Precision of computed tomography and cartilage-reproducing image reconstruction method in generating digital model for potential use in 3D printing of patient- specific radial head prosthesis: a human cadaver study

A prosthetic replacement is a standard treatment for an irreparable radial head fracture; however, the surface mismatch of the commercially available designs is concerned for the long-term cartilage wear. The patient- specific implant created from 3D printing technology could be favorable in replicating the normal anatomy and possibly reduce such sequela. Our study aimed to assess the precision of the computed tomography (CT) and cartilage-reproducing image reconstruction method (CIRM) in generating digital models for potentially use in manufacturing the patient-specific prosthesis from 3D printing. Methods: Eight intact elbows (3 right and 5 left) from 7 formalin-embalmed cadavers (4 males and 3 females) with mean age of 83 years (range, 79–94 years) were used for this study. Computerized 3D models were generated from CT, and CIRM. The cartilage-reproducing image reconstruction method has compensated the cartilage profile based on the distance between the subchondral surfaces of the radial head and surrounding bones in CT images. The models of actual radial head geometry used as the gold standard was generated from CT arthrography (CTA). All models of each specimen were matched by registering the surface area of radial neck along with the tuberosity. The difference of head diameter, head thickness, and articular disc depth among three models was evaluated and analyzed by Friedman ANOVA and multiple comparison test using Bonferroni method for statistical correction. A p- value of less than 0.01 was considered statistically significant. The difference of overall 3D geometry was measured with the root mean square of adjacent point pairs.The analysis displayed the difference of diameter, thickness, and disc depth across the models (p< 0.01). Pairwise comparisons revealed statistically significant difference of all parameters between CTA models and CT models (p< 0.01) whereas no difference was found between CTA models and CIRM models. The mean difference of overall 3D geometry between CTA models and CT models was 0.51±0.24 mm, and between CTA models and CIRM models was 0.24±0.10 mm. Conclusions: CIRM demonstrated encouraging results in reestablish the normal anatomy and could be potentially used in production process of 3D printed patient-specific radial head prosthesis.