Development of Insertion Models Predicting Cochlear Implant Electrode Position.

Abstract

To assess the possibility to define a preferable range for electrode array insertion depth and surgical insertion distance for which frequency mismatch is minimalized. To develop a surgical insertion guidance tool by which a preferred target angle can be attained using preoperative available anatomical data and surgically controllable insertion distance. Multiplanar reconstructions of pre- and post-operative CT scans were evaluated in a population of 336 patients implanted with the CII HiFocus1 or HiFocus1J implant (26 bilaterally implantees included). Cochlear radial distances were measured on four measurement axes on the preoperative CT scan. Electrode contact positions were obtained in angular depth, distance from the round window and to the modiolus center. Frequency mismatch was calculated based on the yielded frequency as a function of the angular position per contact. Cochlear diameters were clustered into three cochlear size groups with K-sample clustering. Using spiral fitting and general linear regression modeling, the feasibility of different insertion models with cochlear size measures and surgical insertion as input parameters was analyzed. The final developed model was internally validated with bootstrapping to calculate the optimism-corrected R. Frequency mismatch was minimalized for surgical insertion of 6.7 mm and insertion depth of 484°. Cochlear size clusters were derived consisting of a "small" (N = 117), "medium" (N = 171), and "large" (N = 74) cluster with mean insertion depths of 506°, 480°, and 441°, respectively. The relation between surgical insertion (LE16) and insertion depth (θE1) differed significantly between the three clusters (p < 0.01). The insertion models based on spiral fitting showed an R of 62% with mean of the residuals of -0.5 mm (SD = 1.2 mm) between the measured and predicted LE16 and a mean of 15° (SD = 83°) for θE1. Using general linear regression modeling resulted in a residual mean of -0.2 μm (SD = 0.9 mm) for measured and predicted LE16 and 0.01° (SD = 33°) for θE1. The model derived from general linear regression modeling resulted in an R of 78.7% and was validated with bootstrapping. An optimism of 0.6% was calculated using this analysis. The optimism-corrected R of 78.1% defined the estimated performance of the final insertion model in future populations. A minimal frequency mismatch for an electrode array design can be calculated to define preferable electrode array position within the cochlea. In general, to achieve a minimal frequency mismatch, the surgeon should attempt to insert the HiFocus 1 or 1J array around 6, 7, or 8 mm in case of a "small," "medium," or "large" cochlea, respectively. Development of different insertion models showed the feasibility of obtaining a surgical guidance tool to lead the surgeon during cochlear implantation depending on individual cochlear size and controllable surgical distance. The developed final insertion model predicted 78.1% of the variation in final HiFocus1 or HiFocus1J implant position.