Machine Learning Models for Predicting Hearing Prognosis in Unilateral Idiopathic Sudden Sensorineural Hearing Loss

Abstract

Objectives. Prognosticating idiopathic sudden sensorineural hearing loss (ISSNHL) is an important challenge. In our study, a dataset was split into training and test sets and cross-validation was implemented on the training set, thereby determining the hyperparameters for machine learning models with high test accuracy and low bias. The effectiveness of the following five machine learning models for predicting the hearing prognosis in patients with ISSNHL after 1 month of treatment was assessed: adaptive boosting, K-nearest neighbor, multilayer perceptron, random forest (RF), and support vector machine (SVM).Methods. The medical records of 523 patients with ISSNHL admitted to Korea University Ansan Hospital between January 2010 and October 2017 were retrospectively reviewed. In this study, we analyzed data from 227 patients (recovery, 106; no recovery, 121) after excluding those with missing data. To determine risk factors, statistical hypothesis tests (e.g., the two-sample t-test for continuous variables and the chi-square test for categorical variables) were conducted to compare patients who did or did not recover. Variables were selected using an RF model depending on two criteria (mean decreases in the Gini index and accuracy).Results. The SVM model using selected predictors achieved both the highest accuracy (75.36%) and the highest F-score (0.74) on the test set. The RF model with selected variables demonstrated the second-highest accuracy (73.91%) and F-score (0.74). The RF model with the original variables showed the same accuracy (73.91%) as that of the RF model with selected variables, but a lower F-score (0.73). All the tested models, except RF, demonstrated better performance after variable selection based on RF.Conclusion. The SVM model with selected predictors was the best-performing of the tested prediction models. The RF model with selected predictors was the second-best model. Therefore, machine learning models can be used to predict hearing recovery in patients with ISSNHL.

Biochemical and Pain Comparisons Between the Laser Lancing Device and Needle Lancets for Capillary Blood Sampling: A Randomized Control Trial

“…Blood sampling with less pain can be useful clinically [25][26][27][28][29][30]. In addition to self-measuring blood glucose in patients with diabetes mellitus, it can be used for gas analysis in the intensive care unit and biochemical component analysis.…”

Contributions and limitations of using machine learning to predict noise-induced hearing loss

“…For instance, two papers suggested that either RF (Statnikov et al 2008 ) or SVM (Statnikov and Aliferis 2007 ) could outperform any other in classification accuracy to diagnose and predict a similar clinical problem. Similarly, Bing et al ( 2018 ) found that deep belief network reached highest performance to predict SSNHL measured by several metrics, whereas SVM was the best classifier compared to predict unilateral SSNHL in another study (Park et al 2020 ).…”

“…Another study constructed a large database ( n = 2,420,330) to analyzed the impact of diverse noise to the generation of NIHL using ANN but unraveled the unsatisfactory performance with less than 65% accuracy, which was no better than LR model (Kim et al 2011 ). The accuracies of some algorithms were also investigated in several studies which either tried to predict hearing loss with specific etiologies, such as sudden hearing loss (Bing et al 2018 ; Park et al 2020 ), ototoxic hearing loss (Tomiazzi et al 2019 ) and cochlear dead regions (Chang et al 2019 ), or predict SNHL by specific auditory measures, such as OAE (de Waal et al 2002 ; Liu et al 2020 ; Ziavra et al 2004 ) and ABR (Acır et al 2006 ; Molina et al 2016 ). Similarly, five studies did not evaluate or describe the significance of input to cochlear dead regions (Chang et al 2019 ; de Waal et al 2002 ; Liu et al 2020 ; Tomiazzi et al 2019 ; Ziavra et al 2004 ).…”

“…In addition, different studies also applied more than one model to predict other types of hearing loss or hearing-related problems. Three types of algorithms achieved good performance in predicting sudden hearing loss (Bing et al 2018 ; Park et al 2020 ), ototoxic hearing loss (Tomiazzi et al 2019 ) and/or SNHL with specific risk factors (Chang et al 2019 ). Also, several studies successfully applied ANN or SVM to clarify different types of SNHL based on the morphology of OAE (de Waal et al 2002 ; Liu et al 2020 ; Ziavra et al 2004 ) or ABR (Acır et al 2006 ).…”

Machine Learning Models for Predicting Facial Nerve Palsy in Parotid Gland Surgery for Benign Tumors

“…These have focused on predicting survival in individuals with oral cavity squamous cell carcinoma, 25 improving prognostic predictions in patients with well-differentiated thyroid cancer, 26 predicting delays in adjuvant radiation in those undergoing surgery for head and neck cancer, 27 analyzing imaging data to create models that predict treatment outcomes in patients with sinonasal squamous cell carcinoma, 28 looking for better models to predict occult nodal metastasis in T1-T2 clinically N0 oral cavity squamous cell carcinoma, 29 and predicting hearing prognosis in sudden sensorineural hearing loss. 30 However, to translate findings from ML to the clinical environment, we need to understand the differences between ML algorithms and the classical statistical models. First, ML focuses on how all the variables interrelate, using this information to make predictions about an unknown variable.…”