Think Smart Automation. Think AI.

The LabSentry AI machine vision system can enhance the reliability of unattended liquid handling by enabling real-time error detection and mitigation.

Using deep learning, the system is trained to monitor critical liquid handling components, recognize common operational patterns and identify anomalies similarly to the human eye.

LabSentry AI actively integrates with equipment software to apply corrective actions to maximize productivity.

Brochure

Determining What to Monitor

Machine vision, like the human eye, can detect far more subtle cues than any single sensor or sensor array alone. Deciding where to implement machine vision — in other words, where to position the cameras — is critical.

On PromoChrom’s SPE-03 automated SPE system, the syringe pumps are responsible for 95% of an extraction method - accurately transferring liquids between solvents, samples, and cartridges. This makes the syringe pumps a powerful indicator for potential anomalies, such as clogging, solvent depletion, or insufficient sample loading. By focusing machine vision on the syringe pumps, we achieve a straightforward yet highly effective way to detect most common errors in real-time.

Why LabSentry AI?

Achieve Peace of Mind with AI

AI-Driven Monitoring
Catch anomalies as they happen and avoid costly reruns with fleet-wide insights.
Remote Access & Control
Oversee extractions from anywhere with intuitive mobile and desktop tools.
Smarter Planning
Shift from reactive troubleshooting to proactive mitigation and predictive workflows.
Streamlined Workflow
Maximize uptime and enhance reproducibility with seamless automation.

Integrated with PromoChrom Pilot, LabSentry AI continuously monitors automated SPE workflows, detecting anomalies in real time and helping laboratories prevent costly reruns.

Discover PromoChrom Pilot

Object Detection

A computer vision model was then trained to detect the pump syringe, plunger, and liquid level. To improve its versatility, the training dataset was augmented with variations in lighting conditions and sample colors. The trained model achieved a mAP50 score of over 99%, indicating highly accurate object detection.