Study of the significance of parameters and their interaction on assessing femoral fracture risk by quantitative statistical analysis

Risk analysis for subsequent fracture of osteoporotic fractures in Chinese women over age 60: a nationwide cross-sectional study

Prevention of subsequent fracture is a major public health challenge in the field of osteoporosis prevention and treatment, and older women are at high risk for osteoporotic fractures. This study aimed to examine factors associated with subsequent fracture in older Chinese women with osteoporosis. We collected data on 9212 older female patients with osteoporotic fractures from 580 medical institutions in 31 provinces of China. Higher odds of subsequent fractures were associated with age of 70-79 years (OR 1.218, 95% CI 1.049-1.414), age ≥ 80 (OR 1.455, 95% CI 1.222-1.732), index fracture site was vertebrae (OR 1.472, 95% CI 1.194-1.815) and hip (OR 1.286, 95% CI 1.041-1.590), index fracture caused by fall (OR 1.822, 95% CI 1.281-2.591), strain (OR 1.587, 95% CI 1.178-2.139), no inducement (OR 1.541, 95% CI 1.043-2.277), and assessed as high risk of fracture (OR 1.865, 95% CI 1.439-2.416), BMD T-score ≤ -2.5 (OR 1.725, 95% CI 1.440-2.067), history of surgery (OR 3.941, 95% CI 3.475-4.471) and trauma (OR 8.075, 95% CI 6.941-9.395). Low risk of fall (OR 0.681, 95% CI 0.513-0.904), use of anti-osteoporosis medication (AOM, OR 0.801, 95% CI 0.693-0.926), and women who had received fall prevention health education (OR 0.583, 95% CI 0.465-0.730) associated with lower risk. The areas under the curve of the prediction model was 0.818. The sensitivity was 67.0% and the specificity was 82.0%. The prediction model showed a good ability to predict the risk of subsequent fracture in older women with osteoporotic fractures and are suitable for early self-measurement which may benefit post-fracture management.

SSDL-an automated semi-supervised deep learning approach for patient-specific 3D reconstruction of proximal femur from QCT images

Deep Learning (DL) techniques have recently been used in medical image segmentation and the reconstruction of 3D anatomies of a human body. In this work, we propose a semi-supervised DL (SSDL) approach utilizing a CNN-based 3D U-Net model for femur segmentation from sparsely annotated quantitative computed tomography (QCT) slices. Specifically, QCT slices at the proximal end of the femur forming ball and socket joint with acetabulum were annotated for precise segmentation, where a segmenting binary mask was generated using a 3D U-Net model to segment the femur accurately. A total of 5474 QCT slices were considered for training among which 2316 slices were annotated. 3D femurs were further reconstructed from segmented slices employing polynomial spline interpolation. Both qualitative and quantitative performance of segmentation and 3D reconstruction were satisfactory with more than 90% accuracy achieved for all of the standard performance metrics considered. The spatial overlap index and reproducibility validation metric for segmentation-Dice Similarity Coefficient was 91.8% for unseen patients and 99.2% for validated patients. An average relative error of 12.02% and 10.75% for volume and surface area, respectively, were computed for 3D reconstructed femurs. The proposed approach demonstrates its effectiveness in accurately segmenting and reconstructing 3D femur from QCT slices.

QCT-based 3D finite element modeling to assess patient-specific hip fracture risk and risk factors

Early assessment of hip fracture risk may play a critical role in designing preventive mechanisms to reduce the occurrence of hip fracture in geriatric people. The loading direction, clinical, and morphological variables play a vital role in hip fracture. Analyzing the effects of these variables helps predict fractures risk more accurately; thereby suggesting the critical variable that needs to be considered. Hence, this work considered the fall postures by varying the loading direction on the coronal plane (α) and on the transverse plane (β) along with the clinical variables-age, sex, weight, and bone mineral density, and morphological variables-femoral neck axis length, femoral neck width, femoral neck angle, and true moment arm. The strain distribution obtained via finite element analysis (FEA) shows that the angle of adduction (α) during a fall increases the risk of fracture at the greater trochanter and femoral neck, whereas with an increased angle of rotation (β) during the fall, the FRI increases by ∼1.35 folds. The statistical analysis of clinical, morphological, and loading variables (αandβ) delineates that the consideration of only one variable is not enough to realistically predict the possibility of fracture as the correlation between individual variables and FRI is less than 0.1, even though they are shown to be significant (p<0.01). On the contrary, the correlation (R2=0.48) increases as all variables are considered, suggesting the need for considering different variables fork predicting FRI. However, the effect of each variable is different. While loading, clinical, and morphological variables are considered together, the loading direction on transverse plane (β) has high significance, and the anatomical variabilities have no significance.