Robotics ｜ Supernatant extraction
AI finds the supernatant-sediment interface and extracts automatically
A robot arm holding a pipette × AI interface detection from camera images × estimation and logging of the supernatant volume. The vessel is back-lit and photographed, and AI detects the interface between supernatant and sediment. All a person specifies is the source and destination wells; how deep to go is decided automatically for each sample.
Demo video
The sequence runs from specifying the source well, imaging, AI interface detection and aspiration, through to dispensing into the destination and logging.
* Recording of the actual hardware, played at 2× speed. No audio.
The problem
Collecting supernatant means judging how far to insert the tip so the sediment is not stirred up. By hand, that judgement varies from person to person, so the volume collected and the degree of carry-over vary.
The interface height also differs between samples. Aspirating at a fixed depth leaves some samples under-collected and draws sediment into others. The more samples there are, the harder it becomes to keep conditions consistent across all of them.
What we built
Interface detection, aspiration, dispensing and logging are automated end to end.
01 Specify source and destination
On the plate map you specify which well to draw from and where to move it. No depth setting is needed.
02 Imaging
The vessel is lit from behind and photographed from the side.
03 Interface detection
From the captured image, AI detects the interface between supernatant and sediment and estimates the supernatant volume. The estimate comes with a range — its uncertainty.
04 Aspirate
The pipette descends to just above the detected interface and draws liquid without touching the sediment layer.
05 Dispense & log
The liquid is dispensed into the destination, and the estimated supernatant volume and the processed well are logged.
Because the interface position is re-measured for every sample, collection stays consistent even when the amount of sediment differs between samples.The estimated supernatant volume is shown with its uncertainty and kept in the record. The screen shows the plate layout and the camera view side by side, so progress through the wells is visible at a glance.
Where it applies
Manufacturing & quality control
Recovering supernatant after a separation step, with every processed well logged. Because the interface is detected per sample, collection conditions stay consistent across lots even when the amount of sediment varies, reducing scatter in analytical values caused by collection variability.
R&D
High-volume, fixed-procedure preparation such as recovering supernatant after centrifugation, and time-course sampling that draws small amounts from the same well at regular intervals. Even when sedimentation progresses differently between samples, nobody has to stand over it.
Swappable components
Vessel (vial → well plate, centrifuge tube) / aspiration (pipette → syringe, pump) / detection target (supernatant-sediment interface → a two-phase boundary, a liquid level). A configuration that keeps your labware and automates only collection and logging is also possible.
Summary
Automating supernatant collection with AI interface detection from camera images lets the system match the interface height of each individual sample. Stirring up sediment and operator-to-operator differences in judgement are both reduced, and the estimated supernatant volume is kept in the record with its uncertainty.
* This system is a proof-of-concept demonstration. Specific configurations, specifications and implementation details are discussed individually on enquiry.
Tell us about the experiment you want automated
We will ask about the target operation, the instruments you have and your sample throughput, then propose where to start.
- Unattended weighing, dispensing and sample exchange
- Control systems built around your existing instruments
- PoCs that connect automatically acquired data to analysis and experimental design