Sartorius came to CloudFactory for help with segmentation and annotation of complex cell imagery to create training data for AI cell identification—work that helps scientific discoveries reach patients faster.
Sartorius supports biotech scientists and engineers as they develop and manufacture medication for incurable diseases. By simplifying workflows, enabling faster research, and reducing human error, Sartorius helps the biopharmaceutical industry bring improved medications to patients quickly and safely.
One way the company does so is through its IncucyteⓇ Live-Cell Analysis System, which quantifies cell behavior and captures cellular changes from within the incubator using high definition phase and fluorescence images.
The IncucyteⓇ first extracts and records image data in real time so scientists, researchers, and engineers can observe complex, biological changes as they happen. It then applies innovative machine learning and artificial intelligence approaches to cell analyses to help customers extract more information from samples in simple, easy ways.
“As biologists, we look at images of cells,” says Gillian Lovell, senior scientist. “We get a lot of information about how healthy they are, where the fluorescence is, and how bright they are. The neural network approach simplifies the process of turning images into hard numbers. With IncucyteⓇ, the plan is that everything will be done automatically with one click, taking a lot of the work out, and making the technology accessible to people without training in cell segmentation.”
The IncucyteⓇ Live-Cell Analysis System automates live-cell imaging—a technique that produces far more microscopic images than any human could realistically analyze. Neural networks can handle such work, given enough annotated data to train on. To create a training dataset, the Sartorius team had to identify and characterize individual cells within microscopic images, each containing hundreds to millions of cells in all shapes and sizes.
Although integrated software enables the customer to automatically segment cells within the IncucyteⓇ Live-Cell Analysis System, the segmentation process required user training and bespoke analysis definition for different cell types. By contrast, the neural network method allows one analysis to be performed on a wide range of cell types or experiments with no required user training. Facing the incredible diversity and sheer number of cells the training dataset would require to get their neural network up and running, the Sartorius team knew they couldn’t tackle annotation alone.
Our images are complex and difficult to annotate. The frequent feedback conversations and short iteration cycles are useful in getting the annotation we want.
Rickard Sjögren, senior scientist, led the initiative to find a solution to the data annotation problem. Sjögren says his team compared a number of vendors, but decided to partner with CloudFactory for three reasons:
Sartorius uses CloudFactory’s Data Annotation Solution, a combination of managed workforce plus the best-in-class Dataloop platform. CloudFactory’s data analysts use Dataloop to perform individual cell segmentation and identify key cell types among the thousands in each image. Nicola Bevan, Manager of Cell Imaging Applications, says that the combination of workforce plus platform contributes to the success of projects. “Dataloop makes it easy for us to share annotations and cross check with key scientists,” she says. “The services and resources provided by CloudFactory enable us to be certain that we’re happy with the work.”
For Sartorius, CloudFactory’s trained annotators apply precision labeling techniques to complex microscopy images, for instance, where cells may be touching, out of focus, in large piles, or overlapping. CloudFactory’s data analysts also support cell tracking projects, drawing bounding boxes around cells as they move between frames in time lapse datasets.
With CloudFactory’s services, Sartorius has been able to considerably speed up projects. As one example, the company developed an open-source dataset of more than 5,000 images—1.6 million individually annotated cells—that has been recently published in Nature Methods.
“This dataset was manually annotated with the help of CloudFactory,” says Sjögren. “It’s about 20 times larger than the second-largest dataset in this domain. The project took CloudFactory six months. Using in-house resources, it would have taken us several years to finalize.”
CloudFactory’s work also helped Sartorius step into the machine learning, artificial intelligence, and image analysis spaces, where more-informed research leads to better health outcomes. “By providing a dynamic platform for pioneers and leading experts in our field, we’re bringing creative minds together for a common goal,” says Bevan. “We’re bringing about breakthroughs that can lead to better health for more people.”
We have 10+ years of experience helping our clients focus on what matters most. See what we can do to help your business.
V7 Labs and CloudFactory created an open dataset for medical practitioners to train ML models and streamline triage.
Quality clinical image annotation enables highly accurate AI diagnostics and eases the burden on medical professionals.