Search

OBJECTIVES: To support future developments of field-based biomechanical load monitoring tools, this study aimed to identify generalised segmental acceleration patterns and their contribution to ground reaction forces (GRFs) across different running tasks. DESIGN: Exploratory experimental design. METHODS: A multivariate principal component analysis (PCA) was applied to a combination of segmental acceleration data from all body segments for 15 team-sport athletes performing accelerated, decelerated and constant low-, moderate- and high-speed running, and 90° cutting trials. Segmental acceleration profiles were then reconstructed from each principal component (PC) and used to calculate their specific GRF contributions. RESULTS: The first PC explained 48.57% of the acceleration variability for all body segments and was primarily related to the between-task differences in the overall magnitude of the GRF impulse. Magnitude and timing of high-frequency acceleration and GRF features (i.e. impact related characteristics) were primarily explained by the second PC (12.43%) and also revealed important between-task differences. The most important GRF characteristics were explained by the first five PCs, while PCs beyond that primarily contained small contributions to the overall GRF impulse. CONCLUSIONS: These findings show that a multivariate PCA approach can reveal generalised acceleration patterns and specific segmental contributions to GRF features, but their relative importance for different running activities are task dependent. Using segmental acceleration to assess whole-body biomechanical loading generically across various movements may thus require task identification algorithms and/or advanced sensor or data fusion approaches. ispartof: JOURNAL OF SCIENCE AND MEDICINE IN SPORT vol:22 issue:12 pages:1355-1360 ispartof: location:Australia status: published

Machine learning-based systems are now part of a wide array of real-world applications seamlessly embedded in the social realm. In the wake of this realisation, strict legal regulations for these systems are currently being developed, addressing some of the risks they may pose. This is the coming of age of the interpretability and explainability problems in machine learning-based data analysis, which can no longer be seen just as an academic research problem. In this tutorial, associated to ESANN 2021 special session on “Interpretable Models in Machine Learning and Explainable Artificial Intelligence”, we discuss explainable and interpretable machine learning as post-hoc and ante-hoc strategies to address these problems and highlight several aspects related to them, including their assessment. The contributions accepted for the session are then presented in this context

Background In-vivo single voxel proton magnetic resonance spectroscopy (SV 1H-MRS), coupled with supervised pattern recognition (PR) methods, has been widely used in clinical studies of discrimination of brain tumour types and follow-up of patients bearing abnormal brain masses. SV 1H-MRS provides useful biochemical information about the metabolic state of tumours and can be performed at short (< 45 ms) or long (> 45 ms) echo time (TE), each with particular advantages. Short-TE spectra are more adequate for detecting lipids, while the long-TE provides a much flatter signal baseline in between peaks but also negative signals for metabolites such as lactate. Both, lipids and lactate, are respectively indicative of specific metabolic processes taking place. Ideally, the information provided by both TE should be of use for clinical purposes. In this study, we characterise the performance of a range of Non-negative Matrix Factorisation (NMF) methods in two respects: first, to derive sources correlated with the mean spectra of known tissue types (tumours and normal tissue); second, taking the best performing NMF method for source separation, we compare its accuracy for class assignment when using the mixing matrix directly as a basis for classification, as against using the method for dimensionality reduction (DR). For this, we used SV 1H-MRS data with positive and negative peaks, from a widely tested SV 1H-MRS human brain tumour database. Results The results reported in this paper reveal the advantage of using a recently described variant of NMF, namely Convex-NMF, as an unsupervised method of source extraction from SV1H-MRS. Most of the sources extracted in our experiments closely correspond to the mean spectra of some of the analysed tumour types. This similarity allows accurate diagnostic predictions to be made both in fully unsupervised mode and using Convex-NMF as a DR step previous to standard supervised classification. The obtained results are comparable to, or more accurate than those obtained with supervised techniques. Conclusions The unsupervised properties of Convex-NMF place this approach one step ahead of classical label-requiring supervised methods for the discrimination of brain tumour types, as it accounts for their increasingly recognised molecular subtype heterogeneity. The application of Convex-NMF in computer assisted decision support systems is expected to facilitate further improvements in the uptake of MRS-derived information by clinicians.

The colour bimodality of galaxies provides an empirical basis for theories of galaxy evolution. However, the balance of processes that begets this bimodality has not yet been constrained. A more detailed view of the galaxy population is needed, which we achieve in this paper by using unsupervised machine learning to combine multi-dimensional data at two different epochs. We aim to understand the cosmic evolution of galaxy subpopulations by uncovering substructures within the colour bimodality. We choose a clustering algorithm that models clusters using only the most discriminative data available, and apply it to two galaxy samples: one from the second edition of the GALEX-SDSS-WISE Legacy Catalogue (GSWLC-2; $z \sim 0.06$), and the other from the VIMOS Public Extragalactic Redshift Survey (VIPERS; $z \sim 0.65$). We cluster within a nine-dimensional feature space defined purely by rest-frame ultraviolet-through-near-infrared colours. Both samples are similarly partitioned into seven clusters, breaking down into four of mostly star-forming galaxies (including the vast majority of green valley galaxies) and three of mostly passive galaxies. The separation between these two families of clusters suggests differences in the evolution of their galaxies, and that these differences are strongly expressed in their colours alone. The samples are closely related, with star-forming/green-valley clusters at both epochs forming morphological sequences, capturing the gradual internally-driven growth of galaxy bulges. At high stellar masses, this growth is linked with quenching. However, it is only in our low-redshift sample that additional, environmental processes appear to be involved in the evolution of low-mass passive galaxies., Comment: Accepted for publication in MNRAS. 22 pages, 15 figures

In this paper we propose to use partial responses derived from an initial multilayer perceptron (MLP) to build an explanatory risk prediction model of in-hospital mortality in intensive care units (ICU). Traditionally, MLPs deliver higher performance than linear models such as multivariate logistic regression (MLR). However, MLPs interlink input variables in such a complex way that is not straightforward to explain how the outcome is influenced by inputs and/or input interactions. In this paper, we hypothesized that in some scenarios, such as when the data noise is significant or when the data is just marginally non-linear, we could find slightly more complex associations by obtaining MLP partial responses. That is, by letting change one variable at the time, while keeping constant the rest. Overall, we found that, although the MLR and MLP in-hospital mortality model performances were equivalent, the MLP could explain non-linear associations that otherwise the MLR had considered non-significant. We considered that, although deeming higher-other interactions as disposable noise could be a strong assumption, building explanatory models based on the MLP partial responses could still be more informative than on MLR.

Differences in the protein composition of fast- and slow-twitch muscle may be maintained by different rates of protein turnover. We investigated protein turnover rates in slow-twitch soleus and fast-twitch plantaris of male Wistar rats (body weight 412 &plusmn, 69 g). Animals were assigned to four groups (n = 3, in each), including a control group (0 d) and three groups that received deuterium oxide (D2O) for either 10 days, 20 days or 30 days. D2O administration was initiated by an intraperitoneal injection of 20 &mu, L of 99% D2O-saline per g body weight, and maintained by provision of 4% (v/v) D2O in the drinking water available ad libitum. Soluble proteins from harvested muscles were analysed by liquid chromatography&ndash, tandem mass spectrometry and identified against the SwissProt database. The enrichment of D2O and rate constant (k) of protein synthesis was calculated from the abundance of peptide mass isotopomers. The fractional synthesis rate (FSR) of 44 proteins in soleus and 34 proteins in plantaris spanned from 0.58%/day (CO1A1: Collagen alpha-1 chain) to 5.40%/day NDRG2 (N-myc downstream-regulated gene 2 protein). Eight out of 18 proteins identified in both muscles had a different FSR in soleus than in plantaris (p &lt, 0.05).

Prediction of ground reaction force (GRF) magnitudes during running-based sports has several important applications, including optimal load prescription and injury prevention in athletes. Existing methods typically require information from multiple body-worn sensors, limiting their ecological validity, or aim to estimate discrete force parameters, limiting their ability to assess overall biomechanical load. This paper presents a neural network method to predict GRF time series from a single, commonly used, trunk-mounted accelerometer. The presented method uses a principal component analysis and multilayer perceptron (MLP) to obtain predictions. Time-series r2 coefficients with test data averaged around 0.9 for each impact, comparing favourably with alternative approaches which require additional sensors. For the impact peak, r2 was 0.74 across activities, comparing favourably with correlation analysis approaches. Several modifications, such as subject-specific training of the MLP, may help to improve results further, but the presented method can accurately predict GRF from trunk accelerometry data without requiring additional information. Results demonstrate the scope of machine learning to exploit common wearable technologies to estimate GRF in sport-specific environments. ispartof: MEDICAL ENGINEERING & PHYSICS vol:78 pages:82-89 ispartof: location:England status: published

Muscle adaptations to exercise are underpinned by alterations to the abundance of individual proteins, which may occur through a change either to the synthesis or degradation of each protein. We used deuterium oxide (2 H2 O) labeling and chronic low-frequency stimulation (CLFS) in vivo to investigate the synthesis, abundance, and degradation of individual proteins during exercise-induced muscle adaptation. Independent groups of rats received CLFS (10 Hz, 24 h/d) and 2 H2 O for 0, 10, 20, or 30 days. The extensor digitorum longus (EDL) was isolated from stimulated (Stim) and contralateral non-stimulated (Ctrl) legs. Proteomic analysis encompassed 38 myofibrillar and 46 soluble proteins and the rates of change in abundance, synthesis, and degradation were reported in absolute (ng/d) units. Overall, synthesis and degradation made equal contributions to the adaptation of the proteome, including instances where a decrease in protein-specific degradation primarily accounted for the increase in abundance of the protein.

This paper discusses the concepts of interpretability and explainability and outlines desiderata for robust interpretability. It then describes a neural network model that meets all criteria, with the addition of global faithfulness.

The aim of the paper is two-fold. First, we show that structure finding with the PC algorithm can be inherently unstable and requires further operational constraints in order to consistently obtain models that are faithful to the data. We propose a methodology to stabilise the structure finding process, minimising both false positive and false negative error rates. This is demonstrated with synthetic data. Second, to apply the proposed structure finding methodology to a data set comprising single-voxel Magnetic Resonance Spectra of normal brain and three classes of brain tumours, to elucidate the associations between brain tumour types and a range of observed metabolites that are known to be relevant for their characterisation. The data set is bootstrapped in order to maximise the robustness of feature selection for nominated target variables. Specifically, Conditional Independence maps (CI-maps) built from the data and their derived Bayesian networks have been used. A Directed Acyclic Graph (DAG) is built from CI-maps, being a major challenge the minimization of errors in the graph structure. This work presents empirical evidence on how to reduce false positive errors via the False Discovery Rate, and how to identify appropriate parameter settings to improve the False Negative Reduction. In addition, several node ordering policies are investigated that transform the graph into a DAG. The obtained results show that ordering nodes by strength of mutual information can recover a representative DAG in a reasonable time, although a more accurate graph can be recovered using a random order of samples at the expense of increasing the computation time.

Neural networks are frequently applied to medical data. We describe how complex and imbalanced data can be modelled with simple but accurate neural networks that are transparent to the user. In the case of a data set on cervical cancer with 753 observations excluding, missing values, and 32 covariates, with a prevalence of 73 cases (9.69%), we explain how model selection can be applied to the Multi-Layer Perceptron (MLP) by deriving a representation using a General Additive Neural Network.

Purpose:To investigate the relationship between whole-body accelerations and body-worn accelerometry during team-sport movements.Methods:Twenty male team-sport players performed forward running and anticipated 45° and 90° side-cuts at approach speeds of 2, 3, 4, and 5 m/s. Whole-body center-of-mass (CoM) accelerations were determined from ground-reaction forces collected from 1 foot–ground contact, and segmental accelerations were measured from a commercial GPS accelerometer unit on the upper trunk. Three higher-specification accelerometers were also positioned on the GPS unit, the dorsal aspect of the pelvis, and the shaft of the tibia. Associations between mechanical load variables (peak acceleration, loading rate, and impulse) calculated from both CoM accelerations and segmental accelerations were explored using regression analysis. In addition, 1-dimensional statistical parametric mapping (SPM) was used to explore the relationships between peak segmental accelerations and CoM-acceleration profiles during the whole foot–ground contact.Results:A weak relationship was observed for the investigated mechanical load variables regardless of accelerometer location and task (R2 values across accelerometer locations and tasks: peak acceleration .08–.55, loading rate .27–.59, and impulse .02–.59). Segmental accelerations generally overestimated whole-body mechanical load. SPM analysis showed that peak segmental accelerations were mostly related to CoM accelerations during the first 40–50% of contact phase.Conclusions:While body-worn accelerometry correlates to whole-body loading in team-sport movements and can reveal useful estimates concerning loading, these correlations are not strong. Body-worn accelerometry should therefore be used with caution to monitor whole-body mechanical loading in the field.

Background Stair falls are a major health problem for older people, but presently, there are no specific screening tools for stair fall prediction. The purpose of the present study was to investigate whether stair fallers could be differentiated from nonfallers by biomechanical risk factors or physical/psychological parameters and to establish the biomechanical stepping profile posing the greatest risk for a stair fall. Methods Eighty-seven older adults (age: 72.1 ± 5.2 years) negotiated an instrumented seven-step staircase and performed a range of physical/psychological tasks. k-Means clustering was used to profile the overall stair negotiation behavior with biomechanical parameters indicative of fall risk as input. Falls and events of balance perturbation (combined “hazardous events”) were then monitored during a 12-month follow-up. Cox-regression analysis was performed to examine whether physical/psychological parameters or biomechanical outcome measures could predict future hazardous events. Kaplan–Meier survival curves were obtained to identify the stepping strategy posing a risk for a hazardous event. Results Physical/psychological parameters did not predict hazardous events and the commonly used Fall Risk Assessment Tool classified only 1/17 stair fallers at risk for a fall. Single biomechanical risk factors could not predict hazardous events on stairs either. On the contrary, two particular clusters identified by the stepping profiling method in stair ascent were linked with hazardous events. Conclusion This highlights the potential of the stepping profiling method to predict stair fall risk in older adults against the limited predictability of single-parameter approaches currently used as screening tools.

Epistasis is a progressive approach that complements the ‘common disease, common variant’ hypothesis that highlights the potential for connected networks of genetic variants collaborating to produce a phenotypic expression. Epistasis is commonly performed as a pairwise or limitless-arity capacity that considers variant networks as either variant vs variant or as high order interactions. This type of analysis extends the number of tests that were previously performed in a standard approach such as Genome-Wide Association Study (GWAS), in which False Discovery Rate (FDR) is already an issue, therefore by multiplying the number of tests up to a factorial rate also increases the issue of FDR. Further to this, epistasis introduces its own limitations of computational complexity and intensity that are generated based on the analysis performed; to consider the most intense approach, a multivariate analysis introduces a time complexity of O(n!). Proposed in this paper is a novel methodology for the detection of epistasis using interpretable methods and best practice to outline interactions through filtering processes. Using a process of Random Sampling Regularisation which randomly splits and produces sample sets to conduct a voting system to regularise the significance and reliability of biological markers, SNPs. Preliminary results are promising, outlining a concise detection of interactions. Results for the detection of epistasis, in the classification of breast cancer patients, indicated eight outlined risk candidate interactions from five variants and a singular candidate variant with high protective association.

Background There are limited data comparing the consistency of angina reporting by patients and clinicians. Methods We performed a retrospective analysis of data from the randomised Stent or Surgery (SoS) trial. The trial required reporting of angina using the Canadian Cardiovascular Society (CCS) classification by both patients and clinicians at baseline and twelve months. We compared paired observations to describe the magnitude and direction of differences in clinician and patient reporting. The difference in CCS grade was expressed as the clinician minus patient value. We also examined the proportion of trial subjects reported as being free from angina (CCS = 0) in clinician and patient reporting. Results Paired CCS data was available for 912 and 887 cases at baseline and 12 months respectively. At baseline, clinicians reported freedom from angina in a single case (1/912 = 0.1%) compared to 70/912 (7.7%) patients (Delta 7.6% 95% CI 5.8 to 9.3, P ≤0.001). At 12 months, the position was reversed, with clinicians reporting 639/887 (72%) angina free compared to 449/887 (50.6%) for patients (Delta −21.4 95% CI −17.1 to −25.8 P ≤ 0.001). For the reported CCS grade at follow-up, the weighted linear kappa for overall agreement was 0.312. Discordant reporting involved the clinician suggesting less angina rather than more (36% v 8% of cases). Conclusions These findings have implications for our perception of previous research which has, in the main, focussed on clinician reporting. This emphasises the importance of patient reporting and a need to better understand reasons for discordance.

Stair falls are a major health problem for older people. Most studies on identification of stair fall risk factors are limited to staircases set in given step dimensions. However, it remains unknown whether the conclusions drawn would still apply if the dimensions had been changed to represent more challenging or easier step dimensions encountered in domestic and public buildings. The purpose was to investigate whether the self-selected biomechanical stepping behaviours are maintained when the dimensions of a staircase are altered. Sixty-eight older adults (>65 years) negotiated a seven-step staircase set in two step dimensions (shallow staircase: rise 15 cm, going 28 cm; steep staircase: rise 20 cm, going 25 cm). Six biomechanical outcome measures indicative of stair fall risk were measured. K-means clustering profiled the overall stair-negotiating behaviour and cluster profiles were calculated. A Cramer's V measured the degree of association in membership between clusters. The cluster profiles revealed that the biomechanically risky and conservative factors that characterized the overall behaviour in the clusters did not differ for the majority of older adults between staircases for ascent and descent. A strong association of membership between the clusters on the shallow staircase and the steep staircase was found for stair ascent (Cramer's V: 0.412, p

© 2020, Springer Nature Switzerland AG. The aim of this paper is to build a computer based clinical decision support tool using a semi-supervised framework, the Fisher Information Network (FIN), for visualization of a set of mammographic images. The FIN organizes the images into a similarity network from which, for any new image, reference images that are closely related can be identified. This enables clinicians to review not just the reference images but also ancillary information e.g. about response to therapy. The Fisher information metric defines a Riemannian space where distances reflect similarity with respect to a given probability distribution. This metric is informed about generative properties of data, and hence assesses the importance of directions in space of parameters. It automatically performs feature relevance detection. This approach focusses on the interpretability of the model from the standpoint of the clinical user. Model predictions were validated using the prevalence of classes in each of the clusters identified by the FIN.

Running impact forces expose the body to biomechanical loads leading to beneficial adaptations, but also risk of injury. High-intensity running tasks, especially, are deemed highly demanding for the musculoskeletal system, but loads experienced during these actions are not well understood. To eventually predict GRF and understand the biomechanical loads experienced during such activities in greater detail, this study aimed to (1) examine the feasibility of using a simple two mass-spring-damper model, based on eight model parameters, to reproduce ground reaction forces (GRFs) for high-intensity running tasks and (2) verify whether the required model parameters were physically meaningful. This model was used to reproduce GRFs for rapid accelerations and decelerations, constant speed running and maximal sprints. GRF profiles and impulses could be reproduced with low to very low errors across tasks, but subtler loading characteristics (impact peaks, loading rate) were modelled less accurately. Moreover, required model parameters varied strongly between trials and had minimal physical meaning. These results show that although a two mass-spring-damper model can be used to reproduce overall GRFs for high-intensity running tasks, the application of this simple model for predicting GRFs in the field and/or understanding the biomechanical demands of training in greater detail is likely limited. ispartof: SPORTS BIOMECHANICS vol:20 issue:5 ispartof: location:England status: published

This case study benchmarks a range of statistical and machine learning methods relevant to computer-based decision support in clinical medicine, focusing on the diagnosis of knee osteoarthritis at first presentation. The methods, comprising logistic regression, Multilayer Perceptron (MLP), Chi-square Automatic Interaction Detector (CHAID) and Classification and Regression Trees (CART), are applied to a public domain database, the Osteoarthritis Initiative (OAI), a 10 year longitudinal study starting in 2002 (n = 4,796). In this real-world application, it is shown that logistic regression is comparable with the neural networks and decision trees for discrimination of positive diagnosis on this data set. This is likely because of weak non-linearities among high levels of noise. After comparing the explanations provided by the different methods, it is concluded that the interpretability of the risk score index provided by logistic regression is expressed in a form that most naturally integrates with clinical reasoning. The reason for this is that it gives a statistical assessment of the weight of evidence for making the diagnosis, so providing a direction for future research to improve explanation of generic non-linear models.