Iowa CI Trauma Tool: A Deep Learning Approach to Cochlear Implant Trauma Assessment

Introduction: Intracochlear cochlear implant (CI) position impacts post-operative hearing outcomes. Currently, tools for CI trauma assessment have limitations including significant user input, limited to proprietary electrode array, and lack of distance metrics between electrode and intracochlear structures. Our aim was to create an automated, open-source tool that segments the CI and cochlea within post-operative CT scans (poCT), and overlays labels for the scala-tympani (ST), scala vestibuli/media (SVM), lateral wall (LW), modiolus, and basilar membrane (BM) to compute CI position relative to these intracochlear structures.

Methods: The tool utilizes two pipelines: 1) Cochlea and CI segmentations derived from poCT; 2) Intracochlear segmentations derived from pre-operative CT. Pipeline 1 was created by training a neural network (nnUNet) on 50 labeled poCTs from CI patients implanted between 2022-24. Pipeline 2 was developed using five co-registered 3D-xray microscopy and CT pairs of the cochlea from cadaveric temporal bones, augmented synthetically to train a separate nnUNet. The result of pipeline 2 is registered to pipeline 1 for CI position analysis. For both pipelines, using cross validation, 20% of the training data were randomly isolated to create a testing dataset.

Results: Accuracy was assessed on the testing dataset using Dice similarity coefficients (DSC), which measures overlap between predicted and ground-truth labels (1 = perfect overlap). Pipeline 1 achieved DSCs of 0.94 (cochlea) and 0.87 (electrode). Pipeline 2 had DSC scores of 0.96, 0.95, 0.83, 0.86, 0.91 for ST, SVM, BM, modiolus, and LW, respectively. On a subset of LW CIs (N=6), mean electrode distances to BM, LW, and modiolus were 0.50 mm, 0.61 mm, and 1.26 mm, respectively.

Conclusions: Using neural networks and image registration techniques, we created an accurate, open-source CI trauma assessment tool that is manufacturer and electrode-agnostic. This platform enables CI position analysis, using clinically available CT technology, to aid in surgical planning and postoperative management and programming.