Detection of Inner Ear Malformations Based on Simple Anatomical Measurements: A Model Approach

Area of Science:

• Otolaryngology
• Medical Imaging
• Anatomy

Background:

• Accurate identification of inner ear malformations is crucial for surgical planning and optimizing hearing outcomes.
• Quantifiable measurements of cochlear morphology, particularly the basal turn, may help differentiate normal from malformed variants.

Purpose of the Study:

• To explore the feasibility of using basal turn morphology principles to differentiate normal cochlear lateral wall geometry from malformed variants.
• To assess the accuracy of simple anatomical measurements in identifying and classifying cochlear malformations.

Main Methods:

• Retrospective analysis of 111 cochleae with inner ear malformations and 141 normal cochleae from cone beam computed tomography (CBCT) scans.
• Manual segmentation to obtain 3D lateral wall spirals and compute basal turn parameters: cochlear diameter (A), width (B), basal turn length, B/A ratio, B-ratio, and height (H).
• Statistical analysis using Mann-Whitney-Wilcoxon tests and logistic regression to evaluate the predictive capability of geometrical parameters for differentiating normal, malformed, and specific malformation types.

Main Results:

• Four basal turn parameters (basal turn length, cochlear height (H), B/A ratio, B-ratio) reliably distinguished normal from malformed cochleae.
• A binary logistic model achieved 94% accuracy, with cochlear height (H) as the dominant predictor.
• A three-class model (normal, incomplete partition, cochlear hypoplasia) retained 90% accuracy, effectively differentiating subtypes like CHIII but showing limitations for IPI vs. IPII and rare subtypes.

Conclusions:

• Simple, basal turn measurements offer a fast, interpretable, and accurate method for detecting cochlear malformations on clinical imaging.
• This approach provides valuable preoperative information for patient-tailored cochlear implantation and highlights areas for future research, such as incorporating additional features for rare subtypes.
• The developed model shows potential for automated recognition of cochlear malformations in preoperative planning.