Smart General Defocusing Particle Tracking: Machine-learning-based 3D cell tracking in acoustofluidic devices

H.C. Ørsted Fellows Programme – H2020-MSCA-COFUND-2015, grant no. 713683
Period: 15.12.2018 - 14.12.2019. External funding: 0.5 MDKK
PI: Assoc.prof. Massimiliano Rossi, DTU Physics
Co-PI: Prof. Henrik Bruus, DTU Physics

About the project
This project intends to combine machine learning and General Defocusing Particle Tracking (GDPT) to track the 3D position and orientation of cells in acoustofluidic devices for in vitro cell-testing in the framework of immunology-oncology research. 3D cell tracking is fundamental in this field for three different reasons: validation of the theoretical and numerical models used in the design of the devices, closed-loop feedback control of acoustofluidic devices for active manipulation of cells, and high-resolution measurements of cell-cell interaction. Compared to existing methods, the proposed approach possess all the key features to work successfully in this field: is a single-view, single-frame approach, can achieve fast post-processing time, works with conventional microscopy and is suitable for non-expert users, as the ones in clinical environments.
     The starting point is the GDPT, a 3D particle-tracking method recently developed by the applicant and coworkers at the Bundeswehr University Munich. GDPT is based on a look-up table approach and works for particle images of arbitrary shapes. The GDPT principle can be extended to indefinite number of particlerelated quantities such as size, orientation, and shape. However, the consequent fast-increasing complexity of the look-up table and of the algorithms required to handle it makes this task very difficult. The key idea of this proposal is to use machine and deep learning to solve this problem. The main objective of this project is the selection of the right classification methods and the assessment of achievable resolution and uncertainty. The objective will be achieved by optimization and testing of different machine learning approaches on synthetic and experimental images. Experiments will be carried out by collaborators at the Biomedical Electronics Group, TranslaTUM, TU München.
     The success of this project will provide a 3D cell tracking method with potential for being a widespread standard technique for experimental research in acoustofluidics and in vitro cell testing.
 
Collaborators during the project
Prof. Oliver Hayden , TU Muinchen (Germany)
Dr. Rune Barnkob, TU Muinchen (Germany)


References

R. Barnkob, C. J. Kähler, and M. Rossi,
General defocusing particle tracking,
Lab Chip 15, 3556 (2015). (doi 10.1039/C5LC00562K)

GDTPlab – how to get it
Institut für Strömungsmechanik und Aerodynamik,
Univeristät der Bundeswher (Germany),
www.unibw.de/lrt7/gdpt-1/gdptlab-how_to_get_it,
 

https://www.staff.dtu.dk/bruus/ExternalFunding/SmartGDPT
7 MARCH 2021