Drug developers could benefit from the alliance to enable standardized central pathology review of certain biomarkers in oncology clinical trials and quantify tumor heterogeneity.

“Tumors can have a lot of heterogeneity. They can have different clonal populations,” Mukhopadhyay said. “So the molecular profile you are extracting can be very heterogeneous, which can, in turn, can affect responder risk to the drug with categories of responder, non-responder, and hyper-progressive. All of these are important for drug development.”

In any event, the rise of digital pathology could have a sweeping impact on the field of medicine at large. “For the past 100 years, pathologists have been looking at glass slides,” said Singh, a pathologist himself. “They would put a tissue slide under the microscope, look at the tissue and try to come to a diagnosis.”

The digitization of such slides in the past decade, however, is a gamechanger, Singh noted. “Instead of looking at a slide in a microscope, pathologists can see the same image in three dimensions on a computer screen,” he explained.

The sheer volume of digital slide data gives rise to increasingly powerful machine learning models. “Let’s say you have 500 breast cancer cases with 500 corresponding slides,” Singh said. “Experts at 4D Path can apply machine learning techniques to develop models that can predict the diagnosis.”

Such models can predict patient prognosis and make treatment recommendations for a given patient. In addition, 4D Path’s platform can also make predictions related to proteins such as estrogen receptors, progesterone receptors and the HER2/neu status.

“If a patient with breast cancer has a HER2/neu status of 3+, they could be given a specific drug,” Singh said. “But if it is 1+ or 2+, they might be given a different treatment.”

“If these algorithms are available for the pathologist to run when they’re looking at the slides on the computer, then they can actually help the clinician decide much faster what drug will be useful for the patient,” Singh concluded.

Rather than using standard AI, 4D Path uses statistical physics to uncover “hidden dynamics” in a frozen snapshot, Mukhopadhyay said. Deciphering information in those hidden dynamics can inform “what targeted drug you should provide or how a patient could respond,” she added.

The platform can also suppress noise obscuring the slide image.

The FDA or the EMA have not approved the 4D Q-plasia OncoReader Breast software for primary diagnosis. The company notes that the software is currently for research use only.

How 4D Path and PathPresenter aim to further AI-enabled digital pathology for breast tumor profiling

Contact supplier

Stay in the loop!