Method: We developed a prototype of a new generation of high content screening platform for simultaneous 3D live imaging of more than 1,000 organoids which relies on the single objective SPIM technique, which was extended to screen the cellular movements of developing gastruloids and liver organoids together with their differentiation state and extracellular matrix spatial distribution.

Method: We designed a freeform lens based optical system that we fabricated (optics, mechanics and electronics) completely in-house. We used injection moulded optical elements and a low cost camera instead of research grade sCMOS camera in order to reduce costs. In a later version, we included a low cost filter wheel to extend the operation of the system to a plurality of fluorescent beads.

Method: We developed a 69 MHz, sub-ps Ti-sapphire laser optimized for fibre delivery in a LMA, double-clad, polarization maintaining optical fibre (P-25/250DC-PM, Thorlabs) meeting the requirements for a head mounted, scanning two-photon microscope. The laser can be operated at around 800 nm, 920 nm and 1000 nm determined by the high reflectance bandwidth of saturable absorber mirrors (SAM) applied for mode-locking of our laser.

Method: For easier experimental implementation, ionising radiation can be replaced with non-ionising laser light. The system is based on an inverted microscope with a CMOS camera for widefield imaging. A 405 nm fibre-coupled laser is integrated through a custom sub-stage unit, where the beam is expanded and digitally scanned to position a sub-micron focus anywhere within a part of the objective’s field of view.

Method: The system is built on an inverted microscope. Its modular, fibre-connected architecture (laser, microscope, Raman spectrometer, Brillouin spectrometer) allows easy swapping of components with minimal optical realignment. Both spectrometers contain no moving parts, ensuring long-term mechanical stability.

The microscope, built around a commercial inverted frame (Olympus, IX-71), contains a 561 nm laser for Brillouin and Raman scattering. Raman spectra are recorded by a custom-designed spectrograph with a CCD camera. The technical challenge in recording Brillouin spectra lies in the closeness of the Brillouin peaks to the peak of elastically scattered photons (Rayleigh peak), which is many orders of magnitude more intensive than the Brillouin peaks. This means that high spectral resolution together with efficient suppression of the Rayleigh peak are necessary. To meet this challenge, we use custom-designed common-path interferometric filters (two in a series) and a virtually imaged phase array (VIPA) spectrograph equipped with a scientific CMOS camera. The design is an improved version of a previously described setup [Karampatzakis A. et al. (2017) npj Biofilms Microbiomes 3, 20] and its detail can be found at our GitHub repository.

Method: FA single photon sensitive light sheet microscope that can both detect and illuminate from either side of the sample was designed and constructed. The system is using a piezo driven 4-dimensional stage to hold the sample in a purpose made cuvette and a single photon sensitive camera whose field of view is split to show both imaging sides simultaneously. The system is also equipped with scanned optogenetic capabilities via a pulsed laser illumination.

Method: Synchronisation is controlled by a microcontroller (Arduino Teensy 4.0) running custom software. The controller sends trigger pulses to both the camera and LEDs, enabling LED alternation on consecutive camera frames. This approach eliminates delays between frames, making acquisition speed dependent only on the camera exposure time. Additionally, the microcontroller can receive logic pulses from external devices, such as a perfusion pump, and record their timing in units of camera frames from the start of acquisition.

Method: We developed and built an epifluorescence microscope capable of tilting in the range of ±90˚ from the vertical axis. A magnetically mounted positioning stage provides fine horizontal and vertical adjustment of the animal relative to the tilted microscope. The system uses a long-working-distance 5x objective, custom green/red LED excitation, and an sCMOS camera, with fast two-colour imaging enabled by fixed dual-band optics and LED-only channel switching. Camera control is handled by Micro-Manager, while a Teensy-based microcontroller synchronizes LED switching with camera exposures and logs external device triggers in frame-based timing. The custom microscope design enabled the integration of additional functionalities essential for its intended application, namely rapid switching between fluorescence channels and synchronization with a perfusion pump.

Method: We designed and built a two-photon microscope that could perform simultaneous two-photon imaging and one- or two-photon optogenetic excitation of cortical circuits in mouse brain. The microscope is equipped with two separate pulsed lasers operating at 920nm and 1060nm, that can be switched for simultaneous imaging/optogenetics.

Method: The microscope we designed and built has two separate parts in a single setup: one tracking the 3D movement of the zebrafish and one imaging the brain of the object. The 3D tracking is achieved by two IR microscopes installed in orthogonal orientation such that analysing the two images they provide the fish’s 3D location can be determined in unambiguously. The second system is a long working distance, high resolution microscope whose field of view can be adjusted by means of a pair of high-speed galvo scanners and an electrically tuneable lens. The system has two high power LED sources to provide the illumination for the high-resolution microscope.

Method: We numerically generated PSFs with varying amounts of aberrations of a given optical system at random locations within its field of view. Simultaneously, numerically generated ground truth objects were convolved with the aberrated PSFs to simulate aberrated images for deep learning training.

Method: Machine learning, in particular, convolutional neural network (CNN) can be applied to automatically distinguish the presence of components in a sample. To do so, the spectra of pure substances are first measurement. Thereafter, data augmentation including background addition, changing amplitude and tilting is performed on the measured spectra. The augmented dataset is then used to train the CNN model, enabling accurate identification of Raman spectral fingerprints.

Method: Reinforcement learning (RL), a machine learning method that enables dynamic learning of optimal actions through interaction with the environment, is employed to generate SPDs for microscope objective lenses. The RL agent iteratively selects optical elements and parameters based on the current design state, with each choice evaluated via a reward signal reflecting its impact on imaging performance. By exploring the solution space beyond conventional heuristics and data-driven approaches, the RL agent can identify optimal SPDs even for complex lens systems.

Method: Raman micro-spectroscopy was performed at a custom confocal micro-spectroscope system built by NOBIC for SCELSE. It was used to identify the chemical composition of individual live bacterial cells in samples cultured with different carbon sources (glucose or acetate) at different times. Raman spectral mapping was performed in selected cells to visualise the spatial distribution of the storage compounds present.

Method: We designed and built an adaptive optical system that can be used on both one and two photon optical microscopes. The optical system possesses a low speed adaptive element (deformable mirror) without the conventional wavefront sensor. It is therefore not possible to correct for aberrations at each point as the system scans from pixel to pixel. However, it is possible to set the deformable mirror to a shape when, even though the point spread of the system at many sample locations is possibly worse than it would be without the adaptive element, the image so obtained contains more information (in terms of Fischer information) than the uncorrected system. The high information content image can hence be calculated from the intensity data obtained via scanning. This can be made faster using trained AI networks, such as CNNs.

Method: After the debridement procedure, an antibody conjugated to a near-infrared fluorescent dye is applied to the wound, where it selectively binds to a biomarker associated with biofilms. A handheld imaging device, equipped with separate illumination and detection arms, then assesses the wound using fluorescence imaging under ambient lighting conditions. The resulting signal indicates the presence or absence of biofilm, enabling an objective evaluation of debridement effectiveness.

Method: We designed an ultra-broadband OCT that advances the state-of-the art in two major aspects: the design and implementation of a broadband spectrometer employing a 4k-pixel line camera achieving a spectral range of greater than 450nm with a central wavelength of 680nm; and an achromatised imaging system that minimises aberrations in the visible-near-infrared wavelength range for imaging mouse retina. The system provides an axial imaging range greater than 0.67mm within retina with a larger than 40° field angle.

Method: The project delivered a reader based on Raspberry Pi computer and camera with integrated optics and ML-based colour recognition to detect colour changes of paper assays. Each individual sample is labelled with a QR code, which is read by the camera to enable accurate identification and tracking during analysis.

Method: Brillouin microscopy is widely used to measure biomechanical properties of biological samples, but it is highly sensitive to sample hydration levels. Optical coherence elastography provides more reliable measurements, however it requires mechanical stimulation of the sample. In this project, we explore using thermal instead of mechanical stimulation by inducing localised heating via two-photon absorption of pulsed laser light. The resulting temperature change is measured using an optical coherence tomograph to probe micromechanical properties. We are currently performing proof-of-concept studies on phantoms and correlating the results with Brillouin spectroscopy.