Machine learning instructed microfluidic synthesis of curcumin-loaded liposomes.

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.

Beating heart implantation of transventricular artificial cordae: How can access site selection and leaflet insertion improve mitral regurgitation correction?

NeoChord-DS1000-System (NC) and The Harpoon-Mitral-Repair-System (H-MRS) are two trans-apical chordal implantation devices developed for the treatment of degenerative mitral valve (MV) regurgitation (DMR) either if as Fibroelastic-Deficiency (FED), Forma-Frusta (FF), or Barlow (B) presentation. The aim of this study is to evaluate some of the advantages and disadvantages of these two different devices by performing numerical simulation analyses focused on different transventricular access sites in all subsets of DMR presentations. By applying a novel approach for the development of patient-specific MV domains we worked out a set of numerical simulations of the artificial chordae implantation. Different leaflet insertions and ventricle access sites were investigated, and resulting contact-area (CA), tensioning-forces (F) and leaflet's stress (LS) were calculated. The analyses showed that: i) NC-approach maintains low LS when performed with a posterior access site and optimizes the overlap between the leaflets at the systolic peak; ii) H-MRS-system presents better results in case of a more anterior ventricular entry site; however, for FED prolapse large variation of F and LS with respect to NC-approach are found; iii) an accidental contact between artificial sutures and the anterior leaflet may occur when valve function is restored through an excessive anterior access site. Present findings set light on specific technical aspects of transapical off-pump chords implantation, either performed with NC and H-MRS systems and highlight the advantages and disadvantages proper to the two devices. Our study also paves the basis for a systematic application of computational methodology, in order to plan a patient-specific mini-invasive approach thus maximizing the outcomes.

Drug delivery: Experiments, mathematical modelling and machine learning.

We address the problem of determining from laboratory experiments the data necessary for a proper modeling of drug delivery and efficacy in anticancer therapy. There is an inherent difficulty in extracting the necessary parameters, because the experiments often yield an insufficient quantity of information. To overcome this difficulty, we propose to combine real experiments, numerical simulation, and Machine Learning (ML) based on Artificial Neural Networks (ANN), aiming at a reliable identification of the physical model factors, e.g. the killing action of the drug. To this purpose, we exploit the employed mathematical-numerical model for tumor growth and drug delivery, together with the ANN - ML procedure, to integrate the results of the experimental tests and feed back the model itself, thus obtaining a reliable predictive tool. The procedure represents a hybrid data-driven, physics-informed approach to machine learning. The physical and mathematical model employed for the numerical simulations is without extracellular matrix (ECM) and healthy cells because of the experimental conditions we reproduce.

Predicting the growth of glioblastoma multiforme spheroids using a multiphase porous media model.

Tumor spheroids constitute an effective in vitro tool to investigate the avascular stage of tumor growth. These three-dimensional cell aggregates reproduce the nutrient and proliferation gradients found in the early stages of cancer and can be grown with a strict control of their environmental conditions. In the last years, new experimental techniques have been developed to determine the effect of mechanical stress on the growth of tumor spheroids. These studies report a reduction in cell proliferation as a function of increasingly applied stress on the surface of the spheroids. This work presents a specialization for tumor spheroid growth of a previous more general multiphase model. The equations of the model are derived in the framework of porous media theory, and constitutive relations for the mass transfer terms and the stress are formulated on the basis of experimental observations. A set of experiments is performed, investigating the growth of U-87MG spheroids both freely growing in the culture medium and subjected to an external mechanical pressure induced by a Dextran solution. The growth curves of the model are compared to the experimental data, with good agreement for both the experimental settings. A new mathematical law regulating the inhibitory effect of mechanical compression on cancer cell proliferation is presented at the end of the paper. This new law is validated against experimental data and provides better results compared to other expressions in the literature.

Optimizing particle size for targeting diseased microvasculature: from experiments to artificial neural networks.

BACKGROUND: Nanoparticles with different sizes, shapes, and surface properties are being developed for the early diagnosis, imaging, and treatment of a range of diseases. Identifying the optimal configuration that maximizes nanoparticle accumulation at the diseased site is of vital importance. In this work, using a parallel plate flow chamber apparatus, it is demonstrated that an optimal particle diameter (d(opt)) exists for which the number (n(s)) of nanoparticles adhering to the vessel walls is maximized. Such a diameter depends on the wall shear rate (S). Artificial neural networks are proposed as a tool to predict n(s) as a function of S and particle diameter (d), from which to eventually derive d(opt). Artificial neural networks are trained using data from flow chamber experiments. Two networks are used, ie, ANN231 and ANN2321, exhibiting an accurate prediction for n(s) and its complex functional dependence on d and S. This demonstrates that artificial neural networks can be used effectively to minimize the number of experiments needed without compromising the accuracy of the study. A similar procedure could potentially be used equally effectively for in vivo analysis.