ICCS 2026 lists a label-free foundation model for automated viability in imaging flow cytometry

Waters Biosciences lists ICCS poster 142, A Label-Free Foundation Model for Automated Viability Assessment and Functional Assays in Imaging Flow Cytometry. The public conference page verifies the project title and its inclusion in the program, but does not publish the underlying model-development dataset or performance table.

CFCM explainer

Image-enabled flow event → label-free image features → foundation model → viability / functional-assay output → human-defined QC and decision layer.

Source/credit: CFCM conceptual schematic based on the Waters ICCS 2026 poster listing. Alt text: imaging-flow event processed by a foundation model to produce a viability or functional output with a downstream QC layer.

AI Evidence Tracker

Instrument harmonisation: not disclosed.
Fluorescence calibration: not disclosed; the model is described as label-free.
Metadata completeness: low in the public conference listing.
External-site validation: not disclosed.
Domain-shift evidence: not disclosed.
OOD/failure handling: not disclosed.

The terms foundation model and automated viability are not evidence of generalizability by themselves. As checked on 6 Oct 2026, the public conference page does not provide sample count, training/validation split, reference viability method, accuracy metrics, instrument diversity, external-site testing, domain-shift analysis or out-of-distribution rejection behavior. CFCM therefore labels this item WATCH.

Imaging flow is naturally suited to representation learning because it generates large numbers of cell images linked to cytometric context. But model transfer can fail when optics, sample preparation, morphology or acquisition settings change. The evidence threshold should therefore be multicentre and failure-aware.