Medical Image Analysis Framework Research Articles

Towards Building a Trustworthy Deep Learning Framework for Medical Image Analysis.

Computer vision and deep learning have the potential to improve medical artificial intelligence (AI) by assisting in diagnosis, prediction, and prognosis. However, the application of deep learning to medical image analysis is challenging due to limited data availability and imbalanced data. While model performance is undoubtedly essential for medical image analysis, model trust is equally important. To address these challenges, we propose TRUDLMIA, a trustworthy deep learning framework for medical image analysis, which leverages image features learned through self-supervised learning and utilizes a novel surrogate loss function to build trustworthy models with optimal performance. The framework is validated on three benchmark data sets for detecting pneumonia, COVID-19, and melanoma, and the created models prove to be highly competitive, even outperforming those designed specifically for the tasks. Furthermore, we conduct ablation studies, cross-validation, and result visualization and demonstrate the contribution of proposed modules to both model performance (up to 21%) and model trust (up to 5%). We expect that the proposed framework will support researchers and clinicians in advancing the use of deep learning for dealing with public health crises, improving patient outcomes, increasing diagnostic accuracy, and enhancing the overall quality of healthcare delivery.

Label-Efficient Self-Supervised Federated Learning for Tackling Data Heterogeneity in Medical Imaging.

The collection and curation of large-scale medical datasets from multiple institutions is essential for training accurate deep learning models, but privacy concerns often hinder data sharing. Federated learning (FL) is a promising solution that enables privacy-preserving collaborative learning among different institutions, but it generally suffers from performance deterioration due to heterogeneous data distributions and a lack of quality labeled data. In this paper, we present a robust and label-efficient self-supervised FL framework for medical image analysis. Our method introduces a novel Transformer-based self-supervised pre-training paradigm that pre-trains models directly on decentralized target task datasets using masked image modeling, to facilitate more robust representation learning on heterogeneous data and effective knowledge transfer to downstream models. Extensive empirical results on simulated and real-world medical imaging non-IID federated datasets show that masked image modeling with Transformers significantly improves the robustness of models against various degrees of data heterogeneity. Notably, under severe data heterogeneity, our method, without relying on any additional pre-training data, achieves an improvement of 5.06%, 1.53% and 4.58% in test accuracy on retinal, dermatology and chest X-ray classification compared to the supervised baseline with ImageNet pre-training. In addition, we show that our federated self-supervised pre-training methods yield models that generalize better to out-of-distribution data and perform more effectively when fine-tuning with limited labeled data, compared to existing FL algorithms. The code is available at https://github.com/rui-yan/SSL-FL.

Open-source image analysis tool for the identification and quantification of cortical interruptions and bone erosions in high-resolution peripheral quantitative computed tomography images of patients with rheumatoid arthritis.

Identification of bone erosions and quantification of erosion volume is important for rheumatoid arthritis diagnosis, and can add important information to evaluate disease progression and treatment effects. High-resolution peripheral quantitative computed tomography (HR-pQCT) is well suited for this purpose, however analysis methods are not widely available. The purpose of this study was to develop an open-source software tool for the identification and quantification of bone erosions using images acquired by HR-pQCT. The collection of modules, Bone Analysis Modules (BAM) - Erosion, implements previously published erosion analysis techniques as modules in 3D Slicer, an open-source image processing and visualization tool. BAM includes a module to automatically identify cortical interruptions, from which erosions are manually selected, and a hybrid module that combines morphological and level set operations to quantify the volume of bone erosions. HR-pQCT images of the second and third metacarpophalangeal (MCP) joints were acquired in patients with RA (XtremeCT, n=14, XtremeCTII, n=22). The number of cortical interruptions detected by BAM-Erosion agreed strongly with the previously published cortical interruption detection algorithm for both XtremeCT (r2=0.85) and XtremeCTII (r2=0.87). Erosion volume assessment by BAM-Erosion agreed strongly (r2=0.95) with the Medical Image Analysis Framework. BAM-Erosion provides an open-source erosion analysis tool that produces comparable results to previously published algorithms, with improved options for visualization. The strength of the tool is that it implements multiple image processing algorithms for erosion analysis on a single, widely available, open-source platform that can accommodate future updates.

Review of Radiomics- and Dosiomics-based Predicting Models for Rectal Cancer.

By breaking the traditional medical image analysis framework, precision medicine–radiomics has attracted much attention in the past decade. The use of various mathematical algorithms offers radiomics the ability to extract vast amounts of detailed features from medical images for quantitative analysis and analyzes the confidential information related to the tumor in the image, which can establish valuable disease diagnosis and prognosis models to support personalized clinical decisions. This article summarizes the application of radiomics and dosiomics in radiation oncology. We focus on the application of radiomics in locally advanced rectal cancer and also summarize the latest research progress of dosiomics in radiation tumors to provide ideas for the treatment of future related diseases, especially 125I CT-guided radioactive seed implant brachytherapy.

A configurable deep learning framework for medical image analysis

A configurable deep learning framework for medical image analysis

Effective Classification of Synovial Sarcoma Cancer Using Structure Features and Support Vectors

In this research work, we proposed a medical image analysis framework with two separate releases whether or not Synovial Sarcoma (SS) is the cell structure for cancer. Within this framework the histopathology images are decomposed into a third-level sub-band using a two-dimensional Discrete Wavelet Transform. Subsequently, the structure features (SFs) such as Principal Components Analysis (PCA), Independent Components Analysis (ICA) and Linear Discriminant Analysis (LDA) were extracted from this sub-band image representation with the distribution of wavelet coefficients. These SFs are used as inputs of the Support Vector Machine (SVM) classifier. Also, classification of PCA + SVM, ICA + SVM, and LDA + SVM with Radial Basis Function (RBF) kernel the efficiency of the process is differentiated and compared with the best classification results. Furthermore, data collected on the internet from various histopathological centres via the Internet of Things (IoT) are stored and shared on blockchain technology across a wide range of image distribution across secure data IoT devices. Due to this, the minimum and maximum values of the kernel parameter are adjusted and updated periodically for the purpose of industrial application in device calibration. Consequently, these resolutions are presented with an excellent example of a technique for training and testing the cancer cell structure prognosis methods in spindle shaped cell (SSC) histopathological imaging databases. The performance characteristics of cross-validation are evaluated with the help of the receiver operating characteristics (ROC) curve, and significant differences in classification performance between the techniques are analyzed. The combination of LDA + SVM technique has been proven to be essential for intelligent SS cancer detection in the future, and it offers excellent classification accuracy, sensitivity, specificity.

Evaluation of bone erosion in rheumatoid arthritis patients by high-resolution peripheral quantitative computed tomography scans: Comparison between two semi-automated programs in a three-dimensional setting.

The aim of this study was to compare OsiriX software with the previous published Medical Image Analysis Framework (MIAF) method to assess the volume of erosion in patients with rheumatoid arthritis (RA). Forty RA patients underwent high-resolution peripheral quantitative computed tomography scans of the second and third metacarpophalangeal joints, and thirty-four patients with any bone erosion were enrolled. Two techniques were applied to erosion evaluation: (a) semi-automated MIAF software, and (b) semi-automated segmentation by free open-source Digital Imaging and Communications in Medicine viewer, OsiriX software. MIAF has been published before, but this is the first time that OsiriX has been used in this way in rheumatology. Bland & Altman plots described agreement between methods. Forty-eight erosions from 34 patients were analyzed. Mean age was 40.74±5.32years and mean disease duration was 10.68±4.96years. Both methods demonstrated a strong correlation regarding erosion volume (r=0.96, P<0.001). Median (interquartile range) of erosion volume was 12.14 (4.5-36.07) when MIAF was considered, and 11.80 (3.45-29.42) when the OsiriX tool was used (P=0.139). MIAF and OsiriX showed good agreement when the Bland & Altman plot was performed. Evaluation by MIAF took 22.69±6.71minutes, whereas OsiriX took only 2.62±1.09minutes (P<0.001). The three-dimensional segmentation of bone erosions can be done by both MIAF and OsiriX software with good agreement. However, because OsiriX is a widespread tool and faster, its method seems to be more feasible for evaluating peripheral bone damage, especially bone erosions.

The ANTsX ecosystem for quantitative biological and medical imaging

The Advanced Normalizations Tools ecosystem, known as ANTsX, consists of multiple open-source software libraries which house top-performing algorithms used worldwide by scientific and research communities for processing and analyzing biological and medical imaging data. The base software library, ANTs, is built upon, and contributes to, the NIH-sponsored Insight Toolkit. Founded in 2008 with the highly regarded Symmetric Normalization image registration framework, the ANTs library has since grown to include additional functionality. Recent enhancements include statistical, visualization, and deep learning capabilities through interfacing with both the R statistical project (ANTsR) and Python (ANTsPy). Additionally, the corresponding deep learning extensions ANTsRNet and ANTsPyNet (built on the popular TensorFlow/Keras libraries) contain several popular network architectures and trained models for specific applications. One such comprehensive application is a deep learning analog for generating cortical thickness data from structural T1-weighted brain MRI, both cross-sectionally and longitudinally. These pipelines significantly improve computational efficiency and provide comparable-to-superior accuracy over multiple criteria relative to the existing ANTs workflows and simultaneously illustrate the importance of the comprehensive ANTsX approach as a framework for medical image analysis.

Bone erosion in the 2nd metacarpophalangeal head: association with its bone mineral density by HR-pQCT in rheumatoid arthritis patients

BackgroundRheumatoid arthritis (RA) is a chronic autoimmune disease depicted by synovial inflammation leading to local and systemic bone loss. The aim of this study was to evaluate by a HR-pQCT (High Resolution Peripheral Quantitative Computed Tomography) study which parameters are associated with volume of bone erosions including bone mineral density (BMD) around erosions (VOI 1 to 4 = volume of interest), BMD of metacarpophalangeal (MCP) head, BMD of radius, presence of osteophytes and joint space width (JSW).MethodsFifty female RA patients (18–50 years) were enrolled in this study. Demographic and disease-specific data, laboratory inflammatory parameters and handgrip test were performed. All patients underwent HR-pQCT of 2nd and 3rd MCP joints and distal radius, according to established protocols. The volume of bone erosions was evaluated by MIAF (Medical Image Analysis Framework) software. Osteophytes were analyzed by manual method.ResultsThe mean of age and disease duration were 40.0 ± 6.0 yrs. and 10.8 ± 4.8 yrs., respectively. According to DAS-28 (Disease Activity Score), 54% (27) of the sample were in remission. However, when SDAI (Simplified Disease Activity Index) was used, only 18% (9) were under remission. The mean of HAQ (Health Assessment Questionnaire), ESR (Erythrocyte sedimentation rate) and CRP (C reactive protein) were 0.9 ± 0.7, 13.9 ± 12.2 mm and 5.6 ± 7.5 mg/mL, respectively. Forty-six bone erosions (0.9 ± 1.2 erosion/patient) and 14 osteophytes (0.3 ± 0.7 osteophyte/patient) were found in 2nd MCP head. The median (IQR-Interquartile range) of volume of erosion and volume of osteophytes were 14.9 (5.7;35.9)mm3 and 3.1 (2.1, 4.3)mm3, respectively. The mean of JSW was 80.5 ± 34.2 mm3. The volume of bone erosions was negatively correlated with BMD of 2nd MCP head, VOI-4 and JSW; and it was positively correlated with osteophytes number. Regarding absence or presence of erosion in 2nd MCP head, a significant difference was found between BMD of MCP head, osteophyte number and JSW. Multiple linear regression analysis showed that only BMD of 2nd MCP head was independently associated with volume of bone erosions.ConclusionBMD of MCP head was independently associated with volume of bone erosion, suggesting that this parameter should be used to analyze and monitoring bone destruction, as well as to evaluate treatment response in RA patients.

Lack of periosteal apposition in the head and neck of femur after menopause in Chinese women with high risk for hip fractures - A cross-sectional study with QCT.

In elderly subjects and in particular in those with osteoporosis the evidence on age related volume changes of the hip is still very limited. Even less is known about bone changes of the femoral head. The aim of this study is to explore associations of bone size of the femoral head and neck with age in postmenopausal women with very high risk of hip fracture and to investigate associations of femoral head and neck bone mineral density. MIAF (medical image analysis framework)-Femur was used for the analysis of CT datasets from 319 females with acute hip fractures age 50 to 98. Integral BMD and volume of the head and neck were assessed. The femoral head was divided into four quadrants to address differential vBMD and volume responses of its superior, inferior, posterior and anterior parts. Areal BMD (aBMD) of femoral neck was also obtained. In this population of postmenopausal women we did not observe age-related changes in bone volume of the femoral head or neck between ages 50 and 98years. Integral vBMD in the head in the 90-98year group was 48.0mg/cm3 lower than that in 50-59year group, which accounts for nearly 30% decrease in vBMD with 40years increase. Age-related vBMD changes in the head quadrants were similar to that in total. With age, the trend line correlation coefficients for vBMD in quadrants were relatively small, but significant (p<0.001) for all. The femoral head integral vBMD correlates well with neck vBMD and FN aBMD. FN aBMD explained 45% of head integral vBMD variance (p<0.0001). Elderly women had relative preservation of femoral head and neck bone volume from 50yrs. over four decades but markedly lower integral vBMD of proximal femur. The findings of our study call in question about the concept of bone expansion with aging even in elderly age.

Methods for segmentation of rheumatoid arthritis bone erosions in high-resolution peripheral quantitative computed tomography (HR-pQCT).

The comparison between different techniques to quantify the 3-dimensional size of inflammatory bone erosions in rheumatoid arthritis(RA) patients. Anti-cyclic citrullinated peptide antibody(ACPA) positive RA patients received high-resolution peripheral quantitative computed tomography (HR-pQCT) scans of the metacarpophalangeal joints (MCP). Erosions were measured by three different segmentation techniques: (1) manual method with calculation by half-ellipsoid formula, (2) semi-automated modified Evaluation Script for Erosions (mESE), and (3) semi-automated Medical Image Analysis Framework (MIAF) software. Bland & Altman plots were used to describe agreement between methods. Furthermore, shape of erosions was classified as regular or irregular and then compared to the sphericity obtained by MIAF. A total of 76 erosions from 65 RA patients (46 females/19 males), median age 57 years, median disease duration 6.1 years and median disease activity score 28 of 2.8 units were analyzed. While mESE and MIAF showed good agreement in the measurement of erosion size, the manual method with calculation by half-ellipsoid formula underestimated erosions size, particularly with larger erosions. Accurate segmentation is particularly important in larger erosions, which are irregularly shaped. In all three segmentation techniques irregular erosions were larger in size than regular erosions (MIAF: 19.7 vs. 3.4mm3; mESE: 15.5 vs. 2.3mm3; manual = 7.2 vs. 1.52mm3; all p < 0.001). In accordance, sphericity of erosions measured by MIAF significantly decreased with their size (p < 0.001). MIAF and mESE allow segmentation of inflammatory bone erosions in RA patients with excellent inter reader reliability. They allow calculating erosion volume independent of erosion shape and therefore provide an attractive tool to quantify structural damage in individual joints of RA patients.

Automative Multi Classifier Framework for Medical Image Analysis

Medical image processing is the technique used to create images of the human body for medical purposes. Nowadays, medical image processing plays a major role and a challenging solution for the critical stage in the medical line. Several researches have done in this area to enhance the techniques for medical image processing. However, due to some demerits met by some advanced technologies, there are still many aspects that need further development. Existing study evaluate the efficacy of the medical image analysis with the level-set shape along with fractal texture and intensity features to discriminate PF (Posterior Fossa) tumor from other tissues in the brain image. To develop the medical image analysis and disease diagnosis, to devise an automotive subjective optimality model for segmentation of images based on different sets of selected skin texture from the unsupervised learning model of extracted features. After segmentation, classification of images is done. The classification is processed by adapting the multiple classifier frameworks in the previous work based on the mutual information coefficient of the selected features underwent for image segmentation procedures. In this study, to enhance the classification strategy, we plan to implement enhanced multi classifier framework for the analysis of medical images and disease diagnosis. The performance parameter used for the analysis of the proposed enhanced multi classifier framework for medical image analysis is Multiple Class intensity, image quality, time consumption.

Improvements in hip trabecular, subcortical, and cortical density and mass in postmenopausal women with osteoporosis treated with denosumab

In the FREEDOM study, denosumab treatment (60mg every 6months) decreased bone resorption, increased bone mineral density (BMD), and reduced new vertebral, nonvertebral, and hip fractures over 36months in postmenopausal women with osteoporosis. In a subset of these women, hip quantitative computed tomography (QCT) was performed at baseline and months 12, 24, and 36. These scans were analyzed using Medical Image Analysis Framework (MIAF) software, which allowed assessment of total hip integral, trabecular, subcortical, and cortical compartments; the cortical compartment was further divided into 2 areas of interest (outer and inner cortex). This substudy reports changes in BMD and bone mineral content (BMC) from baseline and compared placebo with denosumab over 36months of treatment (placebo N=26; denosumab N=36). Denosumab treatment resulted in significant improvements in total hip integral volumetric BMD (vBMD) and BMC from baseline at each time point. At month 36, the mean percentage increase from baseline in total hip integral vBMD and BMC was 6.4% and 4.8%, respectively (both p<0.0001). These gains were accounted for by significant increases in vBMD and BMC in the trabecular, subcortical, and cortical compartments. In the placebo group, total hip integral vBMD and BMC decreased at month 36 from baseline by −1.5% and −2.6%, respectively (both p<0.05). The differences between denosumab and placebo were also significant at months 12, 24, and 36 for integral, trabecular, subcortical, and cortical vBMD and BMC (all p<0.05 to <0.0001). While the largest percentage differences occurred in trabecular vBMD and BMC, the largest absolute differences occurred in cortical vBMD and BMC. In summary, denosumab significantly improved both vBMD and BMC from baseline and placebo, assessed by QCT MIAF, in the integral, trabecular, subcortical, and cortical hip compartments, all of which are relevant to bone strength.

Three-dimensional structural analysis of the proximal femur in an age-stratified sample of women

IntroductionAging is associated with worsening bone structure and increasing risk of hip fracture. However, the commonly used clinical tool, dual-energy x-ray absorptiometry, does not provide information on changes with age or disease separately in trabecular versus cortical bone or in bone geometry. Here we used 3D quantitative computed tomography (QCT) to analyze age-related changes in femoral volumetric bone mineral density (vBMD) and structure in a well characterized, population-based cohort of Rochester, Minnesota women. MethodsMIAF-Femur (MIAF: medical image analysis framework) was used for the analysis of CT datasets from 358 women age 20 to 97years. Integral, “apparent” cortical (rather than true cortical vBMD, due to volume averaging effects) and trabecular vBMD, volume, and bone mineral content (BMC) as well as cortical thickness of the femur head, neck, trochanter, inter-trochanteric, and proximal shaft volumes of interest (VOIs) were measured. In addition, changes in vBMD in the superior, inferior, posterior and anterior quadrants of the femur neck were assessed. ResultsCross-sectional percent decreases in vBMD across life were 2- to 5-fold higher in trabecular versus cortical bone at all sites in the femur, although absolute changes in the trabecular and cortical bone were fairly similar. In addition, the slopes of the relationships of trabecular vBMD with age were generally similar in pre- and postmenopausal women, whereas apparent cortical vBMD in the femur neck, trochanter, inter-trochanteric region, and proximal shaft remained relatively stable in premenopausal women but decreased significantly with age following the menopause. Bone volume increased at all sites, more so in pre- compared to postmenopausal women. Age-related BMC changes were not significant in premenopausal women, but BMC losses were highly significant in postmenopausal women. Detailed analyses of femur neck cortical bone showed that percent apparent vBMD decreases in the superior quadrants were 2- to 3-fold greater than in the inferior quadrants; changes in absolute values were most different (~2-fold) between the superior-posterior and inferior-posterior quadrants. ConclusionsThese data demonstrate that patterns of changes with age within the femur differ in the trabecular versus cortical bone. In the cortical compartment which, due to limitations in spatial resolution, contains some subcortical bone and should be regarded as an “apparent” cortical VOI, the superior quadrants in the femur neck undergo the greatest decreases. These findings may have important implications for understanding the structural basis for increased hip fracture risk with aging.

CT imaging for the investigation of subchondral bone in knee osteoarthritis

In osteoarthritis, magnetic resonance imaging is the method of choice to image articular cartilage and "bone marrow lesions." However, the calcified cartilage, the subchondral bone plate, and trabecular subchondral bone that are mineralized tissues strongly attenuate X-rays and are therefore potentially accessible for analyses using computed tomography (CT). CT images nicely show osseous cardinal signs of advanced osteoarthritis such as osteophytes, subchondral cysts, and subchondral bone sclerosis. But more importantly, CT can help us to better understand the pathophysiology of knee osteoarthritis from the measurement of the density and structure of subchondral mineralized tissues in vivo. For that purpose, we recently developed dedicated image analysis software called Medical Image Analysis Framework (MIAF)-Knee. In this manuscript, our aims are to present current knowledge on CT imaging of the subchondral bone in knee osteoarthritis and to provide a brief introduction to basic technical aspects of MIAF-Knee as well as preliminary results we obtained in patients with knee osteoarthritis as compared to control subjects.

In vivo discrimination of hip fracture with quantitative computed tomography: Results from the prospective European Femur Fracture Study (EFFECT)

In assessing osteoporotic fractures of the proximal femur, the main objective of this in vivo case-control study was to evaluate the performance of quantitative computed tomography (QCT) and a dedicated 3D image analysis tool [Medical Image Analysis Framework--Femur option (MIAF-Femur)] in differentiating hip fracture and non-hip fracture subjects. One-hundred and seven women were recruited in the study, 47 women (mean age 81.6 years) with low-energy hip fractures and 60 female non-hip fracture control subjects (mean age 73.4 years). Bone mineral density (BMD) and geometric variables of cortical and trabecular bone in the femoral head and neck, trochanteric, and intertrochanteric regions and proximal shaft were assessed using QCT and MIAF-Femur. Areal BMD (aBMD) was assessed using dual-energy X-ray absorptiometry (DXA) in 96 (37 hip fracture and 59 non-hip fracture subjects) of the 107 patients. Logistic regressions were computed to extract the best discriminates of hip fracture, and area under the receiver characteristic operating curve (AUC) was calculated. Three logistic models that discriminated the occurrence of hip fracture with QCT variables were obtained (AUC = 0.84). All three models combined one densitometric variable--a trabecular BMD (measured in the femoral head or in the trochanteric region)--and one geometric variable--a cortical thickness value (measured in the femoral neck or proximal shaft). The best discriminant using DXA variables was obtained with total femur aBMD (AUC = 0.80, p = .003). Results highlight a synergistic contribution of trabecular and cortical components in hip fracture risk and the utility of assessing QCT BMD of the femoral head for improved understanding and possible insights into prevention of hip fractures.

Medical Data Fusion for Telemedicine

In this article, we describe a framework for distributed statistical analysis of medical images that operates under minimal bandwidth requirements. This framework implements a distributed dynamic recursive partitioning algorithm for medical image analysis that integrates medical image data repositories that contain multiple studies and are potentially located at spatially distributed clinical sites. The goal is to detect associations among regions of interest (ROIs) in images and clinical properties such as the progression of a disease. Statistical descriptors are computed from the ROIs in order to assist in medical decision making by facilitating automatic characterization, classification, and similarity searches of ROIs (e.g., lesions, tumors, and regions of morphological variability). The system consists of a central information fusion site that coordinates the analysis by communicating with remotely located processing sites. The described system has the advantage of keeping bandwidth requirements to a minimum by reducing the amount of data that need to be transferred, while medical decision making is not affected by the data reduction. This benefit makes the proposed distributed medical image analysis framework very suitable for deploying effective telemedical applications.

Volumetric quantitative computed tomography of the proximal femur: relationships linking geometric and densitometric variables to bone strength. Role for compact bone

In assessing cervical fractures of the proximal femur, this in vitro quantitative computed tomography (QCT) study had three objectives: to compare QCT to dual-energy X-ray absorptiometry (DXA) for predicting the failure load of the proximal femur, to compare the contributions of density and geometry to bone failure load, and to compare the contributions of cortical and trabecular bone to bone failure load. A novel three-dimensional (3D) analysis tool [medical image analysis framework (MIAF-Femur)] was used to analyze QCT scans. The proximal ends of 28 excised femurs were studied (1) using QCT to separately measure bone mineral density (BMD) and geometric variables of trabecular and cortical bone, (2) using mechanical tests to failure in a stance configuration, and (3) using DXA to measure BMD. The variables were described with mean, standard deviation, and range. Correlation matrix and multivariate linear models were computed. Among correlations, cortical thicknesses of the femoral neck were significantly correlated with femoral failure load, especially of the inferoanterior quadrant (r2=0.41; p<0.001), as was cortical volume at the "extended neck" (r2=0.41; p<0.001). Femoral failure load variance was best explained by a combination of QCT variables. Combining densitometric and geometric variables measured by QCT explained 76% of femoral failure load variance compared with 69% with the DXA model. Geometric variables (measured by QCT) explained 43% of femoral failure load variance compared with 72% for densitometric variables (measured by QCT). A model including only trabecular variables explained 52% of femoral failure load variance compared with 59% for a model including only cortical variables. The QCT-MIAF reported here provides analysis of both geometric and densitometric variables characterizing cortical and trabecular bone. Confirmation of our results in an independent sample would suggest that QCT may better explain failure load variance for cervical fracture than the gold standard DXA-provided BMD.