Accuracy of computer-aided geometric 3D reconstruction based on histological serial microgrinding preparation

Abstract

Purpose: For our research on computer-optimised and automated cochlear implant surgery, we pursue a model-based approach to overcome the limitations of currently available clinical imaging modalities. A serial cross section preparation procedure has been developed and evaluated concerning accuracy to serve for modelling of a digital anatomic atlas to make delicate soft tissue structures available for pre-operative planning. Methods: A special grinding tool was developed allowing the setting of a specific amount of abrasion as equidistant slice thickness was considered a crucial step. Additionally, each actual abrasion was accurately measured and used during three-dimensional reconstruction of the serial cross-sectional images obtained via digital photo documentation after each microgrinding step. A well-known reference object was prepared using this procedure and evaluated in terms of accuracy. Results: Reconstruction of the whole sample was achieved with an error less than 0.4%, and the edge lengths in the direction of abrasion could be reconstructed with an average error of 0.6 ± 0.3 mm; both prove the realisation of equidistant abrasion. Using artificial registration fiducials and a custom-made algorithm for image alignment, parallelism and rectangularity could be preserved with average errors less than 0.4° ± 0.3°. Conclusion: We present a systematic, practicable and reliable method for the geometrically accurate reconstruction of anatomical structures, which is especially suitable for the middle and inner ear anatomy including soft tissue structures. For the first time, the quality of such a reconstruction process has been quantified and successfully proven for its usability.