Sartorius uses AI to quantify live-cell images over time

Sartorius applies AI inside a live-cell experiment: images become segmented cells, viability classifications and time courses. Its Incucyte offer is most useful when those measurements answer a specific biological question and remain traceable to the underlying images.

01 / ProductAI is part of a broader life-science toolchain

Sartorius supplies tools for laboratory research and biopharmaceutical development and manufacturing. Its relevant AI offering here is Incucyte live-cell analysis, rather than a general-purpose model service. The useful connection is between controlled image acquisition and the software that converts those observations into repeatable measurements.

The AI Cell Health module classifies live and dead cells from label-free images, with optional further classification using morphological or fluorescence parameters. Sartorius's research explanation describes neural networks for segmentation and live/dead classification. Those are vendor descriptions of the approach; they do not establish accuracy on every cell type or experimental condition.

Incucyte is also an instrument workflow. The S3 system acquires images while cells remain in an incubator and supports repeated observations of cultures. A laboratory therefore evaluates the imaging conditions, compatible vessels and analysis configuration together. Buying an analysis module does not turn arbitrary historical microscopy images into an equivalent validated Incucyte experiment.

02 / AudienceUseful when the time course changes the interpretation

A suitable reader is investigating how a perturbation changes a cell population over time. An endpoint measurement can hide whether cells declined rapidly, recovered or changed morphology before a viability signal became apparent. Repeated imaging offers a way to ask those questions, provided the sampling interval and analysis definition match the biological process.

Core facilities also have a practical interest: several researchers may want to analyse comparable experiments without each creating a separate image-processing pipeline. Shared acquisition and analysis conventions can help, but common software settings are not evidence that different cell models behave identically. A facility should document which applications it has actually evaluated and which require method development.

For organising experiments, samples and research records, Benchling addresses a broader laboratory-data problem. For a company using large-scale experimental biology in its own discovery platform, Recursion provides a different context. Neither comparison makes Incucyte a substitute for an experiment registry or a drug-discovery service; it helps separate the measurement layer from the surrounding research programme.

03 / WorkflowProposed workflow: separate cell detection from biological meaning

Consider a proposed study comparing a cell line under several treatment conditions. First specify the outcome: cell count, proportion classified as viable, or change in a defined morphology. Keep the unit of analysis clear. A large number of imaged cells within one well does not automatically provide the same experimental independence as multiple separately prepared wells.

Check instrument and module compatibility before designing the plate. The AI Cell Health page lists a GPU co-processor requirement and a software-version dependency. Ask Sartorius to map those requirements to the laboratory's actual controller and current software. A platform-level AI description cannot determine whether an older installed instrument is ready for the intended analysis.

Acquire a small representative experiment and inspect the segmentation masks against the original images. Examine crowded fields, sparse cultures, detached cells and changes in focus or illumination. Separate failures to identify a cell boundary from disagreements about whether a detected cell belongs to the live or dead class. These errors affect different downstream measurements and need different remedies.

If the study concerns a distinctive morphology, evaluate the Advanced Label-Free Classification module. Its documented workflow uses segmented objects, morphological features and a classifier trained with control wells. Choose controls that represent the states being compared, then assess additional wells that did not define the classifier. A convenient visual separation of controls may not persist across the full treatment range.

Review the entire time course before selecting the final summary measure. Preserve the images, masks, analysis settings and exclusions alongside the exported results. Investigate a sudden discontinuity rather than assuming it reflects biology: an analysis setting or imaging problem can also change a curve. This workflow is proposed editorial guidance; Sequenced has not performed a Sartorius assay or verified any experimental outcome.

04 / PricingThe module price is only part of the experimental cost

The current product pages use a request-for-quote route. No universal public price for the complete Incucyte AI Cell Health configuration was verified. A useful quotation should identify the instrument or upgrade, GPU requirement, named software modules and any training or service commitments that apply to the proposed study.

The Advanced Label-Free Classification listing says it requires either Cell-by-Cell Analysis or AI Cell Health Analysis. That dependency matters when comparing an additional module with a complete new installation. Ask whether the quoted configuration contains the prerequisite software already owned by the laboratory, and avoid paying twice for an entitlement carried by an existing instrument agreement.

The S3 page describes unlimited free user licences for remote network access. That should not be read as a promise that every specialist analysis module is free. Separate user access from instrument capacity and application entitlements. Also account for compatible vessels, reagents where used, storage, experimental preparation and the scientist's time reviewing uncertain measurements.

05 / DistinctionsLabel-free analysis preserves room for other measurements

The specific appeal of the AI Cell Health route is estimating viability from image information without requiring a viability label for that classification. In a proposed experiment this can leave fluorescent channels available for other measurements. Whether that improves the experiment depends on the cell system and the question; label-free output still needs comparison with suitable reference evidence.

The S3 description places acquisition within a stable incubator workflow. This is a meaningful distinction from repeatedly removing cultures for separate imaging sessions. The analytical opportunity is a connected sequence of observations rather than a set of unrelated snapshots. A laboratory still needs to choose a cadence that is informative without producing unnecessary data and review burden.

Sartorius also separates a supplied viability classifier from application-specific morphology classification. That gives researchers a useful conceptual distinction: one result estimates a predefined state, while another depends on how the experiment defines its classes. Keep both definitions explicit in reports. A morphology class named after a treatment condition should not quietly become a claim about a biological mechanism.

06 / QuestionsCell-model coverage and classifier transfer need evidence

The AI research page describes continuing work on more complex cell models. Research activity is encouraging, but it does not establish that every model is supported by a purchased module. Ask for evidence relevant to the cell line, culture conditions and imaging configuration, particularly when the experiment differs substantially from the examples used to present the product.

A classifier can behave differently when cell density, medium, treatment or morphology changes. Design the evaluation to include the conditions most likely to challenge the measurement. Do not remove difficult fields solely because they disagree with the expected result. Record an exclusion rule before reviewing the treatment comparison, and retain enough original data for another scientist to audit the decision.

Finally, keep research measurements distinct from patient decisions. This blueprint evaluates a laboratory research workflow and makes no claim of diagnostic authorisation or clinical performance. The public sources do not establish a universal sensitivity, a guaranteed reduction in researcher time or a treatment outcome. A successful purchase is one that supports a defensible measurement in the laboratory's specific experiment.

07 / DecisionChoose a measurable assay improvement

Sartorius merits attention when a laboratory needs repeated, interpretable observations of living cells and can define how the AI output will be checked. Start with one assay where the time course matters, resolve the installed hardware and module dependencies, and compare the proposed measurements with an appropriate reference. Extend the workflow only after its limits are understood.