Table 2 Overview of neural function-based studies for CI outcome prediction
From: A systematic review of machine learning approaches in cochlear implant outcomes
Study | Data Characteristics | Outcome Measure and Model Accuracy | Comment |
---|---|---|---|
Children | |||
Lu et al.23 | Stability selection method to select robust features from averaged audiological and radiographic data; Children with cochlear nerve deficiency with anatomically normal cochlea = 70 mean age of 27.31 months (SD = 13.92 months) | Categories of auditory performance, speech intelligibility rating, infant/toddler meaningful auditory integration scale; SVM: Accuracy Hearing rehabilitation = 71% Speech rehabilitation = 93% | Nerve bundle volume and a larger vestibulocochlear nerve area correlated with better CI outcomes. |
Adults | |||
Kyong et al.22 | EEG measures (cAEPs), cSEPs and cVEPs); Adults with CIs= 3 | Latency, amplitude, current source density; SVM model Accuracy: auditory: 82.71%, tactile: 98.88%, visual: 93.5% | Tactile response was a key indicator of speech recognition ability in individuals with CIs. |
Children and Adults | |||
Skidmore et al.24 | eCAP parameters (N1latency, threshold, slope); Children with cochlear nerve deficiency = 23 mean age= 3.42 years, children with normal-sized cochlear nerves = 29 mean age=3.18 years, adults with normal-sized cochlear nerves = 20 mean age= 69.22 years | Cochlear nerve index; Accuracy Linear regression: 91%, SVM: 93%, and logistic regression: 95% | Models predicted two distinct cochlear nerve index distributions for cochlear nerve deficiency and normal-sized cochlear nerves. In adult CI users, cochlear nerve function correlated with speech perception outcomes. |