Label-free imaging and the Power of AI in Biological Medicines

From fluorescence cell imaging in small molecules to label-free cell imaging & Artificial Intelligence in biological medicines – A personal reflection.

In my previous scientific research, I developed live cell imaging applications based on fluorescence microscopy. Working with high-content screening hardware, I developed integrated cell-based strategies for small molecule drug development. These screens combined the power of acquiring large data sets by high throughput techniques with the ability of multiplexed fluorescence microscopy to collect quantitative data on multiple cell readouts.

Fluorescence microscopy is an enormously powerful technology that is already well integrated into all aspects of small molecule pharmaceutical R&D, but it finds only limited application within bioprocessing. ValitaCell’s Artificial Intelligence (AI) tools bring all the power of staining, with none of the practical drawbacks, to the bioprocessing arena.

Fluorescence imaging 

Fluorescence cell imaging has the capacity to generate targeted signal without homogenization, purification or other fully destructive sample preparation procedures. With the right probe you can target any cell, intracellular location, or metabolic pathway. Conjugating a bright fluorophore to your probe allows you to fully ignore the untargeted background and amplify your specific signal to visualize and quantify structural and functional (sub)cellular components or reactions happening in situ in living or fixed cells. Moreover, multispectral or high-content imaging provides powerful multiplexing capability; combining different dyes with non-overlapping wavelengths you can look at multiple targets at the same time. 

Fluorescence imaging limitations

A limitation of fluorescence microscopy is the potential induction of phototoxic stress, caused by illumination of the reporter molecules in the presence of biomolecules sensitive to light.  Another form of toxicity can be chemical induced by the dye’s perturbation of the physiological function of biomolecules. A more practical concern when using fluorophores is photobleaching, which is a light-induced photochemical alteration of a fluorophore molecule that become unable to fluoresce because of overexposure. Dyes leaking from the cells or subcellular compartments of interest can also be an important challenge. Often it is indeed difficult to keep the necessary chemical equilibrium of the dye over time to measure different conditions during an experiment. 

From a wet laboratory point of view, staining protocols – dye concentration, imaging buffer, incubation conditions and duration, plus illumination conditions (filters, exposure time etc.) – need to be carefully studied and optimized for every dye and cell line independently with several experiments every time starting from scratch because optimal setups are dye- and cell-specific. Multiplexing capability is also limited, depending on the dyes, light source and excitation/emission filters available. Common multiplexing applications typically have up to three and rarely up to five dyes combined.

Fluorescence spillover (also called bleedthrough or crosstalk) and chemical interactions of the dyes must be also carefully considered with independent experiments. When multiplexing different readouts, the imaging of several fluorescent dyes into high throughput screenings can also create acquisition, storage, and processing issues because each dye and fluorescent channel needs to be imaged sequentially, creating long imaging sessions and data storage and handling challenges that require dedicated engineer and IT personnel and infrastructures. 

But probably the most concerning negative consequence is the potential health concerns caused by the chemical toxicity for the laboratory operators handling dyes. A dye that every cell-based biotechnology scientist knows well, Trypan Blue, is routinely used worldwide to measure cell viability. This is a known animal carcinogen, and an experimental teratogen, so companies are now massively investing to find safe alternatives and get this out of their labs.

Label-free cell imaging & Artificial Intelligence 

Label-free cell imaging by AI interpretation of brightfield images is becoming more and more popular to study fast and dynamic biological events in real, unperturbed conditions. Making label-free live cell imaging quantitative is an extremely powerful approach to facilitate cellular characterization, maintaining all the great properties of fluorescence microscopy but skipping all the drawbacks of using dyes and fluorophores. In the table below, you will see my views on some of the benefits of label-free imaging when compared to fluorescence imaging. 

Label-free microscopy is a quick and easy approach to cell imaging that reduces workflow complexity (thereby minimizing potential for error and simplifying automation) and saves time when analyzing precious samples by a non-destructive approach.

Label free microscopy does need to be coupled with strong computational capability. Artificial Intelligence, in particular Deep Learning, is often able to extract rich insightful information from label-free images of cells. The simple idea is that algorithms can be trained to identify subtle patterns present in cell morphology, pixel texture, cell-based light refraction and other information present in the images that sometimes are visible but not quantifiable by humans. Computers have shown they can dramatically outperform human computation capability. Often AI can identify robust patterns that are completely undetectable by the eye. 

Most likely, not every fluorescence imaging application will be moved to label-free cell imaging and AI. Because our still limited understanding of AI, in mission critical settings an AI output will not always be an acceptable proxy of a direct measurement. Yet where speed, safety and simplicity are priorities, the gain is significant.

Imaging in Bioprocessing 

If we look at the Biologics and Cell Therapy sectors I currently work in with ValitaCell, imaging has fewer targets compared to the small molecule pharmaceutical industry where it is used to explore a plurality of cell modalities behind disease models. The main standard imaging applications in bioprocessing are: cell viability assessment by Trypan Blue exclusion or fluorescent dyes like propidium iodide and calcein-AM; cell counting, growth and confluence assessment; and assurance of clonality coupled to cell printing or limiting cell dilution in Cell Line Development. Some forms of qualitative morphological characterization for QC purposes are also used more informally in the Cell Therapy space. For certain there is an expanded set of targets – the protein processing machinery looking specifically at the endoplasmic reticulum, for example – that would be measured in Cell Line Development laboratories with rich actionable results if only would be easy and quick enough to do it in a high throughput screening modality.

None of these applications is yet exploiting AI other while crude and simple object segmentation following straightforward image binarization to count cells and/or pixels are the most common image processing approaches. This does not even scratch the enormous potential of AI. When your final products are the cells themselves (i.e. stem cells for Cell Therapy), or the cells are the factory producing your product (CHO in mAb manufacturing), you cannot afford to compromise these because of cytotoxicity events caused by the staining. 


CellAi is ValitaCell’s Artificial Intelligence image analysis software  that enables users to automate their image analysis and quickly extract rich cell insights from simple label-free images. The development and commercialisation of this technology was one of the key reasons I joined the business. This game-changing technology based on Deep Learning enables scientists to analyze cells quickly, in-depth and without destruction. CellAi has a wide range of image applications including nuclear virtual cell staining in MSCs, and predicting cell viability in CHO cells. 

Read more info about CellAi here and get in touch with us to develop your customized CellAi application tailored to your specific lab needs.

Dr. Alessandra Prinelli
Eligio Iannetti, PhD

Eligio Iannetti holds a PhD in cell biology and drug discovery and has extensive experience in cell profiling and drug testing. Eligio worked in a number of biotech startups across Europe before joining ValitaCell focusing on innovation and product development. In his current role as Business Development Manager, Eligio's energy and drive is helping our global customers to accelerate the launch of new, innovative medicines.

Want to talk about our products?

Hi, I’m Eligio, Business Development Manager at ValitaCell. Fill in the form to the right and I’ll get in touch with you soon!