In Silico World releases its first data collection: HFValid


HFValid collection: Hip-Fracture validation collection” is a dataset composed of 101 anonymised calibrated computed tomography scans (CT scans) of whole femurs, the corresponding segmentations, and selected anatomical landmarks. It has been released by a research team of the University of Bologna and IRCCS Istituto Ortopedico Rizzoli, and it’s the first data collection produced within the In Silico World project.

Through the linked third party Istituto Ortopedico Rizzoli (IOR), Dr Alessandra Aldieri, Dr Antonino Amedeo La Mattina, Dr Julia Aleksandra Szyszko, Dr Fabio Baruffaldi and Prof Marco Viceconti from partner UNIBO had access to CT scans of the thigh region, originally collected to provide CT-based surgical planning of a total hip replacement procedure, but under an informed consent that also allowed secondary use for research.

Some subjects experienced a proximal femur fracture within the following 5 years from the initial scan, so the CT scan cohort was used as a validation dataset for prospective patient-specific femur fracture risk estimation models. IOR and the regional ethical committee have granted permission to publicly share a subgroup of the CT scans after full and irreversible anonymisation.

The CT scans were asynchronously calibrated using a set of European Spine Phantom (ESP) CT scans collected at IOR, choosing the phantom with the most similar acquisition parameters for each subject.

A Finite Elements model (FE) of each femur was developed. Starting from the femur geometry, a 10-nodes tetrahedral mesh (2 mm edge size, Octree algorithm) was obtained with Ansys ICEM CFD, and material properties were assigned elementwise with Bonemat (developed by Bioengineering and Computing Laboratory of IOR).

Each FE model was simulated 28 times with different load and boundary condition orientations to explore a wide range of possible femur impact conditions; for each simulation, the critical load that would cause a fracture was calculated. The first estimator of femur fracture risk is the Minimum Sidefall Strength (MSF), defined as the minimum critical load over the 28 orientations.

The FE model was also coupled with a multiscale probabilistic falling model taking into account different patient fall and impact positions and coping mechanisms to estimate the Absolute Risk of Fracture at time 0 (ARF0). HFValid dataset also includes MSF and ARF0 results of the subject-specific FE simulations; together with the MSF, the corresponding intra-extra rotation and add-abduction angles of the thigh at impact are also reported.

We need widely available collections of curated medical data specifically designed for developing and validating in silico trial technologies. Our goal is to define requirements for data used as inputs of In Silico Trials models, analyse contexts of use and define requirements for validation collections, organise collections of existing data and support the formation of prospective collections. This is only the first step in that direction.

Access the full dataset here