Formus Hip

Formus Hip is a Software as a Medical Device that facilitates pre-operative planning for total hip arthroplasty by using AI-driven image segmentation and statistical shape modeling from CT scans to generate patient-specific 3D anatomical models and propose implant configurations. Non-clinical validation demonstrated high accuracy in 3D model generation compared to manual segmentation and strong concordance with surgeon-selected implant sizes.

Formus Hip is a semi-automated Software as a Medical Device (SaMD) designed for pre-operative planning of primary total hip arthroplasty, specifically utilizing the Zimmer Biomet Taperloc G7 system. The software imports patient pre-dimensioned CT scans and employs AI-based automatic image segmentation algorithms, trained on diverse patient data (ages 21-94), to create a 3D model of the patient's femur and pelvis without initial user input. It further uses statistical shape models, trained on subjects aged 18-89, and algorithms to fit the femoral stem and acetabular cup based on the patient's anatomy. The system includes an integrated database of orthopedic hip implant geometries for overlay and allows healthcare professionals to interactively customize the surgical plan via a graphical user interface, with finalized plans exportable as PDF reports. The device's non-clinical performance was validated for image processing accuracy and implant sizing. Image processing accuracy was assessed using 60 US patient CT scans, comparing Formus Hip's 3D models to manual segmentations by radiologists. Results met acceptance criteria, with average Sorensen-Dice scores of 0.95 (hemipelvis) and 0.97 (femur), and mean absolute distances (MAD) below 2mm. Statistical shape modeling accuracy also met targets. Implant sizing validation, conducted on 133 Australian patient images, compared software-recommended sizes to those chosen by orthopedic surgeons using traditional methods. The software achieved 94% agreement for stem sizes and 98% for cup sizes within ±2 sizes of the surgeons' choices, meeting the performance goal of at least 80%.

AI Regulatory Concerns

Powered by

Jev reads the full FDA letter and returns calibrated concern probabilities per regulatory category. These are AI estimates, not regulatory advice.

• Q1. How does the AI-based automatic image segmentation algorithm handle CT scans with significant artifacts or unusual anatomical variations not well-represented in the training dataset (ages 21-94 for segmentation, 18-89 for shape models)?; What is the specific performance of the statistical shape models in cases of extreme anatomical deformities, such as severe dysplasia or post-traumatic changes, compared to typical anatomy within the training data parameters?; Given that implant sizing validation was performed on Australian patient data, what specific measures ensure its generalizability and accuracy across diverse global populations and varying surgical preferences?; What is the typical processing time for Formus Hip to generate a complete pre-operative plan from a patient's CT scan, and how does this integrate into existing clinical workflows?; Beyond the Zimmer Biomet Taperloc G7 system, what is the protocol and anticipated timeline for incorporating and validating additional hip implant systems from other manufacturers into the software's database?