Large Datasets Research Articles

Transforming environmental health datasets from the comparative toxicogenomics database into chord diagrams to visualize molecular mechanisms

In environmental health, the specific molecular mechanisms connecting a chemical exposure to an adverse endpoint are often unknown, reflecting knowledge gaps. At the public Comparative Toxicogenomics Database (CTD; https://ctdbase.org/), we integrate manually curated, literature-based interactions from CTD to compute four-unit blocks of information organized as a potential step-wise molecular mechanism, known as “CGPD-tetramers,” wherein a chemical interacts with a gene product to trigger a phenotype which can be linked to a disease. These computationally derived datasets can be used to fill the gaps and offer testable mechanistic information. Users can generate CGPD-tetramers for any combination of chemical, gene, phenotype, and/or disease of interest at CTD; however, such queries typically result in the generation of thousands of CGPD-tetramers. Here, we describe a novel approach to transform these large datasets into user-friendly chord diagrams using R. This visualization process is straightforward, simple to implement, and accessible to inexperienced users that have never used R before. Combining CGPD-tetramers into a single chord diagram helps identify potential key chemicals, genes, phenotypes, and diseases. This visualization allows users to more readily analyze computational datasets that can fill the exposure knowledge gaps in the environmental health continuum.

The Application of the Piecewise Linear Method for Non-Linear Programming Problems in Ride-Hailing Assignment Based on Service Level, Driver Workload, and Fuel Consumption

Ride-hailing services have grown rapidly, presenting challenges such as increased traffic congestion, inefficient driver workload distribution, and environmental concerns like higher fuel consumption and emissions. This study develops a non-linear ride-hailing assignment model addressing these issues by considering service level, driver workload, and fuel consumption. A piecewise linear method was employed to handle a non-linear programming model, and the method was modified to function autonomously without operator intervention. The model’s performance was evaluated using a publicly accessible dataset of taxi trips in Manhattan, focusing on indicators such as passenger waiting time, driver workload distribution, and fuel consumption. Numerical simulations demonstrated significant improvements: a 15% reduction in average passenger waiting time, a 20% improvement in balancing driver workloads, and a 10% decrease in overall fuel consumption, contributing to reduced emissions and environmental impact. The modified piecewise linear method proved effective in optimizing ride-hailing assignments, providing a more efficient and sustainable solution. The model also showed robustness in handling large datasets, ensuring scalability and applicability to various urban settings. These findings highlight the model’s potential to enhance operational efficiency and promote sustainability in ride-hailing services. By integrating considerations for service level, driver workload, and fuel consumption, the model offers a holistic approach to addressing the key challenges faced by the ride-hailing industry. This study provides valuable insights for future ride-hailing development and implementations of ride-hailing systems, promoting practices that are both efficient and environmentally friendly.

Development Modeling of Phormia regina (Diptera: Calliphoridae)

A series of experiments were conducted on Phormia regina, a forensically important blow fly species, that met the requirements needed to create statistically valid development models. Experiments were conducted over 11 temperatures (7.5 to 32.5 °C, at 2.5 °C intervals) with a 16:8 L:D cycle. Experimental units contained 20 eggs, 10 g of beef liver, and 2.5 cm of sand. Each life stage (egg to adult) had five sampling times. Each sampling time was replicated four times for a total of 20 measurements per life stage. For each sampling time, the cups were pulled from the chambers, and the stage of each maggot was documented morphologically through posterior spiracular slits and cephalopharyngeal development. Data were normally distributed with the later larval (L3m) and pupation stages having the most variation within and transitioning between stages, particularly between 12.5 °C and 20.0 °C. The biological minimum was between 10.0 °C and 12.5 °C, with little egg development and no egg emergence at 7.5 °C and no maturation past L1 at 10.0 °C. Phormia regina did not display increased mortality associated with the upper temperature of 32.5 °C. The development data generated illustrate the advantages of large data sets in modeling blow fly development and the need for curvilinear models in describing development at environmental temperatures near the biological minima and maxima.

International survey on invasive lobular breast cancer identifies priority research questions

There is growing awareness of the unique etiology, biology, and clinical presentation of invasive lobular breast cancer (ILC), but additional research is needed to ensure translation of findings into management and treatment guidelines. We conducted a survey with input from breast cancer physicians, laboratory-based researchers, and patients to analyze the current understanding of ILC, and identify consensus research questions. 1774 participants from 66 countries respondents self-identified as clinicians (N = 413), researchers (N = 376), and breast cancer patients and advocates (N = 1120), with some belonging to more than one category. The majority of physicians reported being very/extremely (41%) to moderately (42%) confident in describing the differences between ILC and invasive breast cancer of no special type (NST). Knowledge of histology was seen as important (73%) and as affecting treatment decisions (51%), and most agreed that refining treatment guidelines would be valuable (76%). 85% of clinicians have never powered a clinical trial to allow subset analysis for histological subtypes, but the majority would consider it, and would participate in an ILC clinical trials consortium. The majority of laboratory researchers, reported being and very/extremely (48%) to moderately (29%) confident in describing differences between ILC and NST. They reported that ILCs are inadequately presented in large genomic data sets, and that ILC models are insufficient. The majority have adequate access to tissue or blood from patients with ILC. The majority of patients and advocates (52%) thought that their health care providers did not sufficiently explain the unique features of ILC. They identified improvement of ILC screening/early detection, and identification of better imaging tools as top research priorities. In contrast, both researchers and clinicians identified understanding of endocrine resistance and identifying novel drugs that can be tested in clinical trials as top research priority. In summary, we have gathered information from an international community of physicians, researchers, and patients/advocates that we expect will lay the foundation for a community-informed collaborative research agenda, with the goal of improving management and personalizing treatment for patients with ILC.

Bibliometric Analysis of Remote Sensing over Marine Areas for Sustainable Development: Global Trends and Worldwide Collaboration

More than two-thirds of the Earth’s surface is covered by oceans and yet only a small portion of these oceans has been directly explored in detail, highlighting the need for powerful tools like remote sensing (RS) technology to bridge this gap. International frameworks, the 2030 Agenda for Sustainable Development, and Ocean Decade point out the significance of marine areas for achieving sustainable growth. This study conducts a bibliometric analysis of RS over marine areas for sustainable development to identify key contributors, collaboration networks, and evolving research themes from the beginning of the 21st century until last year. Using the Web of Science Core Collection database, 499 relevant articles published between 2000 and 2023 were identified. The bibliometric analysis showed a significant increase in scientific productivity related to the field. On an international level, China emerges as the most productive country, but international collaboration has played a crucial role, with 36.87% of articles resulting from international co-authorship, pointing to the global nature of research in this field. RS technology has continuously evolved from airborne sensors to the augmentation of Earth Observation missions. Our findings reveal a shift towards automated analysis and processing of RS data using machine learning techniques to integrate large datasets and develop robust scientific solutions.

Learning from models: high-dimensional analyses on the performance of machine learning interatomic potentials

Machine learning interatomic potential (MLIP) has been widely adopted for atomistic simulations. While errors and discrepancies for MLIPs have been reported, a comprehensive examination of the MLIPs’ performance over a broad spectrum of material properties has been lacking. This study introduces an analysis process comprising model sampling, benchmarking, error evaluations, and multi-dimensional statistical analyses on an ensemble of MLIPs for prediction errors over a diverse range of properties. By carrying out this analysis on 2300 MLIP models based on six different MLIP types, several properties that pose challenges for the MLIPs to achieve small errors are identified. The Pareto front analyses on two or more properties reveal the trade-offs in different properties of MLIPs, underscoring the difficulties of achieving low errors for a large number of properties simultaneously. Furthermore, we propose correlation graph analyses to characterize the error performances of MLIPs and to select the representative properties for predicting other property errors. This analysis process on a large dataset of MLIP models sheds light on the underlying complexities of MLIP performance, offering crucial guidance for the future development of MLIPs with improved predictive accuracy across an array of material properties.

An embedded end-to-end voice assistant

Voice assistants are spreading in various environments, such as houses and cars, bringing the possibility of controlling heterogeneous Internet of Things devices with simple voice commands. However, massive use of the cloud connection for speech processing requires an efficient and robust Internet connection and raises concerns in terms of privacy. Therefore, we propose an end-to-end solution able to work totally offline, based on a system architecture combining different Deep Learning models to implement all the steps of the speech elaboration process. Being interested in targeting the Italian language, we exploited the transfer learning paradigm, which allows leveraging models trained in English on large datasets and fine-tuning them to the target language on a smaller dataset. The proposed system architecture is configurable and easily extensible to other languages. Experimental results in an automotive application use case show that our solution outperforms the other embedded models and achieves performance comparable to state-of-the-art cloud-connected solutions for Automatic Speech Recognition. Moreover, overall latency is significantly reduced by eliminating the need to connect to the cloud.

Classification of Breast Cancer with Transfer Learning on Convolutional Neural Network Models

Breast cancer is a serious medical condition and a leading cause of death among women. Early and accurate diagnosis is crucial for improving patient outcomes. This study explores the use of Convolutional Neural Networks (CNNs) with Transfer Learning using DenseNet121 and ResNet50 models to enhance breast cancer classification via mammography. Transfer Learning enables CNN models to leverage knowledge learned from larger datasets such as ImageNet to improve performance on specific breast cancer datasets. The dataset comprised medical images with three breast variations: benign, malignant, and normal, totaling 531 data points. Data was split with a 70% training and 30% validation ratio. Two CNN models, AlexNet and ResNet50, were evaluated to compare their performance in classifying these breast cancer types. The experimental results show that AlexNet achieved a training accuracy of 98.01%, while ResNet50 achieved 64.07%. AlexNet demonstrated superior performance in identifying complex patterns in mammography images, resulting in more accurate classification of different breast cancer types. These findings highlight the potential of deep learning applications to support more precise and effective medical diagnostics for breast cancer. This research contributes significantly to the development of AI technologies in healthcare aimed at improving early detection of breast cancer. The implications of this study could expand our understanding of Transfer Learning applications in medical contexts, driving further advancements in this field to enhance patient care and prognosis

Evaluation of artificial intelligence-powered screening for sexually transmitted infections-related skin lesions using clinical images and metadata

BackgroundSexually transmitted infections (STIs) pose a significant global public health challenge. Early diagnosis and treatment reduce STI transmission, but rely on recognising symptoms and care-seeking behaviour of the individual. Digital health software that distinguishes STI skin conditions could improve health-seeking behaviour. We developed and evaluated a deep learning model to differentiate STIs from non-STIs based on clinical images and symptoms.MethodsWe used 4913 clinical images of genital lesions and metadata from the Melbourne Sexual Health Centre collected during 2010–2023. We developed two binary classification models to distinguish STIs from non-STIs: (1) a convolutional neural network (CNN) using images only and (2) an integrated model combining both CNN and fully connected neural network (FCN) using images and metadata. We evaluated the model performance by the area under the ROC curve (AUC) and assessed metadata contributions to the Image-only model.ResultsOur study included 1583 STI and 3330 non-STI images. Common STI diagnoses were syphilis (34.6%), genital warts (24.5%) and herpes (19.4%), while most non-STIs (80.3%) were conditions such as dermatitis, lichen sclerosis and balanitis. In both STI and non-STI groups, the most frequently observed groups were 25–34 years (48.6% and 38.2%, respectively) and heterosexual males (60.3% and 45.9%, respectively). The Image-only model showed a reasonable performance with an AUC of 0.859 (SD 0.013). The Image + Metadata model achieved a significantly higher AUC of 0.893 (SD 0.018) compared to the Image-only model (p < 0.01). Out of 21 metadata, the integration of demographic and dermatological metadata led to the most significant improvement in model performance, increasing AUC by 6.7% compared to the baseline Image-only model.ConclusionsThe Image + Metadata model outperformed the Image-only model in distinguishing STIs from other skin conditions. Using it as a screening tool in a clinical setting may require further development and evaluation with larger datasets.

Friedmannodendron rodriguesiana, a new segregate of Dombeya Cav. (Dombeyoideae, Malvaceae), endemic to Rodrigues

ABSTRACT Dombeya rodriguesiana F.Friedmann is an endemic species to Rodrigues (Mascarene Archipelago, Indian Ocean). This species faces severe threats and is currently represented by at least two living individuals. We sequenced a second sample of this species to confirm that its phylogenetic placement renders Dombeya Cav. paraphyletic and, thus, its placement in Dombeya needs revision. The second sample was extracted from a cultivated clone of the last wild individual of D. rodriguesiana. Sequences from this individual for six markers (one nuclear, five plastid) were analyzed using the largest molecular dataset available on Dombeyoideae and Bayesian inference and maximum likelihood methods. Study of all known herbarium material of D. rodriguesiana, and other relevant material, allowed for a new morphological evaluation and description. Our results demonstrate that D. rodriguesiana is closely related to the Malagasy endemic genus, Andringitra Skema. We found that the unique combination of characters of a persistent epicalyx with a persistent perianth provides the synapomorphy for these two sister taxa, independent of the significant morphological differences between them (e.g. flower size, number, arrangement, sex expression). Furthermore, we describe useful field characters to differentiate D. rodriguesiana from the superficially similar D. ferruginea Cav. To maintain a meaningful classification that reflects monophyletic groups in the Dombeyoideae, we propose the establishment of a new genus and combination to accommodate this rare species.

Self-supervised representations and node embedding graph neural networks for accurate and multi-scale analysis of materials

Supervised machine learning algorithms, such as graph neural networks (GNN), have successfully predicted material properties. However, the superior performance of GNN usually relies on end-to-end learning on large material datasets, which may lose the physical insight of multi-scale information about materials. And the process of labeling data consumes many resources and inevitably introduces errors, which constrains the accuracy of prediction. We propose to train the GNN model by self-supervised learning on the node and edge information of the crystal graph. Compared with the popular manually constructed material descriptors, the self-supervised atomic representation can reach better prediction performance on material properties. Furthermore, it may provide physical insights by tuning the range information. Applying the self-supervised atomic representation on the magnetic moment datasets, we show how they can extract rules and information from the magnetic materials. To incorporate rich physical information into the GNN model, we develop the node embedding graph neural networks (NEGNN) framework and show significant improvements in the prediction performance. The self-supervised material representation and the NEGNN framework may investigate in-depth information from materials and can be applied to small datasets with increased prediction accuracy.

Not only domestic spaces: dismantling short‐term occupations in Level 497D of Cova Gran de Santa Linya (Pre‐Pyrenees, Spain)

ABSTRACTThe Early Upper Paleolithic Level 497D of Cova Gran (Pre‐Pyrenees, Spain) comprises large assemblages of lithics, fauna, hearths, ash accumulations and well‐preserved refitting sequences. This provides exceptional information to analyze spatial patterns and deepen our understanding of the socio‐economic behavior of these human groups. This large interdisciplinary dataset has allowed us to carry out a detailed spatial study based on density analysis, geostatistics, fabric analysis and orientation techniques, revealing a structuring of the activities performed within the rock shelter. The distribution and accumulation of lithic artifacts, fauna and refits, and their association with the 10 hearths and ash accumulations, show that different types of activities were developed at the site, as well as showing the different uses of the hearths. 497D is a palimpsest where several short‐term occupations and/or activities could have occurred in a relatively short period of time, avoiding prolonged exposure to biotic and abiotic post‐depositional factors that could have significantly disturbed this well‐preserved assemblage. This level shows a place visited several times in relation to the development of specific activities, with movements throughout the landscape for the exploitation of local and regional resources, and less so its use as a domestic space. This implies that the central settlement, or dwelling space, would have been located somewhere other than Cova Gran.

Refocus the Attention for Parameter-Efficient Thermal Infrared Object Tracking.

Introducing deep trackers to thermal infrared (TIR) tracking is hampered by the scarcity of large training datasets. To alleviate the predicament, a common approach is full fine-tuning (FFT) based on pretrained RGB parameters. Nevertheless, due to its inefficient training pattern and representation collapse risk, some parameter-efficient fine-tuning (PEFT) alternatives have been promoted recently. However, the existing PEFT algorithms typically follow a bottom-up way, where their attention solely relies on the input and lacks the capability of task-guided top-down attention, which provides the task-relevant representation such as the human visual perception system. In this article, we introduce ReFocus, a new PEFT method that adapts the pretrained RGB foundation tracking model to the downstream TIR tracking task through the guidance of high-level task-specific signals in a top-down attention manner. By freezing the entire foundation model and only training query-guided feature selection and top-down blocks, ReFocus achieves state-of-the-art (SOTA) TIR tracking performance while keeping training efficiency. Extensive experiments on five TIR tracking benchmarks demonstrate that ReFocus significantly improves the performance of the foundation tracker. Besides, further ablation studies show the effectiveness and flexible adaptability of the proposed method to lighter foundation models and different tracking frameworks. Compared to FFT and other bottom-up PEFT paradigms, such as head probe, low-rank adaptation (LoRA), and adapter, our method achieves comparable or superior performance with fewer training parameters and reveals the advantage of learning stability.

Whistlerlib: a distributed computing library for exploratory data analysis on large social network datasets

Whistlerlib: a distributed computing library for exploratory data analysis on large social network datasets

The moderating role of ethical awareness in the relationship between nurses’ artificial intelligence perceptions, attitudes, and innovative work behavior: a cross-sectional study

BackgroundArtificial intelligence is rapidly advancing and being integrated into healthcare, potentially revolutionizing patient care and improving outcomes by leveraging large datasets and complex algorithms.AimInvestigate the moderating role of ethical awareness between nurses’ artificial intelligence perceptions, attitudes, and innovative work behaviors.Design and MethodsA cross-sectional descriptive correlational design adhering to STROBE guidelines. A non-probability convenience sample of 415 Alexandria Main University Hospital nurses was analyzed. Statistical methods included one-way ANOVA, the student t-test, and the Pearson coefficient, with results evaluated for significance at the 5% level and internal consistency assessed via Cronbach’s α. Linear regression assessed the predicting and moderating effect between ethical awareness, nurses’ artificial intelligence perceptions, attitudes, and innovative work behavior. The perceptions of using the Artificial Intelligence Scale, general attitudes towards the Artificial Intelligence Scale, ethical awareness of Using Artificial Intelligence, and the Employee Innovative Behavior Scale were used to respond to the research aim.ResultsThe study revealed that perception of AI use among nurses has a mean score of 50.25 (SD = 3.49), attitudes towards AI have a mean score of 71.40 (SD = 4.98), ethical awareness regarding AI use shows a mean score of 43.85 (SD = 3.39), and nurses innovative behavior exhibits a mean score of 83.63 (SD = 5.22). Attitude and ethical awareness were statistically significant predictors of innovation. Specifically, for every one-unit increase in attitude, innovative work behaviors increase by 1.796 units (p = 0.001), and for every one-unit increase in ethical awareness, innovative work behaviors increase by 2.567 units (p = 0.013). The interaction effects between perception, ethical awareness, attitude, and ethical awareness were also examined. Only the interaction between attitude and ethical awareness was found to be significant (p = 0.002), suggesting that the effect of attitude on innovative work behaviors depends on the level of ethical awareness. In other words, ethical awareness moderates the relationship between attitudes and innovative work behaviors rather than perception and innovation.ConclusionThere is a statistically significant correlation between attitude, ethical awareness, and creativity, highlighting that ethical awareness moderates the relationship between attitudes and innovative work behaviors. These findings emphasize the importance of ethical awareness in fostering positive attitudes towards AI and enhancing innovative practices in nursing, ultimately contributing to nurses’ well-being.

What does “palliative care” represent in research using secondary data?

While much research has been done regarding “palliative care” using retrospective cohort studies of large datasets, the data sources may not be capturing specialty palliative care services. This article aims to clarify what source data are used in such studies, how specialty palliative care services are determined to have been provided or not, and mismatches between the nature of the data and the interpretation of researchers. Major US data sources that are examined include cancer registries such as the National Cancer Database; health systems’ internal data; state and nation-level hospital admissions data; and claims data from Medicare and commercial payers. Problematic studies are common. Many used cancer registry data and mischaracterized palliative intent for a given cancer treatment as “palliative care services.” Dozens relied on the diagnosis code for “encounter for palliative care” which lacks adequate validity for use in research. Researchers, peer-reviewers, and research consumers are cautioned about these potential pitfalls that lead to meaningless or misleading research papers. Suggestions are made regarding more rigorous methods and trustworthy data sources and additional research that can lead to consensus among researchers on these issues.

Novel Domain Knowledge-Encoding Algorithm Enables Label-Efficient Deep Learning for Cardiac CT Segmentation to Guide Atrial Fibrillation Treatment in a Pilot Dataset

Background: Segmenting computed tomography (CT) is crucial in various clinical applications, such as tailoring personalized cardiac ablation for managing cardiac arrhythmias. Automating segmentation through machine learning (ML) is hindered by the necessity for large, labeled training data, which can be challenging to obtain. This article proposes a novel approach for automated, robust labeling using domain knowledge to achieve high-performance segmentation by ML from a small training set. The approach, the domain knowledge-encoding (DOKEN) algorithm, reduces the reliance on large training datasets by encoding cardiac geometry while automatically labeling the training set. The method was validated in a hold-out dataset of CT results from an atrial fibrillation (AF) ablation study. Methods: The DOKEN algorithm parses left atrial (LA) structures, extracts “anatomical knowledge” by leveraging digital LA models (available publicly), and then applies this knowledge to achieve high ML segmentation performance with a small number of training samples. The DOKEN-labeled training set was used to train a nnU-Net deep neural network (DNN) model for segmenting cardiac CT in N = 20 patients. Subsequently, the method was tested in a hold-out set with N = 100 patients (five times larger than training set) who underwent AF ablation. Results: The DOKEN algorithm integrated with the nn-Unet model achieved high segmentation performance with few training samples, with a training to test ratio of 1:5. The Dice score of the DOKEN-enhanced model was 96.7% (IQR: 95.3% to 97.7%), with a median error in surface distance of boundaries of 1.51 mm (IQR: 0.72 to 3.12) and a mean centroid–boundary distance of 1.16 mm (95% CI: −4.57 to 6.89), similar to expert results (r = 0.99; p &lt; 0.001). In digital hearts, the novel DOKEN approach segmented the LA structures with a mean difference for the centroid–boundary distances of −0.27 mm (95% CI: −3.87 to 3.33; r = 0.99; p &lt; 0.0001). Conclusions: The proposed novel domain knowledge-encoding algorithm was able to perform the segmentation of six substructures of the LA, reducing the need for large training data sets. The combination of domain knowledge encoding and a machine learning approach could reduce the dependence of ML on large training datasets and could potentially be applied to AF ablation procedures and extended in the future to other imaging, 3D printing, and data science applications.

Development of a laboratory-based model to study the interaction between nutrients and Vibrio cholerae and predicting the spread of cholera outbreaks in the Indian subcontinent.

Cholera is an infectious disease that is transmitted through contaminated water. The disease includes a long back history of epidemics. Despite the numerous hygiene and prevention techniques that have been developed for Cholera, outbreaks of cholera are still reported worldwide. The resolution to this issue lies in promptly identifying the area susceptible to cholera outbreaks, a matter that continues to perplex scientists and medical professionals. It has been reported that Vibrio is effective in nitrogen digestion because it contains the nasA gene. In this study, initially theimpact of nutrients (nitrate and nitrite) on growth of Vibrio cholerae was determined, subsequently a relationship was developed between nutrient substrates and V. cholerae growth rate, using Monod model. Subsequently, the model was applied to large national river quality data set (2012-2014) developed by Central Pollution Control Board (CPCB) and a possible cholerae outbreak zone was predicted. This work will definitely help the policy makers to develop management strategy for keeping rivers safe from future cholera outbreak.

Learning together: Towards foundation models for machine learning interatomic potentials with meta-learning

The development of machine learning models has led to an abundance of datasets containing quantum mechanical (QM) calculations for molecular and material systems. However, traditional training methods for machine learning models are unable to leverage the plethora of data available as they require that each dataset be generated using the same QM method. Taking machine learning interatomic potentials (MLIPs) as an example, we show that meta-learning techniques, a recent advancement from the machine learning community, can be used to fit multiple levels of QM theory in the same training process. Meta-learning changes the training procedure to learn a representation that can be easily re-trained to new tasks with small amounts of data. We then demonstrate that meta-learning enables simultaneously training to multiple large organic molecule datasets. As a proof of concept, we examine the performance of a MLIP refit to a small drug-like molecule and show that pre-training potentials to multiple levels of theory with meta-learning improves performance. This difference in performance can be seen both in the reduced error and in the improved smoothness of the potential energy surface produced. We therefore show that meta-learning can utilize existing datasets with inconsistent QM levels of theory to produce models that are better at specializing to new datasets. This opens new routes for creating pre-trained, foundation models for interatomic potentials.

Finite element models with automatic computed tomography bone segmentation for failure load computation

Bone segmentation is an important step to perform biomechanical failure load simulations on in-vivo CT data of patients with bone metastasis, as it is a mandatory operation to obtain meshes needed for numerical simulations. Segmentation can be a tedious and time consuming task when done manually, and expert segmentations are subject to intra- and inter-operator variability. Deep learning methods are increasingly employed to automatically carry out image segmentation tasks. These networks usually need to be trained on a large image dataset along with the manual segmentations to maximize generalization to new images, but it is not always possible to have access to a multitude of CT-scans with the associated ground truth. It then becomes necessary to use training techniques to make the best use of the limited available data. In this paper, we propose a dedicated pipeline of preprocessing, deep learning based segmentation method and post-processing for in-vivo human femurs and vertebrae segmentation from CT-scans volumes. We experimented with three U-Net architectures and showed that out-of-the-box models enable automatic and high-quality volume segmentation if carefully trained. We compared the failure load simulation results obtained on femurs and vertebrae using either automatic or manual segmentations and studied the sensitivity of the simulations on small variations of the automatic segmentation. The failure loads obtained using automatic segmentations were comparable to those obtained using manual expert segmentations for all the femurs and vertebrae tested, demonstrating the effectiveness of the automated segmentation approach for failure load simulations.