“Our approach enabled us to read a text that was hidden beneath corrosion, perhaps for hundreds of years,” said Alexandre Locquet, an adjunct professor in ECE and researcher at Georgia Tech-CNRS IRL 2958, a joint international research laboratory at the Georgia Tech-Lorraine campus in Metz, France. “Clearly, approaches that access such information without damaging the object are of great interest to archaeologists.”

The study was reported March 2 in the journal Scientific Reports.

The cross, cut from a sheet of lead, was found in a burial plot at an abbey in Remiremont, France — a couple hours drive from the Georgia Tech-Lorraine campus. Known as a croix d’absolution, it is a type of funerary cross that dates to the Middle Ages and has been found at sites in France, Germany, and England.

“This type of cross typically bears inscriptions of prayers or information about the deceased,” said Aurélien Vacheret, director of the Musée Charles-de-Bruyères in Remiremont and co-author on the study. “It is thought their purpose was to seek a person’s absolution from sin, facilitating their passage to heaven.”

The museum loaned the cross to Citrin’s lab in hopes that the team could use imaging techniques to make the invisible visible. Citrin and his group specialize in non-destructive evaluation and develop techniques that allow for detailed examination of an object’s hidden layers without changing or damaging its original form. Although their work often has industrial applications, such as detecting damage to airplane fuselages, the group embraced the opportunity to inspect the cross — a chance to further explore their technology’s applications for archaeological purposes.

Peeking Beneath the Veil of Corrosion

The team used a commercial terahertz scanner to examine the cross every 500 microns (about every half a millimeter) across the object. First, the scanner sent short pulses of terahertz electromagnetic radiation — a form of light that travels on tiny wavelengths — over each section of the cross. Some waves bounced back from the layer of corrosion, while others penetrated through the corrosion, reflecting from the actual surface of the lead cross. This produced two distinct echoes of the same original pulse.

Next, the team used an algorithm to process the time delay between the two echoes into a signal with two peaks. This data revealed how thick the corrosion was in each scanned point. The measurements of the light beams that reflected from the underlying metal were then collected to form images of the lead surface below the corrosion.

Interdisciplinary Insights

Although crucial data was gathered during the scanning process, the raw images were too noisy and jumbled and the inscription remained illegible at the time. But Junliang Dong, then a Ph.D. student in Citrin’s lab, had the insight to process the images in a special way to eliminate the noise. By subtracting and piecing together parts of the images acquired in different frequencies, Dong was able to restore and enhance the images. What was left was a surprisingly readable image containing the text.

Using the processed images, Vacheret was able to identify multiple Latin words and phrases. He determined they were all part of the Pater Noster, commonly known as the Our Father or the Lord’s Prayer.

The team also worked with a conservationist to chemically reverse the corrosion on the cross, confirming the Pater Noster inscription. Comparing their images to the clean cross, the team found their images had revealed parts of the inscription not observable on the original cross. By uncovering additional aspects of the inscriptions that were previously undocumented, their work was able to offer deeper understanding of the cross and further insight into 16^th-century Christianity in Lorraine, France.

“In this case, we were able to check our work afterwards, but not all lead objects can be treated this way,” Citrin said. “Some objects are large, some must remain in situ, and some are just too delicate. We hope our work opens up the study of other lead objects that might also yield secrets lying underneath corrosion.”

Citrin’s group has also used terahertz imaging to look beneath the surface of 17^th-century paintings, elucidating paint layer structure and providing insights into techniques of master painters. They are currently investigating surface coatings on ancient Roman ceramics.

The cross project illustrates that success requires more than just accurate measurement, but also careful data processing and collaboration between researchers from disparate fields. The team’s approach opens new perspectives for terahertz imaging analysis and could produce great boosts for the fields of digital acquisitions and documentation, as well as character recognition, extraction, and classification.

“Despite three decades of intense development, terahertz imaging is still a rapidly developing field,” said Locquet. “While others focus on developing the hardware, our efforts concentrate on making the most of the data that is measured.”

Reprogramming that robot involves hand-labeling thousands of images that show it how to grasp these new mugs, then training the system all over again.

But a new technique developed by MIT researchers would require only a handful of human demonstrations to reprogram the robot. This machine-learning method enables a robot to pick up and place never-before-seen objects that are in random poses it has never encountered. Within 10 to 15 minutes, the robot would be ready to perform a new pick-and-place task.

The technique uses a neural network specifically designed to reconstruct the shapes of 3D objects. With just a few demonstrations, the system uses what the neural network has learned about 3D geometry to grasp new objects that are similar to those in the demos.

In simulations and using a real robotic arm, the researchers show that their system can effectively manipulate never-before-seen mugs, bowls, and bottles, arranged in random poses, using only 10 demonstrations to teach the robot.

“Our major contribution is the general ability to much more efficiently provide new skills to robots that need to operate in more unstructured environments where there could be a lot of variability. The concept of generalization by construction is a fascinating capability because this problem is typically so much harder,” says Anthony Simeonov, a graduate student in electrical engineering and computer science (EECS) and co-lead author of the paper.

Simeonov wrote the paper with co-lead author Yilun Du, an EECS graduate student; Andrea Tagliasacchi, a staff research scientist at Google Brain; Joshua B. Tenenbaum, the Paul E. Newton Career Development Professor of Cognitive Science and Computation in the Department of Brain and Cognitive Sciences and a member of the Computer Science and Artificial Intelligence Laboratory (CSAIL); Alberto Rodriguez, the Class of 1957 Associate Professor in the Department of Mechanical Engineering; and senior authors Pulkit Agrawal, a professor in CSAIL, and Vincent Sitzmann, an incoming assistant professor in EECS. The research will be presented at the International Conference on Robotics and Automation.

Grasping geometry

A robot may be trained to pick up a specific item, but if that object is lying on its side (perhaps it fell over), the robot sees this as a completely new scenario. This is one reason it is so hard for machine-learning systems to generalize to new object orientations.

To overcome this challenge, the researchers created a new type of neural network model, a Neural Descriptor Field (NDF), that learns the 3D geometry of a class of items. The model computes the geometric representation for a specific item using a 3D point cloud, which is a set of data points or coordinates in three dimensions. The data points can be obtained from a depth camera that provides information on the distance between the object and a viewpoint. While the network was trained in simulation on a large dataset of synthetic 3D shapes, it can be directly applied to objects in the real world.

The team designed the NDF with a property known as equivariance. With this property, if the model is shown an image of an upright mug, and then shown an image of the same mug on its side, it understands that the second mug is the same object, just rotated.

“This equivariance is what allows us to much more effectively handle cases where the object you observe is in some arbitrary orientation,” Simeonov says.

As the NDF learns to reconstruct shapes of similar objects, it also learns to associate related parts of those objects. For instance, it learns that the handles of mugs are similar, even if some mugs are taller or wider than others, or have smaller or longer handles.

“If you wanted to do this with another approach, you’d have to hand-label all the parts. Instead, our approach automatically discovers these parts from the shape reconstruction,” Du says.

The researchers use this trained NDF model to teach a robot a new skill with only a few physical examples. They move the hand of the robot onto the part of an object they want it to grip, like the rim of a bowl or the handle of a mug, and record the locations of the fingertips.

Because the NDF has learned so much about 3D geometry and how to reconstruct shapes, it can infer the structure of a new shape, which enables the system to transfer the demonstrations to new objects in arbitrary poses, Du explains.

Picking a winner

They tested their model in simulations and on a real robotic arm using mugs, bowls, and bottles as objects. Their method had a success rate of 85 percent on pick-and-place tasks with new objects in new orientations, while the best baseline was only able to achieve a success rate of 45 percent. Success means grasping a new object and placing it on a target location, like hanging mugs on a rack.

Many baselines use 2D image information rather than 3D geometry, which makes it more difficult for these methods to integrate equivariance. This is one reason the NDF technique performed so much better.

While the researchers were happy with its performance, their method only works for the particular object category on which it is trained. A robot taught to pick up mugs won’t be able to pick up boxes or headphones, since these objects have geometric features that are too different than what the network was trained on.

“In the future, scaling it up to many categories or completely letting go of the notion of category altogether would be ideal,” Simeonov says.

They also plan to adapt the system for nonrigid objects and, in the longer term, enable the system to perform pick-and-place tasks when the target area changes.

This work is supported, in part, by the Defense Advanced Research Projects Agency, the Singapore Defense Science and Technology Agency, and the National Science Foundation.

Now, an interdisciplinary team of engineers, biologists, and geneticists has developed a new way of studying the heart: they’ve built a miniature replica of a heart chamber from a combination of nanoengineered parts and human heart tissue. There are no springs or external power sources — like the real thing, it just beats by itself, driven by the live heart tissue grown from stem cells. The device could give researchers a more accurate view of how the organ works, allowing them to track how the heart grows in the embryo, study the impact of disease, and test the potential effectiveness and side effects of new treatments — all at zero risk to patients and without leaving a lab.

The Boston University-led team behind the gadget — nicknamed miniPUMP, and officially known as the cardiac miniaturized Precision-enabled Unidirectional Microfluidic Pump — says the technology could also pave the way for building lab-based versions of other organs, from lungs to kidneys. Their findings have been published in Science Advances.

“We can study disease progression in a way that hasn’t been possible before,” says Alice White, a BU College of Engineering professor and chair of mechanical engineering. “We chose to work on heart tissue because of its particularly complicated mechanics, but we showed that, when you take nanotechnology and marry it with tissue engineering, there’s potential for replicating this for multiple organs.”

According to the researchers, the device could eventually speed up the drug development process, making it faster and cheaper. Instead of spending millions — and possibly decades — moving a medicinal drug through the development pipeline only to see it fall at the final hurdle when tested in people, researchers could use the miniPUMP at the outset to better predict success or failure.

The project is part of CELL-MET, a multi-institutional National Science Foundation Engineering Research Center in Cellular Metamaterials that’s led by BU. The center’s goal is to regenerate diseased human heart tissue, building a community of scientists and industry experts to test new drugs and create artificial implantable patches for hearts damaged by heart attacks or disease.

“Heart disease is the number one cause of death in the United States, touching all of us,” says White, who was chief scientist at Alcatel-Lucent Bell Labs before joining BU in 2013. “Today, there is no cure for a heart attack. The vision of CELL-MET is to change this.”

Personalized Medicine

There’s a lot that can go wrong with your heart. When it’s firing properly on all four cylinders, the heart’s two top and two bottom chambers keep your blood flowing so that oxygen-rich blood circulates and feeds your body. But when disease strikes, the arteries that carry blood away from your heart can narrow or become blocked, valves can leak or malfunction, the heart muscle can thin or thicken, or electrical signals can short, causing too many — or too few — beats. Unchecked, heart disease can lead to discomfort — like breathlessness, fatigue, swelling, and chest pain — and, for many, death.

“The heart experiences complex forces as it pumps blood through our bodies,” says Christopher Chen, BU’s William F. Warren Distinguished Professor of Biomedical Engineering. “And while we know that heart muscle changes for the worse in response to abnormal forces — for example, due to high blood pressure or valve disease — it has been difficult to mimic and study these disease processes. This is why we wanted to build a miniaturized heart chamber.”

At just 3 square centimeters, the miniPUMP isn’t much bigger than a postage stamp. Built to act like a human heart ventricle — or muscular lower chamber — its custom-made components are fitted onto a thin piece of 3D-printed plastic. There are miniature acrylic valves, opening and closing to control the flow of liquid — water, in this case, rather than blood — and small tubes, funneling that fluid just like arteries and veins. And beating away in one corner, the muscle cells that make heart tissue contract, cardiomyocytes, made using stem cell technology.

“They’re generated using induced pluripotent stem cells,” says Christos Michas (ENG’21), a postdoctoral researcher who designed and led the development of the miniPUMP as part of his PhD thesis.

To make the cardiomyocyte, researchers take a cell from an adult — it could be a skin cell, blood cell, or just about any other cell — reprogram it into an embryonic-like stem cell, then transform that into the heart cell. In addition to giving the device literal heart, Michas says the cardiomyocytes also give the system enormous potential in helping pioneer personalized medicines. Researchers could place a diseased tissue in the device, for instance, then test a drug on that tissue and watch to see how its pumping ability is impacted.

“With this system, if I take cells from you, I can see how the drug would react in you, because these are your cells,” says Michas. “This system replicates better some of the function of the heart, but at the same time, gives us the flexibility of having different humans that it replicates. It’s a more predictive model to see what would happen in humans — without actually getting into humans.”

According to Michas, that could allow scientists to assess a new heart disease drug’s chances of success long before heading into clinical trials. Many drug candidates fail because of their adverse side effects.

“At the very beginning, when we’re still playing with cells, we can introduce these devices and have more accurate predictions of what will happen in clinical trials,” says Michas. “It will also mean that the drugs might have fewer side effects.”

Thinner than a Human Hair

One of the key parts of the miniPUMP is an acrylic scaffold that supports, and moves with, the heart tissue as it contracts. A series of superfine concentric spirals — thinner than a human hair — connected by horizontal rings, the scaffold looks like an artsy piston. It’s an essential piece of the puzzle, giving structure to the heart cells — which would just be a formless blob without it — but not exerting any active force on them.

“We don’t think previous methods of studying heart tissue capture the way the muscle would respond in your body,” says Chen, who’s also director of BU’s Biological Design Center and an associate faculty member at Harvard University’s Wyss Institute for Biologically Inspired Engineering. “This gives us the first opportunity to build something that mechanically is more similar to what we think the heart is actually experiencing — it’s a big step forward.”

To print each of the tiny components, the team used a process called two-photon direct laser writing — a more precise version of 3D printing. When light is beamed into a liquid resin, the areas it touches turn solid; because the light can be aimed with such accuracy — focused to a tiny spot — many of the components in the miniPUMP are measured in microns, smaller than a dust particle.

The decision to make the pump so small, rather than life-size or larger, was deliberate and is crucial to its functioning.

“The structural elements are so fine that things that would ordinarily be stiff are flexible,” says White. “By analogy, think about optical fiber: a glass window is very stiff, but you can wrap a glass optical fiber around your finger. Acrylic can be very stiff, but at the scale involved in the miniPUMP, the acrylic scaffold is able to be compressed by the beating cardiomyocytes.”

Chen says that the pump’s scale shows “that with finer printing architectures, you might be able to create more complex organizations of cells than we thought was possible before.” At the moment, when researchers try to create cells, he says, whether heart cells or liver cells, they’re all disorganized — “to get structure, you have to cross your fingers and hope the cells create something.” That means the tissue scaffolding pioneered in the miniPUMP has big potential implications beyond the heart, laying the foundation for other organs-on-a-chip, from kidneys to lungs.

Refining the Technology

According to White, the breakthrough is possible because of the range of experts on CELL-MET’s research team, which included not just mechanical, biomedical, and materials engineers like her, Chen, and Arvind Agarwal of Florida International University, but also geneticist Jonathan G. Seidman of Harvard Medical School and cardiovascular medicine specialist Christine E. Seidman of Harvard Medical School and Brigham and Women’s Hospital. It’s a breadth of experience that’s benefited not just the project, but Michas. An electrical and computer engineering student as an undergraduate, he says he’d “never seen cells in my life before starting this project.” Now, he’s preparing to start a new position with Seattle-based biotech Curi Bio, a company that combines stem cell technology, tissue biosystems, and artificial intelligence to power the development of drugs and therapeutics.

“Christos is someone who understands the biology,” says White, “can do the cell differentiation and tissue manipulation, but also understands nanotechnology and what’s required, in an engineering way, to fabricate the structure.”

The next immediate goal for the miniPUMP team? To refine the technology. They also plan to test ways to manufacture the device without compromising its reliability.

“There are so many research applications,” says Chen. “In addition to giving us access to human heart muscle for studying disease and pathology, this work paves the way to making heart patches that could ultimately be for someone who had a defect in their current heart.”

Seismic wave data from the events could help researchers learn more about the interior layers of Mars, particularly its core-mantle boundary, researchers from InSight’s Marsquake Service (MQS) report in The Seismic Record.

Anna Horleston of the University of Bristol and colleagues were able to identify reflected PP and SS waves from the magnitude 4.2 event, called S0976a, and locate its origin in the Valles Marineris, a massive canyon network that is one of Mars’ most distinguishing geological features and one of the largest graben systems in the Solar System. Earlier orbital images of cross-cutting faults and landslides suggested the area would be seismically active, but the new event is the first confirmed seismic activity there.

S1000a, the magnitude 4.1 event recorded 24 days later, was characterized by reflected PP and SS waves as well as Pdiff waves, small amplitude waves that have traversed the core-mantle boundary. This is the first time Pdiff waves have been spotted by the InSight mission. The researchers could not definitively pinpoint S1000a’s location, but like S0976a it originated on Mars’ far side. The seismic energy from S1000a also holds the distinction of being the longest recorded on Mars, lasting 94 minutes.

Both marsquakes occurred in the core shadow zone, a region where P and S waves can’t travel directly to InSight’s seismometer because they are stopped or bent by the core. PP and SS waves don’t follow a direct path, but rather are reflected at least once at the surface before traveling to the seismometer.

“Recording events within the core shadow zone is a real steppingstone for our understanding of Mars. Prior to these two events the majority of the seismicity was within about 40 degrees distance of InSight,” said Savas Ceylan, a co-author from ETH Zürich. “Being within the core shadow, the energy traverses parts of Mars we have never been able to seismologically sample before.”

The two marsquakes differ in some important ways. S0976a is characterized by only low frequency energy, like many of the quakes identified so far on the planet, while S1000a has a very broad frequency spectrum. “[S1000a] is a clear outlier in our catalog and will be key to our further understanding of Martian seismology,” Horleston said.

S0976a is likely to have a much deeper origin than S1000a, she noted. “The latter event has a frequency spectrum much more like a family of events that we observe that have been modeled as shallow, crustal quakes, so this event may have occurred near the surface. S0976a looks like many of the events we have located to Cerberus Fossae — an area of extensive faulting — that have depths modeled to be around 50 kilometers or more and it is likely that this event has a similar, deep, source mechanism.”

Compared to the rest of the seismic activity detected by InSight, the two new far-side quakes are true outliers, the researchers said.

“Not only are they the largest and most distant events by a considerable margin, S1000a has a spectrum and duration unlike any other event previously observed. They truly are remarkable events in the Martian seismic catalog,” Horleston said.

Story Source:

Materials provided by Seismological Society of America. Note: Content may be edited for style and length.

Researchers used a unique dataset to get a never-before-seen view of how science funding is spent and the results it produces.

The findings, published today (April 22, 2022) in the journal Science Advances, showed that funding of research by the National Institutes of Health mostly goes to hiring people who work on the projects funded by grants.

But these people — which includes staff and trainees like graduate students — go on to do more than just work on the grants for which they were hired, said Enrico Berkes, co-author of the study and postdoctoral researcher in economics at The Ohio State University.

“We see a great increase in productivity in publications directly linked to a grant but also in new studies that go beyond it,” Berkes said. “There is this ripple effect where people supported by the grant also produce other quality work.”

And the biomedical researchers who are the focus of this study produced more clinical research as the result of receiving more funding — studies directly related to patient care and health, said study co-author Bruce Weinberg, professor of economics at Ohio State.

“Funding is actually producing the kind of research that would lead to improvements in clinical outcomes for patients,” Weinberg said.

The key to this study is the UMETRICS dataset available through the Institute for Research on Innovation and Science. It provides detailed information concerning payments on sponsored research projects at 72 universities.

This allowed the researchers to use payments to identify all people working on research projects funded by NIH — from faculty members to trainees to staff.

Berkes, Weinberg and colleagues then used the PubMed database to find all research publications produced by the scientists in the UMETRICS database. They examined NIH grants between 1985 and 2020.

“We were able to link people to research projects by following the money,” Weinberg said. “This allowed us to answer a question that wasn’t possible before — how money spent on research impacts people.”

Results showed that 68% of grant funding went to spending on employees, which included faculty members, postdoctoral researchers, graduate and undergraduate students, research staff and other staff.

Increases in funding led labs to become more professionalized by hiring more staff and career researchers, findings showed. In fact, research and other staff showed a higher percentage of growth in employment than faculty members when funding increased.

As funding increased and research teams got larger, they produced more scientific papers — and the quality of research did not decline.

“One hypothesis would be that as teams grow larger, they would become more bureaucratic, and it would become more difficult to produce quality science. But we found that labs kept productivity up, likely because they become more professionalized,” Berkes said.

Findings showed that the largest increase in papers comes from studies not directly related to the grant. These are often papers that are not co-authored by the principal investigator, the researcher who is responsible for getting and managing the funding.

One way to understand the ripple effects of funding for medical research is to see how many research papers the people involved with the grants produce.

Unsurprisingly, faculty members have the largest increase in new scientific papers as the result of additional funding. But in relative terms, the trainees — including graduate students and postdoctoral researchers — have a larger-percentage increase in papers they produce.

“We can see how research funding is jump-starting the careers of trainees who take what they learn while working on these funded projects, and the collaborators they met on the grant, and start investigating other important issues,” Weinberg said.

Funding for biomedical research tends to increase both the number of clinical papers directly tied to patient care, and other types of medical research, results showed.

These findings may provide the best view yet of how federal grants awarded for research are spent, Berkes said.

“Funders tend to focus, understandably, on the impact of their money on the specific issue they funded,” Berkes said.

“But they should be aware of how their funding moves through a wide range of people and produces benefits they may have not expected.”

Funding for the study came from the National Institute on Aging, the Office of Behavioral and Social Science Research, the National Institute of General Medical Sciences, the National Center for Advancing Translational Sciences, the National Science Foundation, and the Ewing Marion Kauffman and Alfred P. Sloan foundations.

Other authors on the study were Reza Sattari, a former postdoctoral researcher, and Jung Bae, a PhD graduate, both from Ohio State.