Skip to main content
Premium Trial:

Request an Annual Quote

GE Teams with Harvard on Image-Analysis Software to Extend Range of IN Cell Analyzer


GE Healthcare has enlisted the help of a Harvard bioinformatics team to expand the capabilities of its IN Cell Analyzer cellular imaging system, and hopefully gain an edge over its competitors in the cell-based assay market at a time when high-content screening is gaining a greater foothold in pharmaceutical R&D and academic labs.

Under the terms of a collaboration announced last week, the Harvard Center for Neurodegeneration and Repair's Center for Bioinformatics will develop image-analysis software to extend the range of assays that can run on the IN Cell Analyzer platform. GE Healthcare will have the right of first refusal to commercialize any software developed in the open-ended project, but the company does not yet have any definite plans to do so.

"We'll evaluate what comes out of [the collaboration], and if it's commercially viable, then of course we'll take it up and develop the application and the software," John Burczak, vice president of R&D at GE Healthcare, told BioInform. Burczak added that GE's own software for the platform is "modular," so if the Harvard developers create an application with limited commercial appeal, "it might be something that we pass on, but we want to ensure that the links are there, that the application can run."

"There are lots of ways of making [the software] available," Adrian Ivinson, director of the HCNR, told BioInform, "but the idea is that it's made available to as many people as possible, and whether that's through GE instrumentation or through new software packages, that remains to be seen." Ivinson stressed that the collaboration is "a fairly long-term project."

"There are lots of ways of making [the software] available, but the idea is that it's made available to as many people as possible, and whether that's through GE instrumentation or through new software packages, that remains to be seen."

Rich Fisler, director of biotech consulting firm Beachhead Consulting, said that GE Healthcare stands to benefit competitively from the collaboration even if it doesn't commercialize the software that comes out of it. "It shows the industry that they're pushing the envelope," he said. "They're not just sitting back on their heels and saying, 'Let's wait for someone else to do it.' They're saying, 'Let's really drive the development of applications in very pharmaceutically relevant areas like neurodegenerative disease.'"

High-content screening is gaining a foothold in pharmaceutical R&D and academic research because it provides information about the phenotypic effects of chemical compounds or biological agents, such as siRNA, in live cells. But the approach, in which a typical assay may screen thousands of compounds against millions of cells, relies on sophisticated pattern-recognition algorithms to extract information from very large sets of raw images. These algorithms must be able to identify — and, if possible, quantify — the movement of fluorescent cellular markers, variations in cell morphology, and other subtle changes in order to determine the effects of a compound or biological agent.

One challenge for vendors in this field is that new assays are tightly coupled to the software, so a one-size-fits-all approach to image analysis is not possible. An algorithm to extract cell morphology information about cancer cells, for example, would not work for neurons, which have a completely different shape and structure.

In addition, researchers are looking to run cell-based assays of increasing complexity. The assessment of whether cells are alive or dead, for example, is "pretty straightforward" for most image analysis software packages, Ivinson said. However, he added, "We're hoping to go many levels further than that, where we're looking for much more subtle changes in cells that have been incubated in a variety of different chemicals. That will be much more tricky, and it won't be an all-or-nothing effect. It will be an effect that we try and quantify, and that can become difficult.

"The metabolism of a live cell is very complicated, and the appearance of the cell and the cellular structures is complicated," Ivinson said, "so it's quite a challenging task. Rarely is it an all-or-nothing response that you're looking for."

Sergey Ilyin, bioinformatics group leader at Johnson & Johnson Pharmaceutical Research & Development, said that new experimental applications in pharma are driving demand for new image analysis methods. His group, which uses Cellomics' high-content screening platform, is moving to assays that run on multiple cell types at a time in order to "better mimic the physiology" of biological systems. This approach is expected to serve as a bridge between high-throughput screening and time-consuming in vivo studies using animal models, but "extracting data from these types of experiments is still difficult," he said.

The "big challenge," Ilyin said, "is learning how to interpret the results automatically."

"It's not that [HCS platform vendors] haven't paid attention to the software, but part of it is understanding how all that biological variability can be condensed and quantified to an answer without looking at the images. Nobody wants to look at images — they want the information."

Beachhead's Fisler, who has been tracking the high-content screening market for several years, noted that "software issues" hampered early sales of instruments like GE Healthcare's IN Cell Analyzer and Cellomics' ArrayScan and KineticScan. Many of these initial software limitations have been addressed, Fisler said — a development that may be partially responsible for a pickup the sector is witnessing this year. Fisler said that more instruments have been sold for cell-based screening so far in 2005 than in all of 2004. But biological complexity presents an ongoing challenge for the field.

"It's not that [HCS platform vendors] haven't paid attention to the software," Fisler said, "but part of it is understanding how all that biological variability can be condensed and quantified to an answer without looking at the images. Nobody wants to look at images — they want the information."

GE's Burczak agreed that "the biology is going to run ahead of the software development." This fact drove the company to release its IN Cell Developer Toolbox in June [BioInform 06-20-05], which includes a collection of image-analysis routines that end-user biologists can configure to create algorithms for new applications. The company has seen "very strong" response to the Developer Toolbox so far, he said.

Nevertheless, the partnership with Harvard is expected to "extend the flexibility" of the IN Cell instrument, Burczak said, "in particular, in this case, with applications in neural biology."

As interest in cell-based assays continues to grow, it's likely that image-analysis software will be a key differentiator among competing platforms (see sidebar for a look at some available packages). Fisler noted that the instrumentation is not likely to change over the next several years, which will drive "more software and reagent development" as vendors look to gain an edge.

Burczak agreed that "software development is very important to stay competitive" in the cell-based assay market. The image data that the instruments are generating is large and complicated, he said, "and to get the most out of it, it's going to take a lot of software."

Harvard's Image Analysis Toolbox

The HCNR Center for Bioinformatics, led by Stephen Wong, is working on a number of projects that may be applicable to the IN Cell Analyzer.

Wong said that his bioinformatics team has been developing image analysis methods to keep pace with new assays that HCNR biologists are designing. For several projects — one tracking the cell cycle in cancer cells using time-lapse microscopy, another running whole-genome RNAi screens in Drosophila cells, and another screening Alzheimer's disease drug candidates against neuronal cells — the HCNR researchers "tried all the existing products, and none of them worked," Wong said. "A lot of cell-based assays, or neuron-based assays, will be coming up with new applications, and a lot of the existing algorithms and tools are not helpful."

Image-Analysis Software Packages
for Cell-Based Assays
Beckman Coulter: CytoShop
BioImagene: CellMine
Cellomics: BioApplications
Definiens: Cellenger
Evotec Technologies: Acapella
GE Healthcare: IN Cell Developer Toolbox
Harvard Center for Neurodegeneration
and Repair: Cell IQ
Imstar: Pathfinder
Molecular Devices: MetaXpress
Vala Sciences: Cell-Image Analysis
Whitehead Institute: CellProfiler

Wong said that biological imaging of cells and molecules has lagged far behind that of medical imaging, such as MRI, PET, and CT, but the increasing interest in cell-based assays should help the field catch up a bit. Cellular and molecular imaging "is not as savvy as medical imaging in terms of post-processing," he said. However, biological imaging has historically been limited to studies of one or two cells at a time, so the recent movement toward "millions of cells in parallel, looking at thousands of features at the same time" is attracting computer scientists, like Wong, who previously worked in the medical imaging field, and are looking for a new challenge.

The primary limitations of current image-analysis methods are in "accuracy and specificity," he said. "I came from the medical domain, and in the medical domain we're looking for 99.9 percent accuracy and specificity. But biologists are pretty happy with a 70-percent automated solution."

HCNR's Ivinson said that there's still quite a bit of room for improvement in the field of biological image analysis. "We spend a lot of time in the neuroscience community generating very sophisticated images … but we're not nearly as sophisticated in analyzing those images. For the most part, we still simply eyeball them," he said.

"What we're tying to do, then, is to say, 'What is it that typically the researcher does manually?' And can we define that, can we quantify that, and can we teach the software to do that? This isn't a question of just tweaking it. It's really a question of using the instrument to give us the data, but then going away with the data, off the instrument, and doing a lot of programming and research and development."

GE's Burczak said that the organizations are still in the "definition phase" of the project, so specifics regarding particular assays or applications that will run on the IN Cell "are still being worked out."

— Bernadette Toner ([email protected])

Filed under

The Scan

Genetic Risk Factors for Hypertension Can Help Identify Those at Risk for Cardiovascular Disease

Genetically predicted high blood pressure risk is also associated with increased cardiovascular disease risk, a new JAMA Cardiology study says.

Circulating Tumor DNA Linked to Post-Treatment Relapse in Breast Cancer

Post-treatment detection of circulating tumor DNA may identify breast cancer patients who are more likely to relapse, a new JCO Precision Oncology study finds.

Genetics Influence Level of Depression Tied to Trauma Exposure, Study Finds

Researchers examine the interplay of trauma, genetics, and major depressive disorder in JAMA Psychiatry.

UCLA Team Reports Cost-Effective Liquid Biopsy Approach for Cancer Detection

The researchers report in Nature Communications that their liquid biopsy approach has high specificity in detecting all- and early-stage cancers.