|
Á¢¼ö¹øÈ£ - 270035 4 |
HEARING RECOVERY PREDICTION AND PROGNOSTIC FACTORS OF IDIOPATHIC SUDDEN
SENSORINEURAL HEARING LOSS USING A DEEP NEURAL NETWORK MODEL |
CDEPARTMENT OF OTORHINOLARYNGOLOGY-HEAD AND NECK SURGERY, PUSAN NATIONAL UNIVERSITY COLLEGE OF MEDICINE, PUSAN NATIONAL UNIVERSITY YANGSAN HOSPITAL©ö, DEPARTMENT OF STATISTICS, PUKYONG NATIONAL UNIVERSITY©÷ |
HYUN MIN LEE©ö,TAE WOONG UHM©÷,HYO BEOM JANG©ö, DA HEE PARK©ö, HO BYUNG LEE©ö, SANG HYO LEE©ö, IL WOO LEE©ö |
¸ñÀû: Idiopathic sudden sensorineural hearing loss (ISSHL) is an otologic
emergency. An early prediction of prognosis may facilitate proper
treatment. Therefore, we investigated its prognostic factors for
predicting a recovery in patients with ISSHL using machine learning
models ¹æ¹ý:We retrospectively reviewed the medical records of 433 patients with
ISSHL who were treated at a tertiary medical institution between January
2015 and September 2020. A total of 52 variables were analyzed to
predict hearing recovery. Recovery was defined using Siegel¡¯s criteria,
and the patients were categorized into recovery and non-recovery groups.
Recovery was predicted by various machine learning models. In addition,
the prognostic factors were analyzed using the difference in the loss
function. °á°ú:The comparison of recovery and non-recovery groups revealed
significant differences in variables, including age, hypertension,
history of cardiac disease and previous hearing loss, ear fullness,
dizziness, duration of hospital admission, delay between symptom onset
and treatment, initial hearing level of the affected and unaffected
ears, post-treatment hearing level, and treatment methods. The deep
neural network (DNN) model showed the highest predictive performance.
(accuracy: 89.66%, area under the receiver operating characteristic
curve: 0.9646) As a result of analyzing the prognostic, it was
confirmed that the history of hypertension and previous hearing loss
and dizziness were important in predicting the prognosis. °á·Ð:The DNN model showed the highest predictive performance for recovery in
patients with ISSHL. The factors with prognostic value were also
identified. Further studies using a larger patient population are
warranted. |
|