Abstract:Studies using quantitative computed tomography (QCT) and data-driven image analysis techniques have shown that trabecular and cortical volumetric bone mineral density (vBMD) can improve the hip fracture prediction of dual-energy X-ray absorptiometry areal BMD (aBMD). Here, we hypothesize that (1) QCT imaging features of shape, density and structure derived from data-driven image analysis techniques can improve the hip fracture discrimination of classification models based on mean femoral neck aBMD (Neck.aBMD), and (2) that data-driven cortical bone thickness (Ct.Th) features can improve the hip fracture discrimination of vBMD models. We tested our hypotheses using statistical multi-parametric modeling (SMPM) in a QCT study of acute hip fracture of 50 controls and 93 fragility fracture cases. SMPM was used to extract features of shape, vBMD, Ct.Th, cortical vBMD, and vBMD in a layer adjacent to the endosteal surface to develop hip fracture classification models with machine learning logistic LASSO. The performance of these classification models was evaluated in two aspects: (1) their hip fracture classification capability without Neck.aBMD, and (2) their capability to improve the hip fracture classification of the Neck.aBMD model. Assessments were done with 10-fold cross-validation, areas under the receiver operating characteristic curve (AUCs), differences of AUCs, and the integrated discrimination improvement (IDI) index. All LASSO models including SMPM-vBMD features, and the majority of models including SMPM-Ct.Th features performed significantly better than the Neck.aBMD model; and all SMPM features significantly improved the hip fracture discrimination of the Neck.aBMD model (Hypothesis 1). An interesting finding was that SMPM-features of vBMD also captured Ct.Th patterns, potentially explaining the superior classification performance of models based on SMPM-vBMD features (Hypothesis 2). Age, height and weight had a small impact on model performances, and the model of shape, vBMD and Ct.Th consistently yielded better performances than the Neck.aBMD models. Results of this study clearly support the relevance of bone density and quality on the assessment of hip fracture, and demonstrate their potential on patient and healthcare cost benefits.

Machine learning based finite element analysis (FEA) surrogate for hip fracture risk assessment and visualization

Hip Fracture Risk Assessment in Elderly and Diabetic Patients: Combining Autonomous Finite Element Analysis and Machine Learning

Implementing Machine Learning approaches for accelerated prediction of bone strain in acetabulum of a hip joint

In-Vivo Assessment of Femoral Bone Strength Using Finite Element Analysis (FEA) Based on Routine MDCT Imaging: A Preliminary Study on Patients with Vertebral Fractures

A Staged Approach using Machine Learning and Uncertainty Quantification to Predict the Risk of Hip Fracture

Multi-view information fusion using multi-view variational autoencoder to predict proximal femoral fracture load

A new hip fracture risk index derived from FEA-computed proximal femur fracture loads and energies-to-failure

Machine Learning to Predict Long and Short Term Fracture Risk in Postmenopausal Women

Explainable machine-learning-based prediction of QCT/FEA-calculated femoral strength under stance loading configuration using radiomics features

Fully automated CT imaging biomarkers for opportunistic prediction of future hip fractures

A Novel Fracture Prediction Model Using Machine Learning in a Community‐Based Cohort

3D Analysis of the Proximal Femur Compared to 2D Analysis for Hip Fracture Risk Prediction in a Clinical Population

Improving the Hip Fracture Risk Prediction Through 2D Finite Element Models From DXA Images: Validation Against 3D Models

Opportunistic application of phantom-less calibration methods for fracture risk prediction using QCT/FEA

Predicting fracture risk for elderly osteoporosis patients by hybrid machine learning model

A Computed Tomography–Based Fracture Prediction Model With Images of Vertebral Bones and Muscles by Employing Deep Learning: Development and Validation Study

Introducing Anisotropic Minkowski Functionals for Local Structure Analysis and Prediction of Biomechanical Strength of Proximal Femur Specimens

Hip Fracture Discrimination Based on Statistical Multi-parametric Modeling (SMPM).

The influence of femoral lytic tumors segmentation on autonomous finite element analysis

Machine learning‐based mortality prediction in hip fracture patients using biomarkers

Predicting Fracture Risk for Chinese Elderly Osteoporosis Patients by Hybrid Model