Benchtop mesoSPIM: a next-generation open-source light-sheet microscope for cleared samples

Tissue clearing combined with light-sheet microscopy has enabled rapid, high-contrast, volumetric imaging of intact organs and whole organisms but poses conflicting demands of large field of view and high resolution. As cameras evolve toward larger sensors with rolling-shutter modes suitable for light-sheet imaging, many standard life-science objectives deliver poor image quality at the periphery, and objective performance is rarely quantified independently of excitation properties. The previous mesoSPIM v.5 system, while facility-grade, was constrained by a macro-zoom detection path incompatible with modern large-sensor cameras, a large and expensive footprint, and limited magnification range. This work addresses the need for an open, higher-resolution, larger-FOV, higher-throughput, and easier-to-build light-sheet microscope and establishes objective selection criteria via a new, illumination-independent testing method.

The authors situate their work within the trend of large-sensor imaging in astrophotography and machine vision and cite high-resolution, large-FOV microscopy systems including AMATERAS (122 MP sensor, 2.5 µm resolution), Mesolens (0.7 µm across 24 mm field), and ExA-SPIM achieving 1 µm lateral and 2.5 µm axial resolution with a 151 MP camera. They note the increasing availability of large-sensor sCMOS cameras with rolling-shutter modes suitable for ASLM but highlight incompatibility with common objectives designed for ~18–25 mm fields. Community efforts like QUAREP-LiMi advocate standardized quality metrics; however, detection objective performance is often not assessed independently of light-sheet parameters. Prior mesoSPIM versions used ASLM to maintain uniform axial resolution but were limited by a macro-zoom objective (Olympus MVX-10) causing vignetting and reduced periphery quality on large sensors.

System design: The Benchtop mesoSPIM adopts a modular, compact architecture operable on a lab bench without an optical table. It replaces the macro-zoom detection with fixed-magnification long working distance air objectives and supports a large-sensor sCMOS camera (e.g., Teledyne Photometrics Iris 15). The detection path can be configured with infinity-corrected objective plus tube lens or with industrial telecentric lenses for very low magnifications. The excitation path employs dual symmetric arms using an electrically tunable lens (ETL) for remote focusing and ASLM, galvo scanning, and a modified Nikon 50 mm f/1.4 G objective to form the light-sheet. ASLM: Axially swept light-sheet microscopy is implemented with ETLs synchronized to the camera’s rolling shutter to maintain a thin, uniform beam waist across the FOV. Users tune ETL offset and amplitude in software with the beam parked and no sample present. Mechanics and stages: A stationary immersion chamber on a kinematic mount ensures constant optical paths. Samples are translated in X/Y/Z and rotated using motorized stages (up to 50 × 50 × 100 mm travel; rotary stage for rotation). Vibration isolation is achieved with a lightweight breadboard on sorbothane isolators. Sample handling: A suite of 3D-printed PA12 holders accommodates samples from 3 to 75 mm and multiple clearing media (BABB, DBE, ECI), including cuvette holders, multi-sample SPIM-tower (4 samples per level; stackable), and large cuvette mounts. Objective testing (contrast and field flatness): To decouple detection from illumination performance, the authors adapted full-field lens testing using a 40 lp/mm Ronchi ruling illuminated with diffuse, incoherent light. A 3D stack is acquired by stepping focus (10 µm steps over ±300 µm). Each image is subdivided, and local contrast C(X,Y,Z) is computed (percentile-based). From C(X,Y,Z), the best-focus surface, depth of field, maximum contrast maps, and field flatness are derived. Tests emulate cleared tissue by inserting ~20 mm of n≈1.52 medium between objective and target. Additional chromatic testing uses a halogen lamp with bandpass filters and DBE immersion to quantify channel-dependent focus shifts. Software: Open-source Python (PyQt5) control software supports diverse hardware, multithreaded acquisition, and fast saving to Fiji BigDataViewer HDF5/XML via npy2bdv for efficient stitching and visualization of terabyte-scale datasets. Features include focus interpolation across Z, per-channel refocusing, and modular configuration. Performance quantification: Resolution measured with 0.2 µm fluorescent beads in RI=1.52 media; PSFs fitted to extract lateral and axial FWHM across FOV. Field-of-view and image quality comparisons to v.5 were made using identical magnification settings. Throughput assessed via imaging time for standardized volumes. Additional demonstrations include high-throughput SPIM-tower imaging and imaging of color centers in irradiated CaF2 crystals, with statistical validation of repeated track candidates.

• Optical performance: Achieves 1.5 µm lateral (best case with Mitutoyo G Plan Apo 20x/0.28) and 3.3 µm axial resolution (ASLM-limited) across the entire FOV; lateral resolution ranges 1.5–2.7 µm depending on objective and medium thickness.
• Field of view and sensor: Supports large-sensor sCMOS (25 mm diagonal), providing 1.9× larger sensor area and 3.6× more pixels per image (15 MP vs 4 MP) compared to mesoSPIM v.5 (Orca Flash4.3), with improved field flatness and contrast across the field.
• Throughput: Can image ~1 cm³ volume in ~13 minutes; multi-sample SPIM-tower enabled imaging of 16 Xenopus tadpoles in ~20 minutes.
• Objective selection: Industrial long-WD air objectives from Mitutoyo (Plan Apo BD/M/G series at 2x–20x) outperformed the Olympus MVPLAPO-1x zoom system under cleared-tissue emulation, with higher and more uniform contrast and flatter best-focus surfaces. Thorlabs super apochromats and Olympus XLFluor 4x/0.28 also scored well. Telecentric lenses (0.9x, 1.2x) were effective for low-magnification screening.
• Chromatic effects: Under DBE/RI~1.52 conditions, significant channel-dependent focus offsets were observed—up to ~400 µm between 420 nm and 697/75 nm bands; typical GFP vs RFP offsets ~50–150 µm. Per-channel refocusing is required for optimal performance.
• Applications: Demonstrated single-axon imaging in Thy1-GFP mouse CNS at 5x; high-mag neuronal morphology at 20x; whole-mouse imaging at 0.9x; chicken embryo PNS at 1.2x; APP/PS1 amyloid plaques and vasculature at 1.2x; retrograde AAV labeling; human cortex cytoarchitecture at 5x; whole Xenopus tadpoles for developmental screening. Physics use-case: detection of color centers in irradiated CaF2 crystals at 20x with repeated scans and statistical validation of candidate tracks.
• Practicality and cost: Compact 0.25 m² footprint, travel-friendly modularity (packed 60 kg; reassembly <1 h plus ≤1 h alignment), and reduced parts cost ($100k vs ~$162k for v.5). Upgrade path for older systems’ detection to Benchtop-like performance for ~$27k (including camera).
• Compatibility: Supports multiple clearing protocols and interchangeable sample holders; operates without an optical table; open-source hardware designs and software facilitate adoption.

The Benchtop mesoSPIM addresses the core challenge of combining large FOV with high resolution by pairing ASLM-based uniform axial resolution with a detection path optimized for large sensors and rigorously validated objectives. The illumination-independent contrast mapping method enabled selection of objectives that maintain high, uniform contrast and field flatness under cleared-tissue conditions, directly improving image quality and throughput. Compared to ExA-SPIM and NODO, Benchtop offers lower peak resolution/throughput but at substantially reduced cost and with broader versatility in objective choice; compared to descSPIM, it provides higher resolution, FOV, and throughput at moderate additional cost. The demonstrated biological and physics applications underscore the platform’s relevance across domains, while the open-source control and hardware lower barriers to adoption. The findings support the utility of per-channel refocusing in cleared samples and motivate standardized, objective-centric metrics for system performance independent of excitation. The approach paves the way for scalable, facility-grade cleared tissue imaging and novel use-cases such as color-center tracking in crystals.

This work introduces the Benchtop mesoSPIM, an open, compact, and cost-reduced light-sheet microscope delivering high-resolution, large-FOV imaging with high throughput across diverse cleared samples. Key contributions include an illumination-independent objective testing method producing full-field 3D contrast maps, selection and validation of optimal industrial objectives for large sensors, and a comprehensive, modular hardware/software ecosystem. Demonstrations span neuroscience, developmental biology, pathology, and physics. Future directions include designing long-WD air objectives (NA 0.3–0.5, 5x–20x) with built-in compensation for thick high-index media, further compacting electronics and mechanics (e.g., integrated controllers and motorized turrets), expanding test target orientations and spatial frequencies for richer MTF mapping, and community-driven enhancements to broaden applications and reduce costs.

• Spherical aberration and medium mismatch: Using off-the-shelf long-WD air objectives limits diffraction-limited performance to approximately NA ≤ 0.15 and immersion medium thickness ≤ 15 mm ("15–15 rule"). Higher-NA detection improves resolution primarily for thinner samples (<15 mm total medium plus sample).
• Chromatic focal shifts: Significant channel-dependent focus offsets (up to 400 µm across the visible spectrum in RI1.52 media) necessitate per-channel refocusing, varying by objective and medium.
• Field flatness interpretation: The measured best-focus surface curvature in cleared-tissue emulation reflects combined effects (spherical aberration, coma, astigmatism, non-telecentricity) and should not be conflated with Petzval field curvature.
• Mixed open/closed components: Although many parts are open-source, the system relies on several closed-source components; integration depends on clearly specified interfaces.
• Resolution ceiling: Axial resolution is fundamentally constrained by light-sheet thickness in ASLM; custom objectives corrected for strong spherical aberrations could further improve performance but are not yet standard.