The association of machine learning (ML) tools with the synthesis of nanoparticles has the potential to streamline the development of more efficient and effective nanomedicines. The continuous-flow synthesis of nanoparticles via microfluidics represents an ideal playground for ML tools, where multiple engineering parameters – flow rates and mixing configurations, type and concentrations of the reagents – contribute in a non-trivial fashion to determine the resultant morphological and pharmacological attributes of nanomedicines. Here we present the application of ML models towards the microfluidic-based synthesis of liposomes loaded with a model hydrophobic therapeutic agent, curcumin. After generating over 200 different liposome configurations by systematically modulating flow rates, lipid concentrations, organic:water mixing volume ratios, support-vector machine models and feed-forward artificial neural networks were trained to predict, respectively, the liposome dispersity/stability and size. This work presents an initial step towards the application and cultivation of ML models to instruct the microfluidic formulation of nanoparticles.

The roadmap to fusion electricity of the European scientific program aims at the realization of the future DEMOnstration (DEMO) fusion power plant. In 2020, the pre-concept design phase of DEMO was completed, defining the concept and characteristics of the main magnets and structures of the machine. Sixteen toroidal D-shaped magnets, six poloidal annular coils and a central solenoid constitute the functioning system core. The reactor is subjected to huge mechanical loads, mainly due to the Lorentz force produced by the combination of the high magnetic fields and operative currents. As a consequence, the loading conditions are extremely demanding for the structural components, and it is crucial to complete a comprehensive static and fatigue assessment before proceeding with the next design iteration. This work focuses on the electromagnetic and structural analyses performed on the toroidal field coil system and its support structures to present the methodological approach developed. Exploiting the finite element method, a three-dimensional model has been defined to obtain the electromagnetic loads on the main time points of the reference plasma scenario and then transfer them to a related 3D structural model, corresponding to the discretization of the electromagnetic one. The structural model was used to obtain the displacement and stress fields at the various time points to perform the mechanical evaluation as well as the fatigue assessment.

The highly infiltrating nature of glioma cells is the major cause for the poor prognosis of brain malignancies. Motility, proliferation, and gene expression of cells in natural and synthetic gels have been analyzed by several authors, yet quantitative studies elucidating the role of matrix porosity and rigidity in the development of whole malignant masses are missing. Here, an experimental-computational framework is introduced to analyze the behavior of U87-MG cells and spheroids in compact hyaluronic acid gels (HA), replicating the brain parenchyma; and fibrous collagen gels (COL), resembling the organized structures of the brain. Experimentally it was observed that individual U87-MG cells in COL assumed an elongated morphology within a few hours post inclusion (p.i.) and travelled longer distances than in HA. As spheroids, U87-MG cells rapidly dispersed into COL resulting in infiltrating regions as large as tumor cores (≈600 μm, at 8 days p.i.). Conversely, cells in HA originated smaller and denser infiltrating regions (≈300 μm, at 8 days p.i.). Notably, COL tumor core size was only 20% larger than in HA, at longer time points. Computationally, by introducing for the first time the effects of matrix heterogeneity in our numerical simulations, the results confirmed that matrix porosity and its spatial organization are key factors in priming the infiltrating potential of these malignant cells. The experimental-numerical synergy can be used to predict the behavior of neoplastic masses under diverse conditions and the efficacy of combination therapies simultaneously aiming at killing cancer cells and modulating the tumor microenvironment.

(1) Background: The realization of appropriate aortic replicas for in vitro experiments requires a suitable choice of both the material and geometry. The matching between the grade of details of the geometry and the mechanical response of the materials is an open issue that deserves attention. (2) Methods: To explore this issue, we performed a series of Fluid–Structure Interaction simulations, which compared the dynamics of three aortic models. Specifically, we reproduced a patient-specific geometry with a wall of biological tissue or silicone, and a parametric geometry based on in vivo data made in silicone. The biological tissue and the silicone were modeled with a fiber-oriented anisotropic and isotropic hyperelastic model, respectively. (3) Results: Clearly, both the aorta’s geometry and its constitutive material contribute to the determination of the aortic arch deformation; specifically, the parametric aorta exhibits a strain field similar to the patient-specific model with biological tissue. On the contrary, the local geometry affects the flow velocity distribution quite a lot, although it plays a minor role in the helicity along the arch. (4) Conclusions: The use of a patient-specific prototype in silicone does not a priori ensure a satisfactory reproducibility of the real aorta dynamics. Furthermore, the present simulations suggest that the realization of a simplified replica with the same compliance of the real aorta is able to mimic the overall behavior of the vessel.

In this paper we explore the possibility of using machine learning techniques to identify the characteristics of the initial components of a composite, starting from the knowledge of the macrobehavior of the heterogeneous material or structure. For a periodic medium, we show that the inverse relation can be approximated as easily as the direct one. We focus on the use of artificial neural networks, which are trained with the macroscale properties at the input layer and with the microstructural parameters at the output layer. For the training process, the pairs of macroscale properties- microstructural parameters are obtained by solving a set of boundary value problems for the unit cell, by means of a numerical homogenization procedure. In recall mode, the trained network attributes to the measured or wanted macroproperties at the input the unknown or sought microstructural parameters at the output. The presented method proved computationally efficient and can be a valid alternative when the analytical formulation of the homogenization inverse problem results is very difficult.