Slide kick

PATHOLOGY:New software from an Irish firm won't replace the pathologist but it could prove an invaluable tool, providing a more detailed diagnosis from cell biopsies

A SMALL PIECE of tissue from a patient can provide doctors with answers to important questions: is the tissue diseased? If it's a tumour, how dangerous is it? And what's the best course of action for treatment?

In recent years, scientists have discovered that biopsy tissue can also contain molecular "biomarkers" that yield more and better clues about a disease and could in some cases even predict how a patient will respond to a specific drug.

Now an Irish company is offering computer-aided analysis to help pathologists quantify key biomarkers when there's more to a biopsy slide than meets the eye.

"The vast majority of any reviews of tissue are still done by eye today and I think pathology as a science really hasn't fundamentally changed in over 100 years of practice," says Dr Donal O'Shea, chief executive of SlidePath.

"It's still predominantly about looking at the shapes and sizes of cells, the architecture of cells in tissue, and then figuring out whether it's representative of disease or normal tissue. And they can do that using conventional microscopy."

But in the last five to 10 years new technologies have been developed to generate high-resolution images of the slides and a greater understanding of the molecular markers those slides may contain, he says.

"Previously, people were taking their tissue samples, processing them the way they normally do - cutting wafer-thin sections, putting them on a normal glass slide and staining it up using conventional stains and then reviewing that with a microscope, whereas now, technologies have evolved to allow the capture of that entire biopsy at high resolution."

Many companies build devices to capture those high-resolution images, but SlidePath is more concerned with what the image itself can tell doctors about the biomarkers it contains, says O'Shea, who lectured in biotechnology at Dublin City University before co-founding the company in 2003.

"We are more interested in the resultant image, how we can measure specific things within those images, to provide pathologists with quantitative data instead of qualitative data," he says.

"It ties in with this concept of personalised medicine, tailoring therapies for patients based on their molecular profile."

In practice, this means the software analyses images of biopsies that have been stained to highlight certain biomarkers, which show up as a particular colour on the slide. The colour grading is where the computer can outshine the human, says O'Shea.

"Humans are extremely good at classification, the difference between a car, a house and a tree, for example. By simply looking at an object, we can apply the class to it," he says. "So when it comes to tumour recognition, pathologists are extremely good at saying, 'That's a tumour, that's normal tissue.'"

"But when it comes to the question of how brown is that, is it dark or light brown - it might sound trivial, but in fact this is predominantly where levels of intra-observer variability arise from."

Enter the computer: "Machines are not so good at classifying, but what they are good at is quantitative measurement, enumerating how many tumour cells there are, or the intensity of staining in the material."

SlidePath has been looking at levels of biomarkers, including a breast-cancer marker called HER-2, that can indicate whether a patient will respond well to the drug herceptin.

Trials have found their software was slightly more accurate than the pathologist at quantifying levels of HER-2 in stained slides, says O'Shea. "It's marginal, but of significance. That's good, because it's important that patients get the best possible therapies."

The company has also developed a system for sending images for rapid analysis, he adds. "It allows pathologists to submit image data effectively and securely over the web and get back reports on the images they generate."

At present, the 25-strong company provides decision-support services for over 70 research labs around the world and is based in DCU's Invent centre, with additional offices in the US and Australia. But the plan is to move to larger premises in Dublin as they expand and tap into the clinical arena, particularly in the US.

"[We are] standardising and going through the process of FDA validation for our experimental algorithms for greater use in the clinical space," says O'Shea.

"We are being quite aggressive in establishing ourselves in the US market and we see a much greater role for the technology in the US over the next 12 to 24 months."

So will the computer replace the pathologist's trained eye? Absolutely not, according to O'Shea. "We do not feel we are better than the pathologists, but we are providing the pathologists with quantitative tools that they didn't have," he says. "The pathologists are here to stay; this is about improving their capability to provide quality diagnoses in a world that's getting ever more complicated through the use of molecular markers as part of the diagnostic strategy."