HyperBird, a cutting-edge imaging platform for observing microscopic traits in plants, can detect grape diseases – such as powdery and downy mildews, which cost vineyard owners billions in annual losses globally – days before physical symptoms become visible. 

The advancement will potentially allow plant breeders to speed development of disease-resistant grape varieties, cue growers to target fungicide use early in a disease’s progression and provide vineyard workers better and earlier information for crop management.

Credit: Ryan Young/Cornell University

While regular cameras record three bands of wavelengths (red, blue and green) within the visible light spectrum, hyperspectral imaging records hundreds of narrow bands. With this hyperspectral power, the researchers created a high-throughput phenotyping microscope – meaning it can process hundreds of leaf samples within a few hours for observable traits related to disease resistance. This capability lets it detect changes inside a leaf before they are visible on the surface. A trio of papers, recently published, outline the arc of the technology’s development.  

“It’s harnessing the power of hyperspectral and high-throughput analysis in a much more accessible way for the scientist,” said Katie Gold, assistant professor of grape pathology and Susan Eckert Lynch Faculty Fellow at Cornell AgriTech in the College of Agriculture and Life Sciences, one of three senior scientists on the papers and the HyperBird project. “We now have a physical system that allows us to translate complex questions and answer them with powerful tools in a way that’s so much more accessible than ever before.” 

“Hyperspectral information – beyond the visible range – can provide insights and deepen our understanding about many biological processes between leaf tissues and pathogens,” said Yu Jiang, assistant professor and systems engineer in the School of Integrative Plant Science (SIPS), Horticulture Section, at Cornell AgriTech in CALS, who is also a senior scientist on the papers and project.

The project also solved a major problem limiting hyperspectral imaging at microscopic scales, Jiang said. That issue concerns spatial resolution, which makes it possible to detect and distinguish a very tiny spot on a leaf as a symptom of disease in relation to the surrounding pixels that represent healthy cells.

“Imagine trying to find a Coke can in a photograph of a large room,” Jiang said. “If the image isn’t detailed enough to separate the can from the background, the light reflected by the can gets mixed with light from everything around it. That makes its spectral signature – the distinctive pattern of light it reflects – harder to recognize.”

HyperBird is a hyperspectral extension of Blackbird, a prototype technology developed by Jiang and Lance Cadle-Davidson, Ph.D. ‘03, a research plant pathologist at the USDA’s Agricultural Research Service (ARS), an adjunct associate professor in SIPS, and one of the senior project scientists and authors.

In 2021, the team announced the release of Blackbird, designed to speed up the assessment of thousands of grape leaf samples for evidence of infection, which had been a bottleneck in Cadle-Davidson’s lab’s research to develop powdery mildew resistant grape varieties. Blackbird, a robotic high-throughput phenotyping camera that records in red, blue and green wavelengths, was a first step towards HyperBird’s capabilities.

While Blackbird provided the spatial resolution of an optical microscope, HyperBird supplies roughly 200 times more spectral resolution per pixel. This level of hyperspectral detail allows HyperBird to collect data in three dimensions per pixel. Technicians still need to physically collect dime-sized leaf cutouts for HyperBird, the most time-consuming task in analyzing the samples and which the robotics students are trying to automate.

Credit: Ryan Young/Cornell University

A study published May 20 in the journal Plant Disease, led by Saeed Hosseinzadeh, Ph.D. ’23, a former postdoctoral researcher in Gold’s lab, took a first step to evaluate the use of hyperspectral imaging for automated high-throughput phenotyping for plant pathology and breeding programs. To that effect, the technology collected nine terabytes of high-quality hyperspectral images in a day to show proof-of-principle for detecting disease before it becomes visible. It also detected fungicide residues, to discern which treatment was used to fight mildew and its effectiveness. Finally, it revealed grapevine lineages being bred for disease resistance by identifying subtle optical signatures and relating those cues to a breeding line’s unique genetic makeup, which influences physical – and observable – traits such as the leaf’s pigments, wax structure, water content and internal anatomy. 

HyperBird was then formally developed by Jinhong Yu, a doctoral student in Jiang’s lab and first author of a paper describing the process, published Sept. 8 in the journal Plant Phenomics. It shows how HyperBird offers high-throughput hyperpectral imaging at the microscopic level. The robot is capable of identifying tiny diseased areas and comparing them with the average spectra across the rest of the grape leaf to determine the progression of downy and powder milder diseases. By drilling down to minute detail, HyperBird can detect disease much faster than previous methods.

“Now, we can get the equivalent accuracy at day three that we used to get at day six or day nine,” Gold said. “We’ve just increased the throughput by two to three times.”

And, another study, published June 29 in the journal Plant Disease, with first author Lorenzo Pippi, a former visiting scientist to Gold’s lab from the University of Pisa in Italy, tested HyperBird’s real-world ability to evaluate the performance of conventional fungicides and biofungicides to combat grapevine downy mildew, and compared them with an untreated control. By collecting leaf discs from grapevines in the field, the team found hyperspectral imaging could capture early optical changes associated with infection. HyperBird allowed the researchers to discriminate between the different sprays used, based on leaf residues, and then assess the success of each. Assessing a treatment’s effectiveness can be challenging because mildew infections vary across a vineyard. 

The team is working with Moblanc Robotics to manufacture HyperBird, which has potential applications for other plants and diseases, and researchers have begun discussions with a company in the cheese industry.

The HyperBird project is part of VitisGen3, a grape breeding project funded by a Specialty Crop Research Initiative Competitive Grant of the USDA National Institute of Food and Agriculture (NIFA).  

This research was funded by the USDA-NIFA Specialty Crop Research Initiative, Federal Capacity Funds awarded to Cornell University, the President’s Council of Cornell Women Research fund, the New York Wine and Grape Foundation, and USDA Grape Genetics Research Unit.