How Artificial Intelligence and Machine Learning Is Assisting Us to Extract Meaning from Data on Bone Mechanics?

Dramatic advancements in interdisciplinary research with the fourth paradigm of science, especially the implementation of computer science, nourish the potential for artificial intelligence (AI), machine learning (ML), and artificial neural network (ANN) algorithms to be applied to studies concerning mechanics of bones. Despite recent enormous advancement in techniques, gaining deep knowledge to find correlations between bone shape, material, mechanical, and physical responses as well as properties is a daunting task. This is due to both complexity of the material itself and the convoluted shapes that this complex material forms. Moreover, many uncertainties and ambiguities exist concerning the use of traditional computational techniques that hinders gaining a full comprehension of this advanced biological material. This book chapter offers a review of literature on the use of AI, ML, and ANN in the study of bone mechanics research. A main question as to why to implement AI and ML in the mechanics of bones is fully addressed and explained. This chapter also introduces AI and ML and elaborates on the main features of ML algorithms such as learning paradigms, subtypes, main ideas with examples, performance metrics, training algorithms, and training datasets. As a frequently employed ML algorithm in bone mechanics, feedforward ANNs are discussed to make their taxonomy and working principles more readily comprehensible to researchers. A summary as well as detailed review of papers that employed ANNs to learn from collected data on bone mechanics are presented. Reviewing literature on the use of these data-driven tools is essential since their wider application has the potential to: improve clinical assessments enabling real-time simulations; avoid and/or minimize injuries; and, encourage early detection of such injuries in the first place.

Feedforward backpropagation artificial neural networks for predicting mechanical responses in complex nonlinear structures: A study on a long bone.

Feedforward backpropagation artificial neural networks (ANNs) have been increasingly employed in many engineering practices concerning materials modeling. Despite their extensive applications, how to achieve successfully trained ANNs is not thoroughly explained in the literature, nor are there lucid discussions to delineate influential parameters obtained from analyses. Long bones are composite materials possessing nonhomogeneous and anisotropic properties, and their mechanical responses exhibit dependency on numerous variables. Material complexity hinders researchers from arriving at a consensus in implementing an optimal constitutive model or encourages them to adopt a simple constitutive model including many simplifying assumptions. However, such exceptional features and engineering challenges make long bones materials worth investigating, enriching our comprehension of complex engineering structures using novel techniques where traditional methods may present limitations. This paper reports on the prediction of loading, displacement, load and displacement simultaneously, and strains using feedforward backpropagation ANNs trained with experimental recordings. The technique was used to find optimum network structures (architectures) that encompass the best prediction ability. To enhance predictions, the influence of several elements such as a network training algorithm, injecting noise to datasets prior to training, the level of injected noise which directly affects model fitting and regularization, and data normalization prior to training were investigated and discussed. Essential parameters influencing decision making in identifying well-trained and well-generalized ANNs were elaborated. A considerable emphasis in this study was placed on examining the generalization ability of the already trained and tested ANNs, thus guaranteeing unbiased models that avoided overfitting. Gaining favorable outcomes in this study required three years of performing experiments and data collection before establishing the networks. The subsequent training, testing, and determination of the generalization of more than 60,000 ANNs are promising and will assist researchers in comprehending mechanical responses of complicated engineering structures that exhibit peculiar nonlinear properties.

What can artificial intelligence and machine learning tell us? A review of applications to equine biomechanical research.

Artificial intelligence (AI) and machine learning (ML) are fascinating interdisciplinary scientific domains where machines are provided with an approximation of human intelligence. The conjecture is that machines are able to learn from existing examples, and employ this accumulated knowledge to fulfil challenging tasks such as regression analysis, pattern classification, and prediction. The horse biomechanical models have been identified as an alternative tool to investigate the effects of mechanical loading and induced deformations on the tissues and structures in humans. Many reported investigations into bone fatigue, subchondral bone damage in the joints of both humans and animals, and identification of vital parameters responsible for retaining integrity of anatomical regions during normal activities in all species are heavily reliant on equine biomechanical research. Horse racing is a lucrative industry and injury prevention in expensive thoroughbreds has encouraged the implementation of various measurement techniques, which results in massive data generation. ML substantially accelerates analysis and interpretation of data and provides considerable advantages over traditional statistical tools historically adopted in biomechanical research. This paper provides the reader with: a brief introduction to AI, taxonomy and several types of ML algorithms, working principle of a feedforward artificial neural network (ANN), and, a detailed review of the applications of AI, ML, and ANN in equine biomechanical research (i.e. locomotory system function, gait analysis, joint and bone mechanics, and hoof function). Reviewing literature on the use of these data-driven tools is essential since their wider application has the potential to: improve clinical assessments enabling real-time simulations, avoid and/or minimize injuries, and encourage early detection of such injuries in the first place.

Prediction of load in a long bone using an artificial neural network prediction algorithm.

The hierarchical nature of bone makes it a difficult material to fully comprehend. The equine third metacarpal (MC3) bone experiences nonuniform surface strains, which are a measure of displacement induced by loads. This paper investigates the use of an artificial neural network expert system to quantify MC3 bone loading. Previous studies focused on determining the response of bone using load, bone geometry, mechanical properties, and constraints as input parameters. This is referred to as a forward problem and is generally solved using numerical techniques such as finite element analysis (FEA). Conversely, an inverse problem has to be solved to quantify load from the measurements of strain and displacement. Commercially available FEA packages, without manipulating their underlying algebraic formulae, are incapable of completing a solution to the inverse problem. In this study, an artificial neural network (ANN) was employed to quantify the load required to produce the MC3 displacement and surface strains determined experimentally. Nine hydrated MC3 bones from thoroughbred horses were loaded in compression in an MTS machine. Ex-vivo experiments measured strain readings from one three-gauge rosette and three distinct single-element gauges at different locations on the MC3 midshaft, associated displacement, and load exposure time. Horse age and bone side (left or right limb) were also recorded for each MC3 bone. This information was used to construct input variables for the ANN model. The ability of this expert system to predict the MC3 loading was investigated. The ANN prediction offered excellent reliability for the prediction of load in the MC3 bones investigated, i.e. R2 ≥ 0.98.

Accuracy Quantification of the Reverse Engineering and High-Order Finite Element Analysis of Equine MC3 Forelimb.

Shape is a key factor in influencing mechanical responses of bones. Considered to be smart viscoelastic and inhomogeneous materials, bones are stimulated to change shape (model and remodel) when they experience changes in the compressive strain distribution. Using reverse engineering techniques via computer-aided design (CAD) is crucial to create a virtual environment to investigate the significance of shape in biomechanical engineering. Nonetheless, data are lacking to quantify the accuracy of generated models and to address errors in finite element analysis (FEA). In the present study, reverse engineering through extrapolating cross-sectional slices was used to reconstruct the diaphysis of 15 equine third metacarpal bones (MC3). The reconstructed geometry was aligned with, and compared against, computed tomography-based models (reference models) of these bones and then the error map of the generated surfaces was plotted. The minimum error of reconstructed geometry was found to be +0.135 mm and -0.185 mm (0.407 mm ± 0.235, P > .05 and -0.563 mm ± 0.369, P > .05 for outside [convex] and inside [concave] surface position, respectively). Minor reconstructed surface error was observed on the dorsal cortex (0.216 mm ± 0.07, P > .05) for the outside surface and -0.185 mm ± 0.13, P > .05 for the inside surface. In addition, a displacement-based error estimation was used on 10 MC3 to identify poorly shaped elements in FEA, and the relations of finite element convergence analysis were used to present a framework for minimizing stress and strain errors in FEA. Finite element analysis errors of 3%-5% provided in the literature are unfortunate. Our proposed model, which presents an accurate FEA (error of 0.12%) in the smallest number of iterations possible, will assist future investigators to maximize FEA accuracy without the current runtime penalty.