Project · 2023

Germ Cell Sorter

Designed and built end to end, from the chip to the control loop

MEng final-year project · Imperial College London

For my MEng final-year project I designed and built a complete microfluidic fluorescence-activated cell sorter (µFACS), on my own, from the chip up. The primordial germ cells that fertility research depends on are rare and fragile, and the commercial machines that sort them cost $80,000 to $150,000, run on complex protocols, and shear delicate cells to death. My sorter set out to do it gently and cheaply: focus the cells in a custom chip, single out the GFP-tagged germ cells with a bespoke dual-channel optical system, and sort them with a soft pressure pulse fired by real-time machine vision. I built every subsystem myself, the microfluidics, the Zemax-optimised optics, the electronics, and the LabVIEW control, and got each one working.

System schematic: LED illumination, focusing chip, objective, beam-splitter to dual cameras, LabVIEW control and a pressure pump — The µFACS system: syringe pumps focused the sample in the chip, a telecentric optical path captured fluorescent and brightfield images, LabVIEW classified each cell, and a DAQ-driven pressure pump sorted PGCs into the collection outlet.

The rig

Lab bench: an inverted microscope with LED illumination and a camera imaging the microfluidic chip on the stage, with the live LabVIEW view on a monitor — The optics and detection side: an inverted microscope with LED illumination imaged the chip on the stage, with the live brightfield and fluorescence view running in LabVIEW on the monitor. The board in front drove the sort actuation.

Lab bench: a Harvard Apparatus syringe pump feeding the acrylic manifold and microfluidic chip on the microscope stage — The fluidic side: a Harvard Apparatus syringe pump (set to 40 µl/min here) drove the sheath and sample flows through the acrylic manifold and into the chip under the microscope.

What it does

Built end to end, solo, as a final-year undergraduate: the microfluidic chip, the optics, the electronics, and the real-time control software.
Focuses fragile cells gently in a custom laser-cut chip using 3D hydrodynamic focusing, shrinking the imaging region while keeping shear low.
Images brightfield and fluorescence at once through a 4f telecentric relay I optimised in Zemax and mounted via a 3D-printed, lathe-tapped adapter, so GFP-tagged germ cells stand out from everything else.
Closes the loop in LabVIEW: real-time machine vision confirms a target, then a DAQ-driven pressure pulse redirects it into the collection outlet.

The problem

Primordial germ cells are central to studying the germline and to fertility research, but they are rare, fragile, and arrive mixed with somatic cells that overgrow them in culture. The two cell types are too similar in size for passive methods like filters to tell apart, so sorting them needs active detection of each cell. The catch is that the machines built for that, industry-standard flow cytometers, are costly, run on complex protocols, and are rough enough on cells to kill the fragile ones. The goal was to sort the germ cells actively but gently, at a fraction of the cost and complexity.

What it demonstrated

Tested with 10 µm fluorescent and transparent microspheres as cell surrogates, each subsystem did its job: the chip focused the sample stream, the optics cleanly told fluorescent targets from transparent particles, and a pressure pulse redirected the central streamline into the collection branch on cue. It stopped short of a production sorter, quantifying sort purity and yield was the clear next step, and another student later carried the chip design and sorting evaluation forward. As a solo final-year project spanning microfluidics, custom optics, electronics, machine vision, and real-time control, it is where I learned to take an instrument from a blank chip to a working loop.

A closer look

Labelled 3D model of the focusing microchannel: two Z sheath inlets and two Y sheath inlets squeeze the sample into a central core, which a bifurcation splits into collection and dump outlets — The microfluidics. The chip I designed for 3D hydrodynamic focusing: two pairs of sheath inlets pinch the sample into a central core for gentle, single-plane imaging, then a bifurcation splits collection from waste. Green marks the fluorescent targets, yellow the rest.

Two heatmaps across the channel cross-section: velocity magnitude peaks in the centre, while shear rate is lowest in the centre and highest at the walls — The analysis. Velocity (left) and shear rate (right) I computed across the channel cross-section at the prototype flow rate. The focused sample rides the dark, low-shear core in the centre, which is exactly how the design stays gentle on fragile cells.

Ray-trace layout of the 4f telecentric optical relay: object plane, collector lens, beam splitter, aperture iris, 500 nm filter, detector lens, and image plane, with on-axis and off-axis ray bundles — The optics. The imaging relay I designed and ray-traced in Zemax: a 4f telecentric path that carries one view through a beam splitter, aperture, and 500 nm filter onto two cameras, so brightfield and fluorescence come off the same image at once.

Dimensioned engineering drawing of the custom microscope camera-port adapter, with front, side, and section A-A views and an SM1 thread callout — The optomechanics. No off-the-shelf part joined my optics to the microscope, so I designed one: an SM1-threaded camera-port adapter, 3D-printed then bored and tapped on a lathe to hit the fit.

UML state-machine diagram of the LabVIEW control program, with a Main Loop (Idle, Learn background, Save images, Sort) and a separate Pump Loop running bang-bang pressure control — The control. The LabVIEW software I wrote ran as a queued state machine: a Main Loop handled imaging, detection, and sorting, while a parallel Pump Loop held pressure with bang-bang control over the DAQ card, so detection stayed responsive while the pump tracked its setpoint.

Stack

SOLIDWORKS · CADMicrofluidicsOptical design (Zemax)LabVIEWDAQ · real-time controlFluorescence microscopy