Topological regression as an interpretable and efficient tool for quantitative structure-activity relationship modeling.

Quantitative structure-activity relationship (QSAR) modeling is a powerful tool for drug discovery, yet the lack of interpretability of commonly used QSAR models hinders their application in molecular design. We propose a similarity-based regression framework, topological regression (TR), that offers a statistically grounded, computationally fast, and interpretable technique to predict drug responses. We compare the predictive performance of TR on 530 ChEMBL human target activity datasets against the predictive performance of deep-learning-based QSAR models. Our results suggest that our sparse TR model can achieve equal, if not better, performance than the deep learning-based QSAR models and provide better intuitive interpretation by extracting an approximate isometry between the chemical space of the drugs and their activity space.

Federated learning framework integrating REFINED CNN and Deep Regression Forests.

Summary: Predictive learning from medical data incurs additional challenge due to concerns over privacy and security of personal data. Federated learning, intentionally structured to preserve high level of privacy, is emerging to be an attractive way to generate cross-silo predictions in medical scenarios. However, the impact of severe population-level heterogeneity on federated learners is not well explored. In this article, we propose a methodology to detect presence of population heterogeneity in federated settings and propose a solution to handle such heterogeneity by developing a federated version of Deep Regression Forests. Additionally, we demonstrate that the recently conceptualized REpresentation of Features as Images with NEighborhood Dependencies CNN framework can be combined with the proposed Federated Deep Regression Forests to provide improved performance as compared to existing approaches. Availability and implementation: The Python source code for reproducing the main results are available on GitHub: https://github.com/DanielNolte/FederatedDeepRegressionForests. Contact: ranadip.pal@ttu.edu. Supplementary information: Supplementary data are available at Bioinformatics Advances online.

3D shape reconstruction of the femur from planar X-ray images using statistical shape and appearance models.

Major trauma is a condition that can result in severe bone damage. Customised orthopaedic reconstruction allows for limb salvage surgery and helps to restore joint alignment. For the best possible outcome three dimensional (3D) medical imaging is necessary, but its availability and access, especially in developing countries, can be challenging. In this study, 3D bone shapes of the femur reconstructed from planar radiographs representing bone defects were evaluated for use in orthopaedic surgery. Statistical shape and appearance models generated from 40 cadaveric X-ray computed tomography (CT) images were used to reconstruct 3D bone shapes. The reconstruction simulated bone defects of between 0% and 50% of the whole bone, and the prediction accuracy using anterior-posterior (AP) and anterior-posterior/medial-lateral (AP/ML) X-rays were compared. As error metrics for the comparison, measures evaluating the distance between contour lines of the projections as well as a measure comparing similarities in image intensities were used. The results were evaluated using the root-mean-square distance for surface error as well as differences in commonly used anatomical measures, including bow, femoral neck, diaphyseal-condylar and version angles between reconstructed surfaces from the shape model and the intact shape reconstructed from the CT image. The reconstructions had average surface errors between 1.59 and 3.59 mm with reconstructions using the contour error metric from the AP/ML directions being the most accurate. Predictions of bow and femoral neck angles were well below the clinical threshold accuracy of 3°, diaphyseal-condylar angles were around the threshold of 3° and only version angle predictions of between 5.3° and 9.3° were above the clinical threshold, but below the range reported in clinical practice using computer navigation (i.e., 17° internal to 15° external rotation). This study shows that the reconstructions from partly available planar images using statistical shape and appearance models had an accuracy which would support their potential use in orthopaedic reconstruction.

Reconstruction of the lower limb bones from digitised anatomical landmarks using statistical shape modelling.

BACKGROUND: Bone shapes strongly influence force and moment predictions of kinematic and musculoskeletal models used in motion analysis. The precise determination of joint reference frames is essential for accurate predictions. Since clinical motion analysis typically does not include medical imaging, from which bone shapes may be obtained, scaling methods using reference subjects to create subject-specific bone geometries are widely used. RESEARCH QUESTION: This study investigated if lower limb bone shape predictions from skin-based measurements, utilising an underlying statistical shape model (SSM) that corrects for soft tissue artefacts in digitisation, can be used to improve conventional linear scaling methods of bone geometries. METHODS: SSMs created from 35 healthy adult femurs and tibiae/fibulae were used to reconstruct bone shapes by minimising the distance between anatomical landmarks on the models and those digitised in the motion laboratory or on medical images. Soft tissue artefacts were quantified from magnetic resonance images and then used to predict distances between landmarks digitised on the skin surface and bone. Reconstruction results were compared to linearly scaled models by measuring root mean squared distances to segmented surfaces, calculating differences of commonly used anatomical measures and the errors in the prediction of the hip joint centre. RESULTS: SSM reconstructed surface predictions from varying landmark sets from skin and bone landmarks were more accurate compared to linear scaling methods (2.60-2.95 mm vs. 3.66-3.87 mm median error; p < 0.05). No significant differences were found between SSM reconstructions from bony landmarks and SSM reconstructions from digitised landmarks obtained in the motion lab and therefore reconstructions using skin landmarks are as accurate as reconstructions from landmarks obtained from medical images. SIGNIFICANCE: These results indicate that SSM reconstructions can be used to increase the accuracy in obtaining bone shapes from surface digitised experimental data acquired in motion lab environments.

Improving Musculoskeletal Model Scaling Using an Anatomical Atlas: The Importance of Gender and Anthropometric Similarity to Quantify Joint Reaction Forces.

OBJECTIVE: The accuracy of a musculoskeletal model relies heavily on the implementation of the underlying anatomical dataset. Linear scaling of a generic model, despite being time and cost efficient, produces substantial errors as it does not account for gender differences and inter-individual anatomical variations. The hypothesis of this study is that linear scaling to a musculoskeletal model with gender and anthropometric similarity to the individual subject produces similar results to the ones that can be obtained from a subject-specific model. METHODS: A lower limb musculoskeletal anatomical atlas was developed consisting of ten datasets derived from magnetic resonance imaging of healthy subjects and an additional generic dataset from the literature. Predicted muscle activation and joint reaction force were compared with electromyography and literature data. Regressions based on gender and anthropometry were used to identify the use of atlas. RESULTS: Primary predictors of differences for the joint reaction force predictions were mass difference for the ankle (p < 0.001) and length difference for the knee and hip (p ≤ 0.017). Gender difference accounted for an additional 3% of the variance (p ≤ 0.039). Joint reaction force differences at the ankle, knee, and hip were reduced by between 50% and 67% (p = 0.005) when using a musculoskeletal model with the same gender and similar anthropometry in comparison with a generic model. CONCLUSION: Linear scaling with gender and anthropometric similarity can improve joint reaction force predictions in comparison with a scaled generic model. SIGNIFICANCE: The presented scaling approach and atlas can improve the fidelity and utility of musculoskeletal models for subject-specific applications.

Femur finite element model instantiation from partial anatomies using statistical shape and appearance models.

Accurate models of bone shapes are essential for orthopaedic reconstructions. The commonly used methods of using the contralateral side requires an intact bone and anatomical symmetry. Recent studies have shown that statistical shape and appearance models (SSAMs) as an alternative can predict accurate geometric models, but the accuracy of the mechanical property prediction is typically not addressed. This study compares stress and strain differences under identical loading conditions for reconstructions from partial anatomies. SSAMs representing shape and grey values were created using 40 female cadaveric X-ray computed tomography scans. Finite element models were created for shape reconstructions from partial bone of various lengths with boundary conditions obtained from musculoskeletal simulations. Commonly used anatomical measures, measures of the surface deviations and maximal stresses and strains were used to compare the reconstruction accuracy to the contralateral side. Surface errors were smaller compared to the contralateral side for reconstructions with 90% of the bone and slightly bigger when less bone was available. Anatomical measures were comparable. The contralateral side showed slightly smaller relative errors for strains of up to 6% on average. This study has shown that SSAM reconstructions using partial bone geometry are a possible alternative to the contralateral side.

Anthropometric Scaling of Anatomical Datasets for Subject-Specific Musculoskeletal Modelling of the Shoulder.

Linear scaling of generic shoulder models leads to substantial errors in model predictions. Customisation of shoulder modelling through magnetic resonance imaging (MRI) improves modelling outcomes, but model development is time and technology intensive. This study aims to validate 10 MRI-based shoulder models, identify the best combinations of anthropometric parameters for model scaling, and quantify the improvement in model predictions of glenohumeral loading through anthropometric scaling from this anatomical atlas. The shoulder anatomy was modelled using a validated musculoskeletal model (UKNSM). Ten subject-specific models were developed through manual digitisation of model parameters from high-resolution MRI. Kinematic data of 16 functional daily activities were collected using a 10-camera optical motion capture system. Subject-specific model predictions were validated with measured muscle activations. The MRI-based shoulder models show good agreement with measured muscle activations. A tenfold cross-validation using the validated personalised shoulder models demonstrates that linear scaling of anthropometric datasets with the most similar ratio of body height to shoulder width and from the same gender (p < 0.04) yields best modelling outcomes in glenohumeral loading. The improvement in model reliability is significant (p < 0.02) when compared to the linearly scaled-generic UKNSM. This study may facilitate the clinical application of musculoskeletal shoulder modelling to aid surgical decision-making.

Analysis of shoulder compressive and shear forces during functional activities of daily life.

BACKGROUND: Knowledge of forces acting through the glenohumeral joint during activities of daily living is a prerequisite for improving implant design and aiding rehabilitation planning. Existing data are limited by the number of activities performed and, in some cases, the lack of representation of the glenohumeral loading direction, although high shear force components may cause joint dislocation or implant loosening. This study aims to analyse shoulder compression and shear force components during essential functional activities of daily living. METHODS: This is a combined modelling and experimental study. Motion data and external forces measured from 25 participants for 26 activities of daily living serve as input into an upper limb musculoskeletal model that quantifies glenohumeral loading. FINDINGS: The shoulder contact force exceeds 50% of the body weight in 10/26 activities of daily living with a maximum contact force of 164% of the body weight (SD 69%) for a sit to stand task. The ratio of glenohumeral shear force component to compression force component exceeds 0.5 in 8/26 functional activities, with maximum ratios for reaching across the body (1.09; SD 0.41) and pick and place an everyday object (0.88; SD 0.36). INTERPRETATION: This study demonstrates substantial loads through the glenohumeral joint during activities of daily living. The ratios of glenohumeral shear force component to compression force component are considerable when high loads act at long lever arms and at high angles of arm elevation. These glenohumeral ratios represent a key component of loading that should be considered when designing implants, surgical procedures, or rehabilitation protocols.

The contribution of the glenoid labrum to glenohumeral stability under physiological joint loading using finite element analysis.

The labrum contributes to passive glenohumeral joint stability. Cadaveric studies have demonstrated that this has position and load dependency, which has not been quantified under physiological loads. This study aims to validate subject-specific finite element (FE) models against in vitro measurements of joint stability and to utilise the FE models to predict joint stability under physiological loads. The predicted stability values were within ± one standard deviation of experimental data and the FE models showed a reduction in stability of 10-15% with high, physiological, loads. The developed regression equations provide the first representation of passive glenohumeral stability and will aid surgical decision-making.

Introduction to Force-Dependent Kinematics: Theory and Application to Mandible Modeling.

Knowledge of the muscle, ligament, and joint forces is important when planning orthopedic surgeries. Since these quantities cannot be measured in vivo under normal circumstances, the best alternative is to estimate them using musculoskeletal models. These models typically assume idealized joints, which are sufficient for general investigations but insufficient if the joint in focus is far from an idealized joint. The purpose of this study was to provide the mathematical details of a novel musculoskeletal modeling approach, called force-dependent kinematics (FDK), capable of simultaneously computing muscle, ligament, and joint forces as well as internal joint displacements governed by contact surfaces and ligament structures. The method was implemented into the anybody modeling system and used to develop a subject-specific mandible model, which was compared to a point-on-plane (POP) model and validated against joint kinematics measured with a custom-built brace during unloaded emulated chewing, open and close, and protrusion movements. Generally, both joint models estimated the joint kinematics well with the POP model performing slightly better (root-mean-square-deviation (RMSD) of less than 0.75 mm for the POP model and 1.7 mm for the FDK model). However, substantial differences were observed when comparing the estimated joint forces (RMSD up to 24.7 N), demonstrating the dependency on the joint model. Although the presented mandible model still contains room for improvements, this study shows the capabilities of the FDK methodology for creating joint models that take the geometry and joint elasticity into account.

Non-linear scaling of a musculoskeletal model of the lower limb using statistical shape models.

Accurate muscle geometry for musculoskeletal models is important to enable accurate subject-specific simulations. Commonly, linear scaling is used to obtain individualised muscle geometry. More advanced methods include non-linear scaling using segmented bone surfaces and manual or semi-automatic digitisation of muscle paths from medical images. In this study, a new scaling method combining non-linear scaling with reconstructions of bone surfaces using statistical shape modelling is presented. Statistical Shape Models (SSMs) of femur and tibia/fibula were used to reconstruct bone surfaces of nine subjects. Reference models were created by morphing manually digitised muscle paths to mean shapes of the SSMs using non-linear transformations and inter-subject variability was calculated. Subject-specific models of muscle attachment and via points were created from three reference models. The accuracy was evaluated by calculating the differences between the scaled and manually digitised models. The points defining the muscle paths showed large inter-subject variability at the thigh and shank - up to 26mm; this was found to limit the accuracy of all studied scaling methods. Errors for the subject-specific muscle point reconstructions of the thigh could be decreased by 9% to 20% by using the non-linear scaling compared to a typical linear scaling method. We conclude that the proposed non-linear scaling method is more accurate than linear scaling methods. Thus, when combined with the ability to reconstruct bone surfaces from incomplete or scattered geometry data using statistical shape models our proposed method is an alternative to linear scaling methods.

In Vivo Knee Contact Force Prediction Using Patient-Specific Musculoskeletal Geometry in a Segment-Based Computational Model.

Segment-based musculoskeletal models allow the prediction of muscle, ligament, and joint forces without making assumptions regarding joint degrees-of-freedom (DOF). The dataset published for the "Grand Challenge Competition to Predict in vivo Knee Loads" provides directly measured tibiofemoral contact forces for activities of daily living (ADL). For the Sixth Grand Challenge Competition to Predict in vivo Knee Loads, blinded results for "smooth" and "bouncy" gait trials were predicted using a customized patient-specific musculoskeletal model. For an unblinded comparison, the following modifications were made to improve the predictions: further customizations, including modifications to the knee center of rotation; reductions to the maximum allowable muscle forces to represent known loss of strength in knee arthroplasty patients; and a kinematic constraint to the hip joint to address the sensitivity of the segment-based approach to motion tracking artifact. For validation, the improved model was applied to normal gait, squat, and sit-to-stand for three subjects. Comparisons of the predictions with measured contact forces showed that segment-based musculoskeletal models using patient-specific input data can estimate tibiofemoral contact forces with root mean square errors (RMSEs) of 0.48-0.65 times body weight (BW) for normal gait trials. Comparisons between measured and predicted tibiofemoral contact forces yielded an average coefficient of determination of 0.81 and RMSEs of 0.46-1.01 times BW for squatting and 0.70-0.99 times BW for sit-to-stand tasks. This is comparable to the best validations in the literature using alternative models.

Influence of the fixation stability on the healing time--a numerical study of a patient-specific fracture healing process.

BACKGROUND: The healing outcome of long bone fractures is strongly influenced by the interfragmentary movement of the bone fragments. This depends on the fixation stability, the optimum value of which is still not known. The aim of this study was to simulate a patient-specific human healing process using a numerical algorithm and to retrospectively analyse the influence of the fixation stability on the healing time. METHODS: The healing simulation was processed as an initial value problem. This was iteratively solved based on two mechanical (invariants of the strain tensor, calculated through a finite element analysis) and five biological state variables (local tissue composition and blood perfusion) using a previously published fuzzy logic algorithm. For validation purposes, the calculated interfragmentary movement was compared to in vivo measurements of this patient. By changing clinically adjustable parameters of the fixation device, the influence of the fixation stability on the healing time was analysed. FINDING: The time course showed good agreement of the interfragmentary movement compared with the in vivo measurements. The predicted healing time was strongly influenced by the fixation stability, i.e. by changing the parameters of the fixation device, it was possible to significantly reduce the healing time. INTERPRETATION: The time to heal could be greatly reduced by modification of the fixator design, i.e. increasing the fixation stiffness. When using external fixation devices, this could be achieved by decreasing the free bending length of the pins, using a stiff fixation body and a stiff connection between the pins and the body.

Novel point mutations in the mitochondrial DNA detected in patients with dilated cardiomyopathy by screening the whole mitochondrial genome.

Dilated cardiomyopathy (DCM) is widely accepted as a pluricausal or multifactorial disease. Because of the linkage between energy metabolism in the mitochondria and cardiac muscle contraction, it is reasonable to assume that mitochondrial abnormalities may be responsible for some forms of DCM. We analysed the whole mitochondrial genome in a series of 45 patients with DCM for alterations and compared the findings with those of 62 control subjects. A total of 458 sequence changes could be identified. These sequence changes were distributed among the whole mitochondrial DNA (mtDNA). An increased number of novel missense mutations could be detected nearly in all genes encoding for protein subunits in DCM patients. In genes coding for NADH dehydrogenase subunits the number of mtDNA mutations detected in patients with DCM was significantly increased (p < 0.05) compared with control subjects. Eight mutations were found to occur in conserved amino acids in the above species. The c.5973G > A (Ala-Trp) and the c.7042T > G (Val-Asp) mutations were located in highly conserved domains of the gene coding for cytochrome c oxidase subunit. Two tRNA mutations could be detected in the mtDNA of DCM patients alone. The T-C transition at nt 15,924 is connected with respiratory enzyme deficiency, mitochondrial myopathy, and cardiomyopathy. The c.16189T > C mutation in the D-loop region that is associated with susceptibility to DCM could be detected in 15.6% of patients as well as in 9.7% of controls. Thus, mutations altering the function of the enzyme subunits of the respiratory chain can be relevant for the pathogenesis of dilated cardiomyopathy.