Speaker
Description
The manipulation of three-dimensional (3D) cellular structures such
as organoids and spheroids plays a central role in modern biomedical
applications, including tissue engineering, drug screening, and disease
modeling. However, the accurate and reproducible transfer of these fragile
structures remains a technical bottleneck. Manual handling is limited by user
variability, lacks scalability, and is inherently incompatible with high-
throughput workflows. Laser-assisted bioprinting has emerged as a contactless
and precise alternative to traditional pipetting or extrusion-based methods. In
this context, artificial intelligence (AI) and robotic automation offer the
potential to transform cell handling by enabling intelligent decision-making,
adaptive control, and real-time feedback. We present PickCell(TM), a fully
integrated robotic platform for real-time detection, trajectory optimization,
and laser-assisted bioprinting of organoids.
The system integrates a 6-axis robotic arm, a nanosecond pulsed laser (1 ns,
1064 nm, ~30 µJ), high-resolution optics, and a custom software suite. Object
detection is performed by a YOLOv11 small model (YOLOv11s)[1], trained on 512x512
grayscale image patches and applied to full 1024x1024 images. Depending on
spheroid size and density, the image is divided into 20 to 40 overlapping
patches, ensuring robust detection and minimal false negatives. Inference takes
approximately 2 seconds per image and can be optimized to under 1 second via
batch processing.
Detected coordinates {(x_i, y_i)} are mapped to a user-defined deposition
pattern via affine transformation: [x'_i, y'_i]^T = R * [x_i, y_i]^T + t. A
Traveling Salesman Problem (TSP) based algorithm computes an optimized sequence:
{(x''_i, y''_i)} for i = 1 to N = TSP_Optimize(P_target). Each spheroid is then
transferred via a focused laser pulse, with energy adapted to its diameter.
Robotic alignment ensures sub-20 µm placement precision.
Validation experiments show over 95% detection accuracy with YOLOv11s and
transfer precision below 20 µm. Trajectory optimization reduced operation time
by approximately 40% versus unoptimized paths. The platform supports a range of
plate formats and accommodates spheroids between 50–500 µm. Post-transfer
microscopy confirmed preserved structural integrity.
These results highlight the synergy of AI, photonics, and robotics for advanced
biofabrication. Unlike extrusion or aspiration-based techniques[4, 5], PickCell(TM)
enables high-speed, contactless, and programmable transfers. Its scalability and
modularity make it ideal for automated tissue engineering workflows. Future
directions include viability feedback control and multi-material printing
capabilities[2, 3].
[1] Redmon J. et al. “YOLOv3: An Incremental Improvement.”arXiv:1804.02767, 2018.
[2] Nilsson Hall G. et al. “Laser-assisted bioprinting of targeted cartilaginous spheroids for highdensity bottom-up tissue engineering.” Biofabrication, 2024, 16(4):045029.
[3] Guillemot F. et al. “High-throughput laser printing of cells and biomaterials for tissue engineering.”Acta Biomater, 2010.
[4] Ayan B. et al. “Aspiration-assisted bioprinting of co-cultured osteogenic spheroids.” Biofabrication, 2021, 13(1):015013.
[5] Koch L.et al. “Laser bioprinting of human iPS cells.” Biofabrication, 2018,
10(3):035005.
64057841319