Existing deep learning (DL) networks are primarily trained on adult datasets and may not always generalize to pediatric populations, where growth plays a major role. Here, we investigated improving semantic segmentation outcomes of pediatric hand phalanges from radiographs without relying on fully pediatric training datasets, which are scarce. First, alternative DL networks (FCN-8, FCN-32, U-Net, Inception U-Net, and DeepLabv3+) were trained with manually segmented radiographs of near-skeletally-mature (NSM) subjects and their performances were evaluated using mean intersection-over-union (Mean IoU) and multiclass Dice scores. DeepLabv3+ and Inception U-Net performed the best for NSM segmentation, with Mean IoU scores of 0.899 ± 0.035 and 0.887 ± 0.062, respectively. These networks were then used to investigate zero pediatric data (scaling-based data augmentation) and minimal pediatric data (incremental pediatric data substitution) approaches to improve age-domain generalizability. The minimal pediatric data approach proved effective, with a 20 % pediatric data inclusion leading to an up to 21.1 % increase in Mean IoU for pediatric subjects compared to networks trained exclusively on NSM subjects. Furthermore, no adverse effects of this approach were found when tested on NSM subjects, and there were even improvements in performance for Inception U-Net. To conclude, we highlight that networks utilizing multi-scale filters perform best for the semantic segmentation of hand phalanges. We further demonstrate that a minimal inclusion of pediatric training data can markedly improve age-domain generalizability for semantic segmentation tasks. This removes the difficult task of gathering large training datasets of pediatric subjects, which is often impractical, if not impossible.

Durable interfacing of hard and soft materials is a major design challenge caused by the ensuing stress concentrations. In nature, soft-hard interfaces exhibit remarkable mechanical performance, with failures rarely happening at the interface. Here, we mimic the strategies observed in nature to design efficient soft-hard interfaces. We base our geometrical designs on triply periodic minimal surfaces (i.e., Octo, Diamond, and Gyroid), collagen-like triple helices, and randomly distributed particles. A combination of computational simulations and experimental techniques, including uniaxial tensile and quad-lap shear tests, are used to characterize the mechanical performance of the interfaces. Our analyses suggest that smooth interdigitated connections, compliant gradient transitions, and either decreasing or constraining strain concentrations lead to simultaneously strong and tough interfaces. We generate additional interfaces where the abovementioned toughening mechanisms work synergistically to create soft-hard interfaces with strengths approaching the upper achievable limit and enhancing toughness values by 50%, as compared to the control group.

Living organisms use functional gradients (FGs) to interface hard and soft materials (e.g., bone and tendon), a strategy with engineering potential. Past attempts involving hard (or soft) phase ratio variation have led to mechanical property inaccuracies because of microscale-material macroscale-property nonlinearity. This study examines 3D-printed voxels from either hard or soft phase to decode this relationship. Combining micro/macroscale experiments and finite element simulations, a power law model emerges, linking voxel arrangement to composite properties. This model guides the creation of voxel-level FG structures, resulting in two biomimetic constructs mimicking specific bone-soft tissue interfaces with superior mechanical properties. Additionally, the model studies the FG influence on murine preosteoblast and human bone marrow-derived mesenchymal stromal cell (hBMSC) morphology and protein expression, driving rational design of soft-hard interfaces in biomedical applications.