A novel segmentation of cochlear nerve using region growing algorithm

- This method involves segmentation and measurement of the cochlear nerve.
- Automatic seed point selection based on statistics of histogram is used in region growing algorithm.
- Automatic threshold selection based on kernel density estimation of the image is employed in proposed algorithm.
- Segmentation accuracy of Otsu threshold and proposed region growing segmentation are compared in this method.
- The manual and automated measurements are compared using correlation and Bland-Altman plot.

Sensorineural hearing loss is a hearing impairment which occurs when there is damage to the inner ear, or to the nerve pathways from the inner ear to the brain. Cochlear implants have been developed to benefit children with bilateral or unilateral Sensorineural hearing loss. A very small or absence of cochlear nerve precludes successful outcome of cochlear implant surgery. Hence, segmentation and measurement of the cochlear nerve support the surgeon's decision to predict a normal or poor outcome of the cochlear implant. For this purpose, a modified region growing segmentation algorithm is proposed that segments the cochlear nerve region accurately. The segmentation accuracy is evaluated using parameters like Jaccard, Dice, False Positive Dice, and False Negative Dice. The segmented region is measured and evaluated using long diameter, short diameter, and cross-sectional area. The statistical analyses of intra/inter-observer correlation and limits of agreement are performed on a cross-sectional area of the cochlear nerve to investigate the reproducibility of the automated measurement.