NIH-funded work to explore replacing diagnostic gold standard with 3D histology

A representative sagittal slice of a zebrafish imaged by 3D x-ray histotomography (A). The color indicates segmented brain regions of the organism. Details of brain cell nuclei used as training data for machine learning-based classification are shown in the inset (A'). In section B, the volumes of increased cellular information are demonstrated. Credit: ProvidedAll Rights Reserved.

UNIVERSITY PARK, Pa. — Development of a new, diagnostic, three-dimensional imaging technique is the focus of a new Penn State-led research project funded by a four-year, $2.6 million grant from the National Institutes of Health. If successful, the method could replace current 2D histology — potentially impacting drug development, diagnostics and basic understanding of how genes and environment define phenotype, an organism’s observable characteristics and traits.

Sharon Huang, associate professor of information sciences and technology, and Dr. Keith Cheng, distinguished professor at the Penn State College of Medicine, explained that 2D histology is the current, state-of-the-art method of studying cells, tissues and organs. In histology, a piece of tissue is sliced into thin, two-dimensional slices, and then evaluated under the microscope. This method utilizes less than two percent of the specimen, and yields no three-dimensional data.

“Here’s an analogy,” said Cheng, a member of the Penn State Cancer Institute. “Let’s say you want a really detailed view of a building, including all its hallways, rooms and windows, engineering and furniture. If you just take one-inch thick 2D slices of a building, you could not build a 3D model or do a virtual walk-through.”

He added, “Similarly, if you really want to see what every cell is like in a living organism, and how they are organized in the living organism, then you need a 3D way of imaging every cell.”

In work published this past summer, Cheng’s lab showed that X-ray "histotomography," built on the basic principles used in CT scans, can be used to create 3D images from even mature and pigmented samples at resolutions one thousand times greater than CT. The volume data can be used to visualize structures in 3D.

In the funded next phase of this project, the researchers will apply principles of biology, physics and machine learning to improve image resolution, processing speed and analytics. First, they will aim to improve the resolution of histotomography and increase the size of samples that can be imaged. They will use zebrafish — which are about the same size as human biopsy samples — as model organisms to define the range of normal phenotypic variation and to inform a range of potential applications. Huang will apply the power of machine learning to create a 3D geometric representations and a 3D visualization of every cell. This will be used as a foundation for mathematically defining the cells that characterize each tissue in an organism, both normal and abnormal, in high volume.

“The advantage of this 3D histology method is that it’s able to image every single cell within an organism,” said Huang. “So when we have the 3D imaging technique plus the 3D analytics and visual framework, it definitely has the potential to replace 2D histology, today’s gold standard for diagnosis.”

Huang and Cheng said they think that this work has the potential to save lives. And beyond disease diagnosis, the researchers’ technique could have applications in other areas, such as examining chemical toxicity and phenotyping full sets of vertebrate genes.

“The final phenotype of any animal will be dependent on how the genotype combines with environmental factors,” said Huang. “We will be able to more precisely define toxicity caused by chemical contaminants of our environment, to help detect crises that would otherwise cause people harm.”

She continued, “If you have zebrafish that grew up in a normal environment versus zebrafish that exposed to chemicals, in the end, we will have the analytical framework to look at every single cell in those samples. Then we can compare the two classes to tell us what the effects of the toxins or chemicals are.”

Cheng and Huang are both affiliated with the Penn State Huck Institutes of the Life Sciences. The Penn State researchers are working in collaboration with Patrick La Riviere, associate professor of radiology at the University of Chicago, on the project.

Last Updated January 22, 2020