A neural network was trained to generate synthetic images of severe and moderate psoriatic plaques, after being trained on 375 photographs of patients with psoriasis taken in a clinical setting. A latent w-space vector was identified that allowed the degree of severity of the psoriasis in the generated images to be modified. A second latent w-space vector was identified that allowed the size of the psoriasis plaque to be modified and this was used to show the potential to alleviate bias in the training data. With appropriate training data, such an approach could see a future application in a clinical setting where a patient is able to observe a prediction for the appearance of their skin and associated skin condition under a range of treatments and after different time periods, hence allowing an informed and data-driven decision on optimal treatment to be determined.

Abstract Introduction and aims Narrowband ultraviolet B therapy (NB-UVB) is an effective treatment for psoriasis. Currently, NB-UVB is delivered in whole-body cabinets to all of the patient’s skin, thus increasing the risk of normal skin sunburning and the potential risk of skin cancer development. This limits the NB-UVB doses that can be used and requires patients attend sessions three times weekly for up to 10 weeks in order to complete treatment. An artificial intelligence (AI)-mediated, targeted NB-UVB approach would avoid these risks and allow use of higher NB-UVB doses for the areas of the skin affected by psoriasis. We sought to train an artificial neural network on digital images of psoriasis, and subsequently test this trained model in order to assess its accuracy at distinguishing psoriasis from unaffected skin. Methods Overall, 104 digital images of psoriasis from 14 patients with this skin condition, who attended the dermatology department at University Hospital Southampton NHS Foundation Trust, were used to train a conditional generative adversarial network– ‘pix2pix’. Areas of psoriasis were delineated within the images prior to the images being augmented by randomly cropping, rotating and altering colour intensity, thereby increasing the total number to 23 111 images to train and test the neural network. Results The trained neural network was able to identify psoriasis in unseen digital images with 96.73% accuracy at the pixel level across the dataset. The output of the neural network (i.e. a computer-generated image of psoriasis) acted as a signal source for a digital micromirror device (DMD) that was able to shape visible light and project this light over the original image of psoriasis, while avoiding the uninvolved skin. Conclusions It is possible to differentiate psoriasis from unaffected skin within digital images using an artificial neural network. We demonstrated that AI-mediated targeted phototherapy using a DMD is feasible.

This work combines two existing technologies to demonstrate the possibility for automated, targeted, phototherapy of psoriasis and other skin conditions: 1) Image-to-image translation via a neural network (NN) as a method of image segmentation. 2) Light control using a digital micromirror device (DMD). With a small dataset of just 104 patient photographs (labeled by expert dermatologists) our NN model was trained to identify regions of psoriasis that required treatment, achieving an average accuracy of 96.6%. The image output of the NN model was applied to a DMD and precise control over the shape of the illuminated region was demonstrated. In the proposed automated phototherapy device this would target treatment to the affected regions, minimizing exposure of healthy skin and the associated risks of patient harm.

This paper reports the fabrication and first demonstration of all-dielectric crystalline grating–waveguide reflectors comprising a Sc2O3 waveguide grown on a sub-wavelength-patterned sapphire substrate. Rigorous coupled-wave analysis is employed to simulate the operation of the structure, suggesting a 100% resonance reflectivity in theory. Structuring of the sapphire substrate is achieved using inductively coupled plasma etching, whilst pulsed laser deposition is used for epitaxial growth of the Sc2O3 crystalline waveguide. Devices with distinct TE- and TM-polarisation resonances around 1030-nm for an angle of incidence near 10° are demonstrated, with reflectance approaching 90%. The discrepancy in reflectivity is attributed to the waveguide thickness variation and surface roughness. Refinement of the fabrication processes and tolerances should lead to improvement in the surface quality of the crystalline grating–waveguide structure and operation closer to the ideal resonance reflectivity.

Practical application of coherent beam combination of multiple fibres necessitates phase identification and optimisation in real-time. Here, we solve this mathematical challenge via deep learning, and hence demonstrate the potential for real-time bespoke beam shaping.

Pollen grains dehydrate during their development and following their departure from the host stigma. Since the size and shape of a pollen grain can be dependent on environmental conditions, being able to predict both of these factors for hydrated pollen grains from their dehydrated state could be beneficial in the fields of climate science, agriculture, and palynology. Here, we use deep learning to transform images of dehydrated Ranunculus pollen grains into images of hydrated Ranunculus pollen grains. We also then use a deep learning neural network that was trained on experimental images of different genera of pollen grains to identify the hydrated pollen grains from the generated transformed images, to test the accuracy of the image generation neural network. This pilot work demonstrates the first steps needed towards creating a general deep learning-based rehydration model that could be useful in understanding and predicting pollen morphology.

Catheter-associated urinary tract infection (CAUTI) is a leading cause of hospital-acquired infections worldwide causing debilitating illness for patients as well as a significant financial and treatment burden on health services. CAUTI is linked with the build-up of biofilms on catheter surfaces which act as a reservoir for infection. Additionally, urease-producing bacteria such as Gram-negative Proteus mirabilis (PM), can form crystalline biofilms which encrust catheter surfaces ultimately leading to blockages which require immediate removal of the catheter. Currently there are limited treatments available to prevent the formation of biofilms by PM as well as other urinary tract infection causing bacteria. A novel concept for a light-guiding urinary catheter is presented where a silicone elastomer waveguide incorporated along the length of the catheter is used to irradiate the catheter surfaces with antimicrobial blue light (405 nm) to prevent biofilm formation in situ. The prototype device is mass producible while also easy to fabricate in a lab setting for research studies. The inhibitory effect of blue light on PM biofilm formation over a range of irradiances is described for the first time showing an LD90 at 192-345 J/cm(2) and total inhibition at 1,700 J/cm(2) In vitro studies show that the light-guiding catheter (LGC) prototypes exhibit a 98% inhibition in PM biofilm formation inside the catheter lumen at an average estimated irradiance of 30-50 mW/cm(2) (324-540 J/cm(2) fluence) showing that the concept is highly effective, promising to be a powerful and economical antimicrobial approach to prevent catheter associated biofilm development and blockage.

Lateral flow devices (LFDs) or lateral flow tests (LFTs) are one of the most widely used biosensor platforms for point-of-care (POC) diagnostics. The basic LFD design has remained largely unchanged since its first appearance, and this has limited LFD use in clinical applications due to a general lack of analytical sensitivity. We report here a comprehensive study of the use of laser-patterned geometric control barriers that influence the flow dynamics within an LFD, with the specific aim of enhancing LFD sensitivity and lowering the limit of detection (LOD). This control of sample flow produces an increase in the time available for optimizing the binding kinetics of the implemented assay. The geometric modification to the flow path is in the form of a constriction that is produced by depositing a photo-sensitive polymer onto the nitrocellulose membrane which when polymerized, creates impermeable barrier walls through the depth of the membrane. Both the position of the constriction within the flow path and the number of constrictions allow for an increase in the sensitivity because of a slower overall flow rate within the test and a larger volume of sample per unit width of the test line. For these high sensitivity LFDs (HS-LFD), through optimization of the constriction position and addition of a second constriction we attained a 62% increase in test line color intensity for the detection of procalcitonin (PCT) and were also able to lower the LOD from 10 ng/mL to 1 ng/mL. In addition, of relevance for future commercial exploitation, this also significantly decreases the antibody consumption per device leading to reduced costs for test production. We have further tested our HS-LFD with contrived human samples, validating its application for future clinical use.

Laser processing techniques such as laser machining, marking, cutting, welding, polishing and sintering have become important tools in modern manufacturing. A key step in these processes is to take the intended design and convert it into coordinates or toolpaths that are useable by the motion control hardware and result in efficient processing with a sufficiently high quality of finish. Toolpath design can require considerable amounts of skilled manual labor even when assisted by proprietary software. In addition, blind execution of predetermined toolpaths is unforgiving, in the sense that there is no compensation for machining errors that may compromise the quality of the final product. In this work, a novel laser machining approach is demonstrated, utilizing reinforcement learning (RL) to control and supervise the laser machining process. This autonomous RL-controlled system can laser machine arbitrary pre-defined patterns whilst simultaneously detecting and compensating for incorrectly executed actions, in real time.

This paper reports a study of the effect of laser pulse repetition rate and substrate temperature on the properties of crystalline Sc 2 O 3 films grown on <0001>-oriented sapphire via pulsed laser deposition (PLD). For the range of substrate laser-heating powers investigated, optimum <111>-film growth was obtained at the highest available repetition rate of 100 Hz. Scanning electron microscopy and X-ray diffraction measurements revealed that a decrease in the repetition rate favours highly textured/island growth. The repetition rate was also proven to affect the lattice constant of the films, with a difference up to 0.3% between the films grown at high and low repetition rates. The general trend indicates that the out-of-plane lattice constant of the film shifts closer to the bulk value when the pulse repetition rate is increased. However, the lattice constant could also be reduced through in-situ post-growth annealing, which had the additional benefit of reducing the width of the (222) rocking curve peak. This work provides further evidence that energetic-PLD dynamics leads to higher quality PLD-grown crystalline films and for this material undermines the old adage that lower pulse repetition rates lead to higher quality thin-film crystalline growth.

Focused ion beam (FIB) milling is an important rapid prototyping tool for micro- and nanofabrication and device and materials characterization. It allows for the manufacturing of arbitrary structures in a wide variety of materials, but establishing the process parameters for a given task is a multidimensional optimization challenge, usually addressed through time-consuming, iterative trial-and-error. Here, we show that deep learning from prior experience of manufacturing can predict the postfabrication appearance of structures manufactured by focused ion beam (FIB) milling with >96% accuracy over a range of ion beam parameters, taking account of instrument- and target-specific artifacts. With predictions taking only a few milliseconds, the methodology may be deployed in near real time to expedite optimization and improve reproducibility in FIB processing.

Since the pollen of different species varies in shape and size, visualizing the 3-dimensional structure of a pollen grain can aid in its characterization. Lensless sensing is useful for reducing both optics footprint and cost, while the capability to image pollen grains in 3-dimensions using such a technique could be truly disruptive in the palynology, bioaerosol sensing, and ecology sectors. Here, we show the ability to employ deep learning to generate 3-dimensional images of pollen grains using a series of 2-dimensional images created from 2-dimensional scattering patterns. Using a microscope to obtain 3D Z-stack images of a pollen grain and a 520 nm laser to obtain scattering patterns from the pollen, a single scattering pattern per 3D image was obtained for each position of the pollen grain within the laser beam. In order to create a neural network to transform a single scattering pattern into different 2D images from the Z-stack, additional Z-axis information is required to be added to the scattering pattern. Information was therefore encoded into the scattering pattern image channels, such that the scattering pattern occupied the red channel, and a value indicating the position in the Z-axis occupied the green and blue channels. Following neural network training, 3D images were formed from collated generated 2D images. The volumes of the pollen grains were generated with a mean accuracy of ∼84%. The development of airborne-pollen sensors based on this technique could enable the collection of rich data that would be invaluable to scientists for understanding mechanisms of pollen production climate change and effects on the wider public health.

A novel subtractive laser patterning approach, actuated by reinforcement learning-controlled motorized XYZ stages, enables path-planning for quasi-arbitrary target patterns and an ability to supervise the machining process and perform self-correction following inadvertent pulse misfires.

Reinforcement learning was carried out in a simulated environment to learn continuous velocity control over multiple motor axes. This was then applied to a real-world optical tweezers experiment with the objective of moving a laser-trapped microsphere to a target location whilst avoiding collisions with other free-moving microspheres. The concept of training a neural network in a virtual environment has significant potential in the application of machine learning for experimental optimization and control, as the neural network can discover optimal methods for problem solving without the risk of damage to equipment, and at a speed not limited by movement in the physical environment. As the neural network treats both virtual and physical environments equivalently, we show that the network can also be applied to an augmented environment, where a virtual environment is combined with the physical environment. This technique may have the potential to unlock capabilities associated with mixed and augmented reality, such as enforcing safety limits for machine motion or as a method of inputting observations from additional sensors.

Interactions between light and matter during short-pulse laser materials processing are highly nonlinear, and hence acutely sensitive to laser parameters such as the pulse energy, repetition rate, and number of pulses used. Due to this complexity, simulation approaches based on calculation of the underlying physical principles can often only provide a qualitative understanding of the inter-relationships between these parameters. An alternative approach such as parameter optimisation, often requires a systematic and hence time-consuming experimental exploration over the available parameter space. Here, we apply neural networks for parameter optimisation and for predictive visualisation of expected outcomes in laser surface texturing with blind vias for tribology control applications. Critically, this method greatly reduces the amount of experimental laser machining data that is needed and associated development time, without negatively impacting accuracy or performance. The techniques presented here could be applied in a wide range of fields and have the potential to significantly reduce the time, and the costs associated with laser process optimisation.

As the SARS-CoV-2 pandemic continues to spread, the necessity for rapid, easy diagnostic capabilities could never have been more crucial. With this aim in mind, we have developed a cost-effective and time-saving testing methodology/strategy that implements a sensitive reverse transcriptase loop-mediated amplification (RT-LAMP) assay within narrow, commercially available and cheap, glass capillaries for detection of the SARS-CoV-2 viral RNA. The methodology is compatible with widely used laboratory-based molecular testing protocols and currently available infrastructure. It employs a simple rapid extraction protocol that lyses the virus, releasing sufficient genetic material for amplification. This extracted viral RNA is then amplified using a SARS-CoV-2 RT-LAMP kit, at a constant temperature and the resulting amplified product produces a colour change which can be visually interpreted. This testing protocol, in conjunction with the RT-LAMP assay, has a sensitivity of ∼100 viral copies per reaction of a sample and provides results in a little over 30 min. As the assay is carried out in a water bath, commonly available within most testing laboratories, it eliminates the need for specialised instruments and associated skills. In addition, our testing pathway requires a significantly reduced quantity of reagents per test while providing comparable sensitivity and specificity to the RT-LAMP kit used in this study. While the conventional technique requires 25 μl of reagent, our test only utilises less than half the quantity (10 μl). Thus, with its minimalistic approach, this capillary-based assay could be a promising alternative to the conventional testing, owing to the fact that it can be performed in resource-limited settings, using readily available apparatus, and has the potential of increasing the overall testing capacity, while also reducing the burden on supply chains for mass testing.

The identification and imaging of pollen grains in-flight was performed via illumination of the pollen grains with three collinear laser beams that had central wavelengths of 450 nm, 520 nm and 635 nm. Two neural networks are reported here; the first neural network was able to categorise pollen grain species from their scattering patterns with ∼86% accuracy, while the second neural network generated images of the pollen grains from their scattering patterns. This work demonstrates the potential application of laser scattering and deep learning for real-world in-flight pollen identification.

Pollen can lead to individuals suffering from allergic rhinitis, with a person’s vulnerability being dependent on the species and the amount of pollen. Therefore, the ability to precisely quantify both the number and species of pollen grains in a certain volume would be invaluable. Lensless sensing offers the ability to classify pollen grains from their scattering patterns, with the use of very few optical components. However, since there could be 1000 s of species of pollen one may wish to identify, in order to avoid having to collect scattering patterns from all species (and mixtures of species) we propose using two separate neural networks. The first neural network generates a microscope equivalent image from the scattering pattern, having been trained on a limited number of experimentally collected pollen scattering data. The second neural network segments the generated image into its components, having been trained on microscope images, allowing pollen species identification (potentially allowing the use of existing databases of microscope images to expand range of species identified by the segmentation network). In addition to classification, segmentation also provides richer information, such as the number of pixels and therefore the potential size of particular pollen grains. Specifically, we demonstrate the identification and projected area of pollen grain species, via semantic image segmentation, in generated microscope images of pollen grains, containing mixtures and species that were previously unseen by the image generation network. The microscope images of mixtures of pollen grains, used for training the segmentation neural network, were created by fusing microscope images of isolated pollen grains together while the trained neural network was tested on microscope images of actual mixtures. The ability to carry out pollen species identification from reconstructed images without needing to train the identification network on the scattering patterns is useful for the real-world implementation of such technology.

Background Controlling the spread of SARS-CoV-2 is problematic because of transmission driven by asymptomatic and pre-symptomatic individuals. Community screening can help identify these individuals but is often too expensive for countries with limited health care resources. Low-cost ELISA assays may address this problem, but their use has not yet been widely reported. Methods We developed a SARS-CoV-2 nucleocapsid ELISA and assessed its diagnostic performance on nose and throat swab samples from UK hospitalised patients and sputum samples from patients in Ghana. Results The ELISA had a limit of detection of 8.4 pg/ml antigen and 16 pfu/ml virus. When tested on UK samples (128 positive and 10 negative patients), sensitivity was 58.6% (49.6-67.2) rising to 78.3% (66.7-87.3) if real-time PCR Ct values > 30 were excluded, while specificity was 100% (69.2-100). In a second trial using the Ghanaian samples (121 positive, 96 negative), sensitivity was 52% (42.8-61.2) rising to 72.6% (61.8-81.2) when a > 30 Ct cut-off was applied, while specificity was 100% (96.2-100). Conclusions: Our data show that nucleocapsid ELISAs can test a variety of patient sample types while achieving levels of sensitivity and specificity required for effective community screening. Further investigations into the opportunities that this provides are warranted.