In conventional DXA (Dual-energy X-ray Absorptiometry) analysis, pixel bone mineral density (BMD) is often averaged at the femoral neck. Neck BMD constitutes the basis for osteoporosis diagnosis and fracture risk assessment. This data averaging, however, limits our understanding of localised spatial BMD patterns that could potentially enhance fracture prediction. DXA region free analysis (RFA) is a validated toolkit for pixel-level BMD analysis. We have previously deployed this toolkit to develop a spatio-temporal atlas of BMD ageing in the femur. This study aims first to introduce bone age to reflect the overall bone structural evolution with ageing, and second to quantify fracture-specific patterns in the femur. The study dataset comprised 4933 femoral DXA scans from White British women aged 75 years or older. The total number of fractures was 684, of which 178 were reported at the hip within a follow-up period of five years. BMD maps were computed using the RFA toolkit. For each BMD map, bone age was defined as the age for which the L2-norm between the map and the median atlas at that age is minimised. Next, bone maps were normalised for the estimated bone age. A t-test followed by false discovery rate (FDR) analysis was applied to compare between fracture and non-fracture groups. Excluding the ageing effect revealed subtle localised patterns of loss in BMD oriented in the same direction as principal tensile curves. A new score called f-score was defined by averaging the normalised pixel BMD values over the region with FDR q-value less than 1e–6. The area under the curve (AUC) was 0.731 (95% confidence interval (CI)=0.689–0.761) and 0.736 (95% CI=0.694–0.769) for neck BMD and f-score. Combining bone age and f-score improved the AUC significantly by 3% (AUC=0.761, 95% CI=0.756–0.768) over the neck BMD alone (AUC=0.731, 95% CI=0.726–0.737). This technique shows promise in characterizing spatially-complex BMD changes, for which the conventional region-based technique is insensitive. DXA RFA shows promise to further improve fracture prediction using spatial BMD distribution

Introduction. Up to 60% of total hip arthroplasties (THA) in Asian populations arise from avascular necrosis (AVN), a bone disease that can lead to femoral head collapse. Current diagnostic methods to classify AVN have poor reproducibility and are not reliable in assessing the fracture risk. Femoral heads with an immediate fracture risk should be treated with a THA, conservative treatments are only successful in some cases and cause unnecessary patient suffering if used inappropriately. There is potential to improve the assessment of the fracture risk by using a combination of density-calibrated computed tomographic (QCT) imaging and engineering beam theory. The aim of this study was to validate the novel fracture prediction method against in-vitro compression tests on a series of six human femur specimens. Methods. Six femoral heads from six subjects were tested, a subset (n=3) included a hole drilled into the subchondral area of the femoral head via the femoral neck (University of Leeds, ethical approval MEEC13-002). The simulated lesions provided a method to validate the fracture prediction model with respect of AVN. The femoral heads were then modelled by a beam loaded with a single joint contact load. Material properties were assigned to the beam model from QCT-scans by using a density-modulus relationship. The maximum joint loading at which each bone cross-section was likely to fracture was calculated using a strain based failure criterion. Based on the predicted fracture loads, all six femoral heads (validation set) were classified into two groups, high fracture risk and low fracture risk (Figure 1). Beam theory did not allow for an accurate fracture load to be found because of the geometry of the femoral head. Therefore the predicted fracture loads of each of the six femoral heads was compared to the mean fracture load from twelve previously analysed human femoral heads (reference set) without lesions. The six cemented femurs were compression tested until failure. The subjects with a higher fracture risk were identified using both the experimental and beam tool outputs. Results. The computational tool correctly identified all femoral head samples which fractured at a significantly low load in-vitro (Figure 2). Both samples with a low experimental fracture load had an induced lesion in the subchondral area (Figure 3). Discussion. This study confirmed findings of a previous verification study on a disease models made from porcine femoral heads (Preutenborbeck et al. I-CORS2016). It demonstrated that fracture prediction based on beam theory is a viable tool to predict fracture. The tests confirmed that samples with a lesion in the weight bearing area were more likely to fracture at a low load however not all samples with a lesion fractured with a low load experimentally, indicating that a lesion alone is not a sufficient factor to predict fracture. The developed tool takes both structural and material properties into account when predicting the fracture risk. Therefore it might be superior to current diagnostic methods in this respect and it has the added advantage of being largely automated and therefore removing the majority of user bias. For any figures or tables, please contact the authors directly

To explore a novel machine learning model to evaluate the vertebral fracture risk using Decision Tree model and train the model by Bone Mineral Density (BMD) of different compartments of vertebral body. We collected a Computed Tomography image dataset, including 10 patients with osteoporotic fracture and 10 patients without osteoporotic fracture. 40 non-fracture Vertebral bodies from T11 to L5 were segmented from 10 patients with osteoporotic fracture in the CT database and 53 non-fracture Vertebral bodies from T11 to L5 were segmented from 10 patients without osteoporotic fracture in the CT database. Based on the biomechanical properties, 93 vertebral bodies were further segmented into 11 compartments: eight trabecular bone, cortical shell, top and bottom endplate. BMD of these 11 compartments was calculated based on the HU value in CT images. Decision tree model was used to build fracture prediction model, and Support Vector Machine was built as a compared model. All BMD data was shuffled to a random order. 70% of data was used as training data, and 30% left was used as test data. Then, training prediction accuracy and testing prediction accuracy were calculated separately in the two models. The training accuracy of Decision Tree model is 100% and testing accuracy is 92.14% after trained by BMD data of 11 compartments of the vertebral body. The type I error is 7.14% and type II error is 0%. The training accuracy of Support Vector Machine model is 100% and the testing accuracy is 78.57%. The type I error is 17.86% and type II error is 3.57%. The performance of vertebral body fracture prediction using Decision Tree is significantly higher than using Support Vector Machine. The Decision Tree model is a potential risk assessment method for clinical application. The pilot evidence showed that Decision Tree prediction model overcomes the overfitting drawback of Support Vector Machine Model. However, larger dataset and cohort study should be conducted for further evidence

Advances in cancer therapy have prolonged patient survival even in the presence of disseminated disease and an increasing number of cancer patients are living with metastatic bone disease (MBD). The proximal femur is the most common long bone involved in MBD and pathologic fractures of the femur are associated with significant morbidity, mortality and loss of quality of life (QoL). Successful prophylactic surgery for an impending fracture of the proximal femur has been shown in multiple cohort studies to result in longer survival, preserved mobility, lower transfusion rates and shorter post-operative hospital stays. However, there is currently no optimal method to predict a pathologic fracture. The most well-known tool is Mirel's criteria, established in 1989 and is limited from guiding clinical practice due to poor specificity and sensitivity. The ideal clinical decision support tool will be of the highest sensitivity and specificity, non-invasive, generalizable to all patients, and not a burden on hospital resources or the patient's time. Our research uses novel machine learning techniques to develop a model to fill this considerable gap in the treatment pathway of MBD of the femur. The goal of our study is to train a convolutional neural network (CNN) to predict fracture risk when metastatic bone disease is present in the proximal femur. Our fracture risk prediction tool was developed by analysis of prospectively collected data of consecutive MBD patients presenting from 2009–2016. Patients with primary bone tumors, pathologic fractures at initial presentation, and hematologic malignancies were excluded. A total of 546 patients comprising 114 pathologic fractures were included. Every patient had at least one Anterior-Posterior X-ray and clinical data including patient demographics, Mirel's criteria, tumor biology, all previous radiation and chemotherapy received, multiple pain and function scores, medications and time to fracture or time to death. We have trained a convolutional neural network (CNN) with AP X-ray images of 546 patients with metastatic bone disease of the proximal femur. The digital X-ray data is converted into a matrix representing the color information at each pixel. Our CNN contains five convolutional layers, a fully connected layers of 512 units and a final output layer. As the information passes through successive levels of the network, higher level features are abstracted from the data. The model converges on two fully connected deep neural network layers that output the risk of fracture. This prediction is compared to the true outcome, and any errors are back-propagated through the network to accordingly adjust the weights between connections, until overall prediction accuracy is optimized. Methods to improve learning included using stochastic gradient descent with a learning rate of 0.01 and a momentum rate of 0.9. We used average classification accuracy and the average F1 score across five test sets to measure model performance. We compute F1 = 2 x (precision x recall)/(precision + recall). F1 is a measure of a model's accuracy in binary classification, in our case, whether a lesion would result in pathologic fracture or not. Our model achieved 88.2% accuracy in predicting fracture risk across five-fold cross validation testing. The F1 statistic is 0.87. This is the first reported application of convolutional neural networks, a machine learning algorithm, to this important Orthopaedic problem. Our neural network model was able to achieve reasonable accuracy in classifying fracture risk of metastatic proximal femur lesions from analysis of X-rays and clinical information. Our future work will aim to externally validate this algorithm on an international cohort

Advances in cancer therapy have prolonged cancer patient survival even in the presence of disseminated disease and an increasing number of cancer patients are living with metastatic bone disease (MBD). The proximal femur is the most common long bone involved in MBD and pathologic fractures of the femur are associated with significant morbidity, mortality and loss of quality of life (QoL). Successful prophylactic surgery for an impending fracture of the proximal femur has been shown in multiple cohort studies to result in patients more likely to walk after surgery, longer survival, lower transfusion rates and shorter post-operative hospital stays. However, there is currently no optimal method to predict a pathologic fracture. The most well-known tool is Mirel's criteria, established in 1989 and is limited from guiding clinical practice due to poor specificity and sensitivity. The goal of our study is to train a convolutional neural network (CNN) to predict fracture risk when metastatic bone disease is present in the proximal femur. Our fracture risk prediction tool was developed by analysis of prospectively collected data for MBD patients (2009–2016) in order to determine which features are most commonly associated with fracture. Patients with primary bone tumors, pathologic fractures at initial presentation, and hematologic malignancies were excluded. A total of 1146 patients comprising 224 pathologic fractures were included. Every patient had at least one Anterior-Posterior X-ray. The clinical data includes patient demographics, tumor biology, all previous radiation and chemotherapy received, multiple pain and function scores, medications and time to fracture or time to death. Each of Mirel's criteria has been further subdivided and recorded for each lesion. We have trained a convolutional neural network (CNN) with X-ray images of 1146 patients with metastatic bone disease of the proximal femur. The digital X-ray data is converted into a matrix representing the color information at each pixel. Our CNN contains five convolutional layers, a fully connected layers of 512 units and a final output layer. As the information passes through successive levels of the network, higher level features are abstracted from the data. This model converges on two fully connected deep neural network layers that output the fracture risk. This prediction is compared to the true outcome, and any errors are back-propagated through the network to accordingly adjust the weights between connections. Methods to improve learning included using stochastic gradient descent with a learning rate of 0.01 and a momentum rate of 0.9. We used average classification accuracy and the average F1 score across test sets to measure model performance. We compute F1 = 2 x (precision x recall)/(precision + recall). F1 is a measure of a test's accuracy in binary classification, in our case, whether a lesion would result in pathologic fracture or not. Five-fold cross validation testing of our fully trained model revealed accurate classification for 88.2% of patients with metastatic bone disease of the proximal femur. The F1 statistic is 0.87. This represents a 24% error reduction from using Mirel's criteria alone to classify the risk of fracture in this cohort. This is the first reported application of convolutional neural networks, a machine learning algorithm, to an important Orthopaedic problem. Our neural network model was able to achieve impressive accuracy in classifying fracture risk of metastatic proximal femur lesions from analysis of X-rays and clinical information. Our future work will aim to validate this algorithm on an external cohort

Effectiveness and long term stability of hip resurfacing and total hip arthroplasty for osteoarthritis patients are still debated nowadays. Several clinical and biomechanical issues have to be considered, including pain relief, return to function, femoral neck fractures, impingement and prosthesis loosening. Normally, patients with hip arthroplasties are facing gait adaptation and at risk of fall. Sudden impact loading and twisting during sideway falls may lead to femoral fractures and joint failures. The purposes of this study are (i) to investigate the stress behavior of hip resurfacing and total hip arthroplasty, and (ii) to predict pattern of femoral fractures during sideway falls and twisting configurations. Computed tomography (CT) based images of a 54-year old male were used in developing a 3D femoral model. The femur model was designed to be inhomogeneous material as defined by Hounsfield Unit of the CT images. CAD data of hip arthroplasties were imported and aligned to represent RHA and THA femur modelas shown in Fig.1. Prosthesis stem is modeled as Ti-6Al-4V material while femoral ball as Alumina properties. Meanwhile, RHA implant is assigned as Co-Cr-Mo material. Four types of loading and boundary conditions were assigned to demonstrate different falling (FC) and twisting (TC) configurations (see Fig.2). Finite element analysis combined with a damage mechanics model was then performed to predict bone fractures in both arthroplasty models. Different loading magnitudes up to 4BW were applied to extrapolate the fracture patterns. Prediction of femoral fracture for RHA and THA femurs are discussed in corresponding to maximum principal stress and damage formation criterion. The load bearing strain was set to 3000micron, the physiological bone loading that leads to bone formation. The test strength was wet to 80% of the yield strength determined from the CT images. Different locations of fracture are predicted in each configuration due to different loading direction and boundary conditions as shown in Fig.3. For falling configurations, fractures were projected at trochanteric region for intact and RHA femur, while THA femurs experience fracture at inner proximal region of bone. Differs to twisting configurations, both arthroplasties were predicted to fracture at the distal end of femurs