Machine learning is the art of combining a set of measurement data and predictive variables to forecast future events. Every day, new model approaches (with high levels of sophistication) can be found in the literature. However, less importance is given to the crucial stage of validation. Validation is the assessment that the model reliably links the measurements and the predictive variables. Nevertheless, there are many ways in which a model can be validated and cross-validated reliably, but still, it may be a model that wrongly reflects the real nature of the data and cannot be used to predict external samples. This manuscript shows in a didactical manner how important the data structure is when a model is constructed and how easy it is to obtain models that look promising with wrong-designed cross-validation and external validation strategies. A comprehensive overview of the main validation strategies is shown, exemplified by three different scenarios, all of them focused on classification.

Human exposure to nano- and microplastics (NMPs) has raised major societal concerns, yet no framework to assess the risks of NMPs for human health exists. A substantial proportion of plastic produced worldwide is not properly disposed and persists in the environment for decades while degrading. Plastic degradation generates a size continuum of fragments, including nano- and microplastic particles, with numerous associated environmental pollutants and plastic additives, and microbial communities colonising their surfaces. The ubiquitous presence of NMPs, their availability for uptake by organisms and their potential to act as vectors for toxicants and pathogens render risk assessment a priority on the political agenda at the global level. We provide a new, fully integrated risk assessment framework tailored to the specificities of NMPs, enabling an assessment of current and future human health risks from NMPs. The framework consists of four novel paradigms to the traditional risk assessment methodology. These paradigms deal with techniques in NMP analysis, gaps in empirical data, theoretical and modelling approaches and stakeholder engagement. Within the proposed framework, we propose how we can use research experiences gained so far to carry out the different steps of the assessment process, and we define priorities for further research.

We have developed a self-contained, liquid tunable microlens based on polyacrylate membranes integrated with compact on-chip thermo-pneumatic actuation fabricated using full-wafer processing. Silicone oil is used as the optical liquid, which is pushed or pulled into the lens cavity via an extended microfluidic channel structure without any pumps, valves or other mechanical means. The heat load generated by the thermal actuator is physically isolated from the lens chamber. The back focal length may be tuned from infinity to 4 mm with a maximum power consumption of 300 mW. The principal application is fine tuning of the back focal length, for which tuning time constants as small as 100 ms are suitable.

An integrated tunable microlens, whose focal length may be varied over a range of 3 to 15 mm with total power consumption below 250 mW, is presented. Using thermo-pneumatic actuation, this adaptive optical microsystem is completely integrated and requires no external pressure controllers for operation. The lens system consists of a liquid-filled cavity bounded by a distensible polydimethyl-siloxane membrane and a separate thermal cavity with actuation and sensing elements, all fabricated using silicon, glass and polymers. Due to the physical separation of thermal actuators and lens body, temperature gradients in the lens optical aperture were below 4°C in the vertical and 0.2°C in the lateral directions. Optical characterization showed that the cutoff frequency of the optical transfer function, using a reference contrast of 0.2, varied from 30 lines/mm to 65 lines/mm over the tuning range, and a change in the numerical aperture from 0.067 to 0.333. Stable control of the focal length over a long time period using a simple electronic stabilization circuit was demonstrated.

Employing Bessel beams in imaging takes advantage of their self-reconstructing properties to achieve small focal points while maintaining a large depth of focus. Bessel beams are efficiently generated using axicons, and their utility in scanning imaging systems, such as optical coherence tomography (OCT), has been demonstrated. As these systems are miniaturized to allow, for example, endoscopic implementations, micro-axicons are required to assure the maintenance of a large depth of focus. We demonstrate here the design, fabrication, and application of molded micro-axicons for use in silicon-based micro-optical benches. It is shown that arrangements of multiple convex and concave axicons may be implemented to optimize the depth of focus in a miniaturized OCT system, using a telescopic optical arrangement of considerably shorter optical system length than that achievable with classical micro-optics.