Why in‑process imaging matters now Cell & gene therapy manufacturing runs on short timelines and fragile materials. Batches are small. A minor shift in cell health or contamination risk can ruin an entire lot. Traditional checks happen at intervals and miss what occurs between samples. Continuous, non‑invasive insight is essential. It allows action while there is still time to save a run. Process Analytical Technology (PAT) for real‑time monitoring has long promised this benefit. Industry voices point to faster decisions, reduced variability and earlier alerts when a process drifts. Guidance and expert discussions stress that PAT in cell and gene production must follow a risk‑based framework. Teams need clear critical quality attributes (CQAs), critical process parameters (CPPs) and validation plans. Developers report challenges such as limited datasets, complex methods and the need for stepwise comparability when processes change. Regulatory drivers that shape the approach EU GMP Annex 1 now applies. It calls for a strong Contamination Control Strategy (CCS) and promotes technologies that prevent, detect and control risk. The aim is to move beyond sporadic checks to continuous assurance. The Food and Drug Administration (FDA/CBER) has issued draft guidance on manufacturing changes and comparability for cell therapies including gene‑based products. Sponsors must manage lifecycle changes with analytical comparability first. Non‑clinical or clinical data follow only when needed. This approach fits well with data‑rich, in‑process monitoring. Sector roadmaps highlight in‑line and at‑line analytics, feedback control and data analytics as critical for reliable, lower‑cost production of types of cell products over the next decade. These include immune cell therapies, stem cells, and based gene therapies for rare disorders. Read more: Validation‑Ready AI for GxP Operations in Pharma What “edge imaging” means in practice Edge imaging places compact compute units near bioreactors, closed systems or sterile transfer points. Cameras or microscopes stream images to these on‑site nodes for real‑time analysis. No raw frames leave the suite. Only events and compact features move to higher‑level systems. This reduces latency, protects privacy and simplifies compliance with data‑residency rules. Unlike fixed, off‑line snapshots, edge imaging runs continuously. Systems extract morphology features, density, confluence, debris profiles and motion cues that link to viability or unwanted activation states. Models then raise interpretable alerts such as “unexpected debris signature rising” or “cell cluster size distribution outside expected band.” Each alert includes a short reason and a link to the relevant SOP. Operators stay in control. QA keeps oversight with an audit‑ready trail. For particle risk, flow imaging microscopy remains a known tool. Recent commentary argues for CGT‑specific particulate monitoring during development and manufacturing. This is a reminder to include sub‑visible risk, not just microbiology and viability, in the monitoring plan. Where edge imaging fits in Annex 1 CCS Annex 1’s CCS spans design, people, environment, process and monitoring. Edge imaging supports the monitoring and process pillars by providing continuous, context‑rich evidence. It can sit alongside viable and non‑viable particle counting and environmental monitoring to create a full picture of contamination risk. Links to risk assessments and SOPs must be documented. Industry guidance notes that Annex 1 represents a holistic, risk‑based shift. The CCS acts as the steering instrument that connects measures and shows interactions. Continuous, explainable imaging fits this model. Read more: AI in Genetic Variant Interpretation: From Data to Meaning A practical system blueprint Acquisition and optics. Select optics that match the vessel and process. Bright‑field or phase optics suit adherent growth. Inline holographic units suit stirred systems. Ensure sanitary design, stable illumination and serviceable mounts so technicians can maintain the kit during turnarounds. Edge compute. Deploy on‑premise compute with GPU acceleration at the capture point. Process frames locally. Emit compact events such as “morphology out of bounds” with time, model version, confidence and a thumbnail where allowed. This supports quick review and keeps the record lean. Explainability. Use human‑readable features and overlays so supervisors can verify why an alert fired. This aligns with PAT guidance for cell and gene therapy and supports validation. Validation and change control. Treat the imaging and analytics stack as a validated system. Write a URS tied to process risks. Define acceptance criteria per alert type. Lock model, data and configurations in a signed build for production. Monitor for drift and promote updates through a formal comparability and validation pack. This mirrors FDA/CBER lifecycle thinking for gene therapy trials and manufacturing changes. Security and privacy. Segment networks, sign artefacts and enforce role‑based access. Record only what is necessary for quality and inspection. Retain raw frames only when justified. Continuous monitoring under Annex 1 must not become continuous surveillance. Read more: Predicting Clinical Trial Risks with AI in Real Time Use cases that deliver quick value Early contamination cues. Debris patterns, micro‑bubbles and subtle turbidity changes often appear before failure. Inline imaging can surface these signals hours earlier than manual checks. Teams gain time to redirect, quarantine or investigate before value is lost. Viability and activation insights. In autologous therapy, donor variability complicates set‑points. Imaging‑derived morphology and motion descriptors help identify when cells under‑ or over‑activate. This informs gentler agitation, media refresh timing or temperature adjustments. Media changes and hold steps. Edge analytics can confirm stabilisation after a feed, detect stratification in a hold vessel and flag when oxygenation or mixing deviates. Sub‑visible particle watch. Pairing imaging with particle techniques such as flow imaging microscopy raises assurance, especially in later stages or for drug product handling. These use cases apply across types of gene therapy and types of cell products, from bone marrow‑derived blood cell therapies to chimeric antigen receptor T‑cell products and donor allogenic grafts. They also support regenerative medicine workflows and gene therapy work that uses lentiviral vector systems to deliver a single gene or multiple types of gene edits into cancer cells or other targets. KPIs that matter Detection lead time between the first imaging alert and a confirmed deviation. False alarm rate and operator acceptance. Batch‑to‑batch comparability metrics built from imaging features, used as part of analytical comparability evidence when processes evolve. Release cycle time and deviation rates for the monitored step. SOP adherence time from alert to corrective action. Read more: Generative AI in Pharma: Compliance and Innovation Avoiding common pitfalls Opaque models slow adoption and create audit friction. Keep features interpretable and document operating ranges. Noisy alerts fatigue teams. Combine thresholds with smoothing and require human confirmation for low‑confidence events. Validation left too late invites rework. Build the validation pack as you go. Privacy missteps undermine trust. Edge processing, event‑first records and redaction are safer defaults under Annex 1. How to roll out fast Start in one suite and one step where imaging can reduce a top risk. Define a narrow URS and two or three acceptance criteria that QA and operations co‑own. Capture a small challenge set and run a shadow phase where the system raises alerts but does not drive decisions. Hold weekly triage with operators and QA to adjust thresholds and explanations. When KPIs meet the bar, lock the configuration, complete the validation pack and switch to production with human‑in‑the‑loop confirmation. Plan how imaging events will link to the CCS, how SOPs will reference alerts and how records will appear during inspection. Annex 1 expects a risk‑based, holistic posture. Tie the imaging system into that structure from day one. Read more: AI for Pharma Compliance: Smarter Quality, Safer Trials Read more: AI in Life Sciences Where TechnoLynx fits TechnoLynx designs and delivers edge‑imaging solutions for cell & gene therapy suites. The team builds interpretable analytics that run on‑premise, integrates events with MES/EBR and QMS, and supplies validation artefacts that fit existing templates. Deployments are privacy‑first and inspection‑ready: event‑centric logs, signed configurations and change control that mirrors CGT comparability practices. Solutions support cell therapies including autologous and donor allogenic workflows, based gene therapies, and regenerative medicine platforms. We handle imaging for immune cell products, stem cells, and gene and cell combinations. Our solutions also align with the strict requirements of gene therapy trials and manufacturing for chimeric antigen receptor T‑cells. The result is continuous, explainable visibility over cell health and contamination risk—without slowing work on the shop floor. References BioPharm International (2024) Process Analytical Technologies for Manufacturing Cell and Gene Therapies. Available at: [link] (Accessed: 18 September 2025). CASSS CGTP Roundtable Notes (2024) Developing PAT to Support Advances in Cell Therapy Manufacturing. Available at: [link] (Accessed: 18 September 2025). Cell Manufacturing USA (2023) Roadmap to 2030. Available at: [link] (Accessed: 18 September 2025). FDA/CBER (2023) Manufacturing Changes and Comparability for Human Cellular and Gene Therapy Products (Draft Guidance). Available at: [link] (Accessed: 18 September 2025). European Commission (2022) EU GMP Annex 1: Manufacture of Sterile Medicinal Products. Available at: [link] (Accessed: 18 September 2025). Fluid Imaging (2023) Flow Imaging Microscopy in Cell and Gene Therapy. Available at: [link] (Accessed: 18 September 2025). Image credits: DC Studio. Available at Freepik
