Articular surface segmentation using active shape models for intraoperative implant assessment

Area of Science:

• Orthopedic Surgery
• Medical Imaging
• Computational Anatomy

Background:

• Intra-articular implant placement in orthopedic surgery is linked to higher revision rates and arthritis risk.
• Accurate intraoperative assessment is crucial for preventing such complications.

Purpose of the Study:

• To develop and validate an automatic detection approach for intra-articular implants using cone-beam computed tomography (CBCT).
• To support intraoperative assessment and correction of surgical implant placement.

Main Methods:

• Utilized multiple active shape models (ASM) to segment joint spaces and articular surfaces.
• Integrated fast and robust methods within ASM for intraoperative accuracy.
• Employed cylinder detection to identify metal implants and their intersections with articular surfaces to detect intra-articular collisions.

Main Results:

• Achieved average surface distance errors of 0.59 mm and 0.46 mm for calcaneal articular surface segmentations on 50 patient images.
• Demonstrated a 22% and 25% average error reduction using model-independent segmentation.
• Localized articular perforations in cadaver tests with an average accuracy of 0.80 mm, compensating for significant translation and rotation errors.

Conclusions:

• Presented a novel concept for automatic intraoperative detection of intra-articular implants in CBCT images.
• Validated reliable segmentation of articular surfaces in patient data.
• Confirmed accurate localization of misplaced implants in cadaver tests.