, which was compared to that of FL in terms of accuracy and area under the receiver operating characteristic curve (AUROC). RESULTS: The autoencoder-based network successfully transformed the original data into latent representations with no domain knowledge applied. These altered data were different from the original data in terms of the feature space and data distributions, indicating appropriate data security. The loss of performance was minimal when using an overcomplete autoencoder; accuracy loss was 1.2%, 8.89%, and 1.23%, and AUROC loss was 1.1%, 0%, and 1.12% in the Adult income, Schwannoma, and eICU dataset, respectively. CONCLUSIONS: We proposed an autoencoder-based ML model for vertically incomplete data. Since our model is based on unsupervised learning, no domain-specific knowledge is required in individual sites. Under the circumstances where direct data sharing is not available, our approach may be a practical solution enabling both data protection and building a robust model.

Feasibility of Asynchronous and Automated Telemedicine in Otolaryngology: Prospective Cross-Sectional Study.

BACKGROUND: COVID-19 often causes respiratory symptoms, making otolaryngology offices one of the most susceptible places for community transmission of the virus. Thus, telemedicine may benefit both patients and physicians. OBJECTIVE: This study aims to explore the feasibility of telemedicine for the diagnosis of all otologic disease types. METHODS: A total of 177 patients were prospectively enrolled, and the patient's clinical manifestations with otoendoscopic images were written in the electrical medical records. Asynchronous diagnoses were made for each patient to assess Top-1 and Top-2 accuracy, and we selected 20 cases to conduct a survey among four different otolaryngologists to assess the accuracy, interrater agreement, and diagnostic speed. We also constructed an experimental automated diagnosis system and assessed Top-1 accuracy and diagnostic speed. RESULTS: Asynchronous diagnosis showed Top-1 and Top-2 accuracies of 77.40% and 86.44%, respectively. In the selected 20 cases, the Top-2 accuracy of the four otolaryngologists was on average 91.25% (SD 7.50%), with an almost perfect agreement between them (Cohen kappa=0.91). The automated diagnostic model system showed 69.50% Top-1 accuracy. Otolaryngologists could diagnose an average of 1.55 (SD 0.48) patients per minute, while the machine learning model was capable of diagnosing on average 667.90 (SD 8.3) patients per minute. CONCLUSIONS: Asynchronous telemedicine in otology is feasible owing to the reasonable Top-2 accuracy when assessed by experienced otolaryngologists. Moreover, enhanced diagnostic speed while sustaining the accuracy shows the possibility of optimizing medical resources to provide expertise in areas short of physicians.

Machine learning approach for prediction of hearing preservation in vestibular schwannoma surgery.

In vestibular schwannoma patients with functional hearing status, surgical resection while preserving the hearing is feasible. Hearing levels, tumor size, and location of the tumor have been known to be candidates of predictors. We used a machine learning approach to predict hearing outcomes in vestibular schwannoma patients who underwent hearing preservation surgery: middle cranial fossa, or retrosigmoid approach. After reviewing the medical records of 52 patients with a pathologically confirmed vestibular schwannoma, we included 50 patient's records in the study. Hearing preservation was regarded as positive if the postoperative hearing was within serviceable hearing (50/50 rule). The categorical variable included the surgical approach, and the continuous variable covered audiometric and vestibular function tests, and the largest diameter of the tumor. Four different algorithms were lined up for comparison of accuracy: support vector machine(SVM), gradient boosting machine(GBM), deep neural network(DNN), and diffuse random forest(DRF). The average accuracy of predicting hearing preservation ranged from 62% (SVM) to 90% (DNN). The current study is the first to incorporate machine learning methodology into a prediction of successful hearing preservation surgery. Although a larger population may be needed for better generalization, this study could aid the surgeon's decision to perform a hearing preservation approach for vestibular schwannoma surgery.

Automated diagnosis of ear disease using ensemble deep learning with a big otoendoscopy image database.

BACKGROUND: Ear and mastoid disease can easily be treated by early detection and appropriate medical care. However, short of specialists and relatively low diagnostic accuracy calls for a new way of diagnostic strategy, in which deep learning may play a significant role. The current study presents a machine learning model to automatically diagnose ear disease using a large database of otoendoscopic images acquired in the clinical environment. METHODS: Total 10,544 otoendoscopic images were used to train nine public convolution-based deep neural networks to classify eardrum and external auditory canal features into six categories of ear diseases, covering most ear diseases (Normal, Attic retraction, Tympanic perforation, Otitis externa±myringitis, Tumor). After evaluating several optimization schemes, two best-performing models were selected to compose an ensemble classifier, by combining classification scores of each classifier. FINDINGS: According to accuracy and training time, transfer learning models based on Inception-V3 and ResNet101 were chosen and the ensemble classifier using the two models yielded a significant improvement over each model, the accuracy of which is in average 93·67% for the 5-folds cross-validation. Considering substantial data-size dependency of classifier performance in the transfer learning, evaluated in this study, the high accuracy in the current model is attributable to the large database. INTERPRETATION: The current study is unprecedented in terms of both disease diversity and diagnostic accuracy, which is compatible or even better than an average otolaryngologist. The classifier was trained with data in a various acquisition condition, which is suitable for the practical environment. This study shows the usefulness of utilizing a deep learning model in the early detection and treatment of ear disease in the clinical situation. FUND: This research was supported by Brain Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science and ICT(NRF-2017M3C7A1049051).

The Feasibility of a Modified Exclusive Endoscopic Transcanal Transpromontorial Approach for Vestibular Schwannomas.

Objective We evaluated the feasibility of an exclusive endoscopic transcanal transpromontorial approach (EETTA) for the treatment of small vestibular schwannomas (VSs) limited to the internal auditory canal (IAC), and introduced a modification without external auditory canal closure. Methods Between June 2016 and June 2017, seven patients with VS underwent surgery using a modified EETTA. Treatment outcomes, including efficacy of tumor resection, preservation of function, operation time, and quality of life (QOL), were evaluated. Results The patients preoperatively exhibited Koos Grade I/II tumors and severe-to-profound hearing loss. Gross total resection was accomplished in all cases. There were no major complications, and all patients exhibited normal facial nerve function immediately after surgery. The mean follow-up period was 12.9 months. The operation time (average 196.3 ± 64.9 minutes) and hospitalization period (average 7.4 ± 1.0 days) were favorable. Short Form-36 scores for QOL showed unremarkable results compared with previous reports. Conclusions The modified EETTA was effective in the removal of VSs in the IAC. It can be an alternative surgical option for small VSs.

Transoral Robotic Surgery With Transoral Retropharyngeal Lymph Node Dissection in Patients With Tonsillar Cancer: Anatomical Points, Surgical Techniques, and Clinical Usefulness.

OBJECTIVE: The purpose of this study was to evaluate the clinical usefulness of TORS and transoral robotic retropharyngeal lymph node (RPLN) dissection in tonsillar cancer patients with suspicious RPLN metastasis. METHODS: From April 2008 to March 2014, 71 patients with tonsillar cancer underwent transoral robotic surgery and standard neck dissection at the Yonsei Head and Neck Cancer Center. RESULTS: Three patients underwent transoral robotic ropharyngectomy with transoral robotic RPLN dissection because of suspicious RPLN metastasis. The mean age of the patients was 42 years (range, 31-50 years). There were no cases of wound infection or serious complications related to wound healing. Mild nasal regurgitation was observed during an oral diet immediately after surgery, but all patients spontaneously resolved without surgical treatment. There was no significant bleeding due to great vessel injury during surgery or swallowing difficulty due to cranial nerve IX injury. CONCLUSION: Although the oncologic stability and usefulness of this technique should be confirmed based on large-scale research, RPLN can be easily accessed and resected through our approach with less morbidity compared to the conventional surgical approach. In addition, because RPLN metastasis can be performed pathologically based on obtained specimens, it will be helpful to explore whether to perform adjuvant radiation.

The long-term oncological and functional outcomes of transoral robotic surgery in patients with hypopharyngeal cancer.

OBJECTIVE: We conducted a prospective clinical trial of transoral robotic surgery in patients with hypopharyngeal cancer and herein report the long-term oncological and functional outcomes. MATERIALS AND METHODS: Between April 2008 and March 2014, 45 patients diagnosed with hypopharyngeal cancer participated in this prospective study. RESULTS: All patients were male with a mean age of 66.7years. The median follow-up period was 60months. Patients were classified using the staging system of the American Joint Commission on Cancer, as follows: Stage I, 7.9%; Stage II, 5.3%; Stage III, 15.8%; Stage IV, 71.1. Of all 38 patients, 17 (44.7%) were alive with no evidence of disease at the last follow-up. Seven patients (18.4%) died of TNM-related disease and fourteen (36.8%) from other causes. The 5-year disease-specific survival rate of stage I and II patients was 100.0%, and that of stage III and IV patients was 74.0%. The 5-year disease-free survival rate was 100.0% for stage I and II patients and 68.6% for stage III and IV patients. CONCLUSIONS: Patients who underwent TORS exhibited oncological outcomes comparable to those of conventional therapies and rapid functional recovery with low surgical morbidity. TORS and simultaneous neck dissection, with or without adjuvant therapy, may be effective alternatives to existing treatment methods.

A New Clinical Trial of Neoadjuvant Chemotherapy Combined With Transoral Robotic Surgery and Customized Adjuvant Therapy for Patients With T3 or T4 Oropharyngeal Cancer.

BACKGROUND: A prospective clinical trial of combination neoadjuvant chemotherapy, transoral robotic surgery (TORS), and customized adjuvant therapy for patients with locally advanced oropharyngeal cancer was conducted. METHODS: Between July 2009 and October 2016, 31 patients were enrolled in this clinical trial. RESULTS: The primary lesions were located in the tonsils of 27 patients and in the base of the tongue of 4 patients. Of the 31 patients, 16 (51.6%) were classified as T3 and 15 patients (48.4%) as T4a. Three patients (9.7%) had stage 3 disease, and 28 (90.3%) had stage 4 disease. The 5-year overall survival rate was 78.7%; the 5-year disease-specific survival rate was 85%; and the 5-year disease-free survival rate was 80.8%. At the final follow-up visit, 26 patients were alive with no evidence of disease, and 1 was alive with disease. Four patients died during the study: two of tumor-node-metastasis (TNM)-related disease and two of another condition. All the patients tolerated an oral diet at an average of 7.4 days postoperatively. At the subjective swallowing evaluation using the Functional Outcome Swallowing Scale score, 83.9% of the patients exhibited favorable outcomes. No patient was permanently dependent on a feeding tube. All the patients breathed and phonated in the absence of a permanent tracheotomy at the final follow-up evaluation. CONCLUSIONS: The treatment strategy in this study afforded good oncologic and functional outcomes for patients with locally advanced oropharyngeal cancer. Although future large-scale multicenter studies with longer follow-up periods are needed, this study showed that neoadjuvant chemotherapy combined with TORS is useful for treating advanced oropharyngeal cancer.