MIT student team CERBERUZ wins NASA LunaRecycle Centennial Challenge, turning lunar trash into tools

Establishing a human presence on the moon will require state-of-the-art spacecraft, spacesuits, life-support systems, and some of the most sophisticated technology humans have ever built. It will also require figuring out where to put the garbage.

A future lunar outpost will receive a steady stream of supplies from Earth, bringing packaging and other materials that could quickly become a mountain of waste. Meanwhile, astronauts will need tools, replacement parts, and materials to maintain the outpost, and shipping supplies from Earth is astronomically expensive. An MIT student team has won first prize in a NASA competition for developing a way to turn that trash into useful materials for lunar living.

The Composites for Extraterrestrial Recycling By Engineering the Reuse and Upcycling of Zotek (CERBERUZ) team was named first-prize winner in both the Prototype and Digital Twin tracks of NASA’s LunaRecycle Challenge. The challenge — part of NASA’s Centennial Challenges series — asked participants to design and develop solutions that can reuse, recycle, or upcycle space waste and improve the sustainability of longer-duration lunar missions. Teams were encouraged to envision solutions that not only addressed recycling in deep space, but could also have applications on Earth.

The CERBERUZ team’s system turns otherwise unusable waste into new composite materials, grinding it down, melting it, and molding it into useful parts. The approach reduces the need for costly shipments from Earth and keeps bulky packaging from piling up in lunar landfills.

“Waste packaging materials will inevitably accumulate when a lunar outpost is built, and at the same time, outpost construction and maintenance will require a steady supply of new materials, including replacement parts,” says George Lordos, a research scientist in the MIT Department of Aeronautics and Astronautics and System Design and Management, and the lead advisor for the CERBERUZ team. “Recycling waste materials into such parts can help reduce materials needed to be launched from Earth.”

Waste not, want not

CERBERUZ separates non-biological waste into two streams: thermoplastics and everything else. All materials are then ground into a fine powder and mixed together. The powder mixture is loaded into an injection molding machine, where the thermoplastics melt with the remaining materials suspended in the molten plastic. The mixture is finally injected into a mold, producing high-quality composite parts that can be quickly and reliably replicated.
 

The team’s approach is especially notable for its treatment of Zotek F30, a closed-cell thermoset foam commonly used in aerospace packaging. Mixed thermoplastics and thermoset foams like Zotek F30 have no established recycling solutions, even on Earth. CERBERUZ takes a unique approach to this challenge by incorporating the Zotek as a reinforcing filler for the molten feedstock.

Zotek presents another peculiar engineering challenge that makes the need to break it down even more pressing: It is so lightweight that enormous quantities are needed to produce a useful mass of material. “We had big shipments of Zotek coming in, and one person can carry them on their head like it’s nothing,” says José Soto, a graduate student in AeroAstro and the team's pre-processing lead. “The box is the size of a person.”

The team designed the process around the constraints of operating far from Earth. Because CERBERUZ uses mechanical methods powered by electricity, it doesn’t rely on water-intensive recycling processes or chemical depolymerization, and produces no secondary chemical waste. This could allow packaging and other materials delivered to a lunar outpost to become a local source of manufacturing feedstock, giving astronauts the ability to produce needed items from material they already have on hand. 

“We set out to build something that would actually make sense on the moon, not just in a lab,” says Lanie McKinney, a graduate student in AeroAstro and team lead for CERBERUZ. “That means no water and no chemicals you can’t replace. Every design choice we made was driven by those constraints. And we don’t want to recycle just for recycling’s sake; we want to make useful things. Part of how we framed the problem was to create a system that can process nearly any feedstock and turn it into a wide array of components, giving more utility than producing a single item.”

Alongside the physical recycling system, the team developed DEIMOS, or Digital Environment for Intelligent Modeling of Operational Systems, a digital twin of its injection molding machine. Using data collected during actual production runs, DEIMOS models how different combinations of recycled materials and part geometries will behave during manufacturing, predicting factors including mold fill, warpage, and shrinkage risk. Ultimately, the digital twin could help astronauts determine how recycled materials will perform before committing limited resources to manufacturing a part on the moon.

Where ideas become hardware

The CERBERUZ team grew out of the MIT Space Resources Workshop, a co-curricular initiative of students passionate about designing, building, and testing systems to support future moon and Mars exploration. Founded in 2017, the workshop is well-known for building teams that take on, and conquer, grand challenges in NASA-sponsored competitions.

The team brings together undergraduate and graduate students from four MIT departments: Aeronautics and Astronautics, Mechanical Engineering, Electrical Engineering and Computer Science, and Materials Science and Engineering. Professor Jeffrey Hoffman, a former NASA astronaut, and Lordos served as advisors for the team. 

“What impresses me most about this team is the quality of leadership,” says Lordos. “Lanie McKinney has been an outstanding team lead, and the co-leads — Palak Patel on manufacturing and materials characterization, Lilly Etzenbach on the digital twin, and José Soto on pre-processing — have each built and managed subteams doing cutting-edge research in lunar sustainability.”

The win caps nearly two years of work on the challenge. CERBERUZ was selected as a Phase 1 winner in 2025 before advancing through Phase 2 and earning milestone prizes in both tracks on its way to the final demonstration in summer 2026. The team earned $925,000 in prize money over the course of the challenge, and hopes to reinvest their winnings in the Space Resources Workshop, funding equipment and facilities that will enable future teams to rise to ever-more-complex challenges. Those advances will help turn a return to the moon into something more permanent: a human presence, built to last.

“If we’re able to reduce the amount of waste that’s inevitably going to be left on the moon, rather than letting it pile up like a landfill,” Soto says, “we can preserve both the beauty of the moon and make exploration more efficient and sustainable. And perhaps we can inspire people back on Earth to take another look at our trash and find more ways to recycle or reuse it.”

CERBERUZ team leads are Lanie McKinney, Palak B. Patel, Lilly Etzenbach, and José Soto. Team members include Sophia Augier, Sophie Berkvist, Ariella Blackman, Keira Boone, Constantine Bulavenko, Seth Burrowes, Andrew Chen, Alex Chen, Chris Choe, Johnson Cristales, Brady Cruse, Hillel Dei, Natalie Hatzigeorgiu, Intia Ibnah, Avishai Jeselsohn, Mollie Johnson, Uma Kirk, Christopher Kwon, Jessica Meza, Juan Orrego, Giuliana Pertuz, Jordan Prawira, Brandy Romine, Dustin Wang, Madison Wood, and Adrian Yang. 

de MIT News https://ift.tt/rp0Mjs6

Beaver Works Summer Institute: A leader in transformational STEM education

Eleven years ago, MIT Lincoln Laboratory staff member Robert Shin was challenged by a colleague to find a way to extend the resources of Beaver Works — a joint education and research center created by the MIT School of Engineering and the laboratory — to some of the best and brightest engineering students around the country. A year later, in summer 2016, the MIT Beaver Works Summer Institute (BWSI) was born. 

BWSI, which celebrated its 10th anniversary this summer, is a four-week program designed to offer high school students a transformational hands-on experience while teaching them skills to fuel their future studies and careers in science, technology, engineering, and mathematics (STEM).

"As I look back at the last 10 years of BWSI, I am humbled and grateful to be a part of this amazing journey," says Shin, who has served as the founding director and then the STEM program director for BWSI since 2016. "What started as an idea has evolved into something much bigger and more impactful than we ever imagined. Each year, our students have exceeded every expectation through their curiosity, hard work, willingness to accept challenges, and perseverance." 

Originally launched with a class of 46 students from 24 high schools across the country, the BWSI program has seen steady growth each year, both in the number of students enrolled and the total number of courses offered. In the first year of the program, BWSI offered just a single course — the Rapid Autonomous Complex Environment Competing Ackermann-Steering Robots (RACECAR) course — that challenged students to program vehicles to navigate a racetrack full of obstacles. A decade later, BWSI now offers 14 courses spanning fields such as artificial intelligence, data science, cybersecurity, quantum computing, and microelectronics to more than 440 students. 

"The nation needs innovators, problem-solvers, and leaders to tackle the national security challenges of tomorrow in a rapidly evolving world," says Melissa Choi, director of Lincoln Laboratory. "By bridging the gap between classroom theory and practical application, BWSI inspires the next generation of scientists and engineers and exemplifies the laboratory’s commitment as a federally funded research and development center to help develop the future STEM workforce." 

Beaver Works Operational Manager Lisa Kelley has been with the program from the start, for both the founding of Beaver Works in 2010 and the launch of BWSI in 2016. She recalls welcoming the first class of 46 students, adding that she never could have imagined the impact that BWSI would have a decade later. 

"Today, I look at the thousands of students who have come through our programs with immense pride and gratitude," says Kelley. "What makes me most proud is that we built something both exceptional and accessible, a program that challenges students at the highest level while expanding opportunity." 

Beyond the summer curriculum, BWSI has expanded throughout the years by offering eight-week outreach programs in the spring and fall that build technical foundations for high school students in grades 9–11 prior to entering the competitive summer program. One of these fall programs is "Girls Who Can," which focuses on engaging young girls with STEM. Additionally, in 2022 and 2023, BWSI launched national challenges that adapted the content of two of their courses (Build a CubeSat and Designing for Assistive Technology) into an eight-month team-based format with mentors. 

The BWSI curriculum has inspired collaborators to create similar programs in countries such as Greece, Japan, Canada, Mexico, Italy, and Spain, as well as parallel programs in schools and universities across the country. 

Beaver Works Manager Joel Grimm says that the growth of the program has continued to challenge and inspire him from year to year. "To me, Beaver Works and BWSI mean embracing change — saying 'yes' to something new and innovative," Grimm says. "I encourage students to step out of their comfort zone and see what’s possible, and I hope that more students understand that BWSI is a unique opportunity to take a risk and stretch beyond their expectations." 

After participating in BWSI, many students continue to engage with the program. One example is Diptanshu Sikdar, who participated in BWSI for two summers as a high school student before returning as a teaching assistant for the Quantum Software and Software Development courses. 

"BWSI transformed my curiosity about quantum computing into a strong foundation and introduced me to the beauty of algorithm design and computer science," says Sikdar, who went on to pursue a PhD in quantum science and credits his experience with BWSI for building connections that led to his current job at MITRE. "It was much more than a summer program for me; it was the point of origin for my quantum journey and the foundation for my growth as a student, researcher, teacher, and leader." 

It’s this type of transformational experience that Shin hopes BWSI students will continue to have well into the future. 

"I believe the last 10 years of BWSI were just the beginning," Shin says. "The BWSI program at MIT will continue to be impactful, and we see a future where all students interested in programs like BWSI will have an opportunity to participate and learn. Wouldn’t that be wonderful? It is not only possible, but we owe that to our next generation." 

To learn more about BWSI and how to support its mission, visit the BWSI website or email the administrator.

de MIT News https://ift.tt/majAOvt

MIT spinout makes surgical stitching more like applying a sticky note

Dozens of surgical procedures require doctors to connect two pieces of tissue together in a delicate stitching process. It happens in triple heart bypass, C-sections, and surgeries to repair hernias and tendons. Today, those procedures are mostly done by surgeon-controlled robots or by hand, and the margin for error is small. Even small errors in suturing or stapling can lead to leaks with devastating consequences to patients.

SanaHeal wants to make connecting tissues more like using tape. The company has developed a bioadhesive technology that can rapidly stick to wet and uneven surfaces and form air- and fluid-tight seals to simplify surgical procedures. The company’s product, first developed at MIT, was inspired by sticky materials in nature like barnacles and spider webs.

SanaHeal’s technology has already gained regulatory approval to stop trauma bleeding, and next year it will enter human trials to help control bleeding during surgeries and for repair of various injured tissues.

The startup, which collaborates with larger companies for clinical trials and commercialization, believes those applications only scratch the surface of the technology’s potential.

“If someone is struggling to put something on biologic tissues, we can help,” SanaHeal founder and chief technology officer Hyunwoo Yuk SM ’16, PhD ’21 says. “The current way of doing this is by suturing through tissues, which is painful and difficult to do perfectly during surgery. Our solution is more like a Post-it note. It’s much less invasive and easier to use with surgical robots. Also, faster and more efficient surgical procedures have potential to help lower ever-increasing costs of surgeries that are significantly straining the health care system.”

A sticking problem

Yuk conducted his master’s and PhD in the lab of Xuanhe Zhao, MIT’s Uncas (1923) and Helen Whitaker Professor of Mechanical Engineering. Initially, his research focused on the mechanical properties of soft materials called hydrogels, but about halfway through the program, he pivoted to study their biomedical applications, in part motivated by a serious accident his family member was involved in.

“We were looking at how to make these hydrogels better adhesives, and because we are in Boston, we started working with doctors and medical schools in the surrounding area,” Yuk recalls.

Those conversations made him realize how often surgeons have to seal or attach tissues. That process usually involves staples or sutures, which can cause more damage to tissues, trigger immune responses, and lead to scar tissue or infections.

“These are often tiny tissues that sometimes require needles thinner than a human hair to be used during microsurgery,” Yuk explains. “They are very difficult surgeries, and they can lead to all kinds of complications.”

Yuk realized his hydrogels offered a better solution. The materials, which are made from polymers and water, are biocompatible, meaning they minimize immune responses and scar tissue buildup. Early versions of the technology acted like double-sided tape, absorbing water from wet tissues using polyacrylic acid — an absorbent material used in diapers — to adhere to them in a matter of seconds.

Yuk had planned to stay in academia after his PhD, but as he learned more about what it would take for the technology to reach patients, he decided to focus on leading that work.

“It’s very difficult for innovations to make it into the clinic and help patients if you’re in an academic research lab,” Yuk says. “We didn’t know anything about startups, so we used resources like MIT’s Deshpande Center, the Venture Mentoring Service, and MIT’s Technology Licensing Office in hope that there might be external interest in our technology. We quickly learned the regulatory hurdles presented too much risk at the early stages for potential partners, so we had to commercialize it ourselves.”

The founders received funding from the Deshpande Center to start developing the company. SanaHeal was officially formed in 2021.

Yuk and his collaborators initially focused on applying their technology to wounds to stop bleeding. But they soon shifted to focus on surgical applications, which are more numerous.

“It’s a major unmet need,” Yuk says. “We are giving surgeons more options for a host of complex procedures. They are really excited about this.”

Partnering for impact

SanaHeal provides its technology to larger medical device company partners, who are developing products for a wide range of applications including in open, laparoscopic, and endoscopic surgeries as well as in combination with existing devices for sutureless implantation in applications like hernia and nerve repairs.

“We have envisioned ourselves as a platform company from the beginning,” Yuk says. “We don’t have the resources to develop 10 products at once, so we partner with bigger companies to commercialize solutions.”

Yuk, who has been working on the technology for nearly a decade, admits the commercialization process has taken longer than he expected. But he believes the payoff — seeing a technology he first developed in a lab have an impact on patients — will make the journey worth it.

“My life goal as a researcher is to see the full lifecycle of a new technology, from the scientific discovery, to research, to understanding how it works, to engineering it, to working with subject matter experts, and finally developing a product that gets approved to help humans,” Yuk explains. “It’s rare, but it would be incredibly rewarding and would be a great story to tell to inspire my research colleagues.”

de MIT News https://ift.tt/WbVmC6t

Symposium highlights MIT’s wide range of innovations for aging brain health

From its beginning in 2015, the premise of MIT’s Aging Brain Initiative (ABI) has been that neurodegenerative diseases such as Alzheimer’s are complex, “system-wide” breakdowns that will require advances of many kinds from many sources to address. The six talks at the ABI’s symposium “Innovations for the Aging Brain” on Sept. 25 illustrated what that’s come to mean at MIT, as scientists and engineers described their latest work to improve health for the aging brain.

“The ABI is unique because it brings together researchers from a diverse array of fields,” said ABI Director Li-Huei Tsai, Picower Professor in the the Picower Institute for Learning and Memory and Department of Brain and Cognitive Sciences (BCS). “Today’s symposium showcases MIT’s collective brilliance across neuroscience, engineering, chemistry, and beyond.”

One of the talks, for instance, featured a team effort including Giovanni Traverso, the Karl Van Tassel (1925) Professor of Mechanical Engineering and a physician at Brigham and Women’s Hospital, who has joined forces with Tsai and other collaborators to develop an artificial intelligence system that could improve personalized Alzheimer’s care by matching patients to the best prevention interventions for them. The system integrates data from clinical trials, patient lifestyle histories, and multi-omics data (e.g., genomics, proteomics) to predict an individual’s Alzheimer’s progression and likely response to interventions. 

Speaking alongside project leads Arvid Gollwitzer, an engineer-scientist in the Broad Institute of MIT and Harvard, and MIT-Novo Nordisk AI Fellow Adrián Noriega de la Colina, Traverso described how the team engineered an AI model, called FINGERS-7B, to integrate these multiple highly diverse data sources into its training, thereby amassing 8 trillion “tokens” to inform its predictions.

“Every individual is very unique in their biology, and that links to how we respond to those interventions,” said Noriega. Not only could AI predictions help tailor individual prevention measures, he added, but also it could inform clinical trial designs by better pinpointing the most appropriate populations to target with a prospective drug or other treatment.

Picower Professor Earl K. Miller focused on prevention of a different problem for the aging brain: cognitive side effects from exposure to too much general anesthesia. Noting high rates of postoperative delirium in older adults after major surgery, Miller described his research with ABI founding member and Institute Professor Emery N. Brown, a Picower Institute and BCS colleague, to develop a closed-loop system that uses brain wave-based measurements of unconsciousness to optimize anesthesia dosing. Miller and Brown’s labs have shown that regardless of their molecular means of action, multiple anesthesia drugs induce unconsciousness in the same way: by shifting brainwaves out of phase with each other, disrupting the coordination of communication across the brain’s cortex that sustains cognition and consciousness. Informed by such brain wave “signatures” of unconsciousness, the system could help anesthesiologists to safely tailor dosing during surgery.

In her talk, associate professor of mechanical engineering Ritu Raman, the Eugene Bell Career Development Chair of Tissue Engineering, described how she’s approaching the study of neuromuscular diseases, including amyotrophic lateral sclerosis (ALS) and muscular dytrophies, by developing advanced lab models of muscle tissue that also integrate nerves and vasculature. For instance, for ALS, her lab has been able to make the cultures from induced pluripotent stem cells derived from patients. The lab models give her team an opportunity to examine neuromuscular junctions in detail and to model how they change in response to potentially therapeutic drugs.

In many neurodegenerative diseases associated with aging, such as Alzheimer’s and Parkinson’s, a major pathology is misfolding and aggregation of intrinsically disordered proteins, such as amyloid-beta or alpha-synuclein. In her talk, William R (1964) and Daniel L. (1995) Young Career Development Assistant Professor of Chemistry Oleta Johnson described her lab’s work to prevent aggregation before it starts. In pursuit of that strategy, she said her lab’s research is focused on developing the chemical knowledge and means to identify proteins at issue, detect that they are moving into a problematic state, and then control their conformation.

Eugene McDermott Professor in the Brain Sciences and Human Behavior Alan Jasanoff, a member of the BCS and Biological Engineering departments as well as the McGovern Institute for Brain Research, described a very different application of chemistry to aging brain research. He described how his lab develops novel chemical probes to make molecules in the brain visible in magnetic resonance imaging. These include the neuromodulatory enzyme cholinesterase, which is the target of many of the few drugs approved to treat Alzheimer’s, and also lipids, which several brain cell types mishandle in Alzheimer’s, Tsai and others have shown. His lab has also developed technology to genetically encode a novel probe in brain cells to improve MRI visualization of brain activity.

And while many researchers focus on the prominent problems that proteins or lipids present in neurodegenerative disease, Novartis Professor of Chemistry Laura Kiessling discussed a less-appreciated but still important component of cellular function: sugars called glycans that line the surface of cells. In particular, researchers have linked a glycan called heparan sulfate to Alzheimer’s disease. Kiessling’s lab, working with ABI founding member Ed Boyden, Y. Eva Tan Professor of Neurotechnology, is working to characterize and understand glycans. Her lab, for instance, has shown that glycans of different sequences organize on cell membranes so that they can bind and recruit specific proteins to distinct regions on cells.

“We are really excited about the fact that we’ve observed this now for the first time and have a toolkit to study other diseases,” she said. Understanding glycan binding could even give scientists a means to target different cells and cellular processes.

Following the six talks, more than a dozen trainees at a poster session presented even more aging brain-related projects from around the Institute, further highlighting MIT’s diversity of research in the area, something the ABI exists to encourage.

From massive AI datasets to the sugars on individual cells, the symposium highlighted how MIT scientists are pushing the frontiers of science and engineering to improve understanding of, and interventions for, the health of the aging brain.

de MIT News https://ift.tt/zDl0vOY

Why teaching AI about symmetry could make prenatal diagnosis more reliable

Despite millions of years of evolution, the development of the human fetus remains a complex process that is vulnerable to genetic, environmental, and metabolic disruptions. These disruptions could result in developmental abnormalities, starting with one of the first organs to begin developing in the fetus that continues to develop after birth: the brain.

In a routine prenatal ultrasound, the sonographer will flag possible brain abnormalities for additional magnetic resonance imaging (MRI). But imperfect imaging can result in missed or incorrect diagnoses. 

To assist with the early detection and treatment of these abnormalities, researchers at MIT, Boston Children’s Hospital, and Inria have developed and trained a deep learning model called E(3)-Pose that helps image the fetus’s head while it’s in motion to capture a high-resolution image that the radiologist can examine. A paper about the work was presented at the 2026 European Conference on Computer Vision. 

“One of the problems with having the baby move is that the slices are all over the place,” says first author Ramya Muthukrishnan. Muthukrishnan is an MIT PhD student in the lab of Sunlin (1966) and Priscilla Chou Professor of Electrical Engineering and Computer Science Polina Golland, who is a principal investigator at the MIT Abdul Latif Jameel Clinic for Machine Learning in Health (Jameel Clinic) and the MIT Computer Science and Artificial Intelligence Laboratory (CSAIL).

By slices, Muthukrishnan is referring to a series of 2D images taken during a fetal MRI scan. These 2D planes stacked together effectively make up one high-resolution, 3D image of the fetus’s head. But if the head shifts during imaging, each 2D image “slice” could end up being taken from a different angle, resulting in gaps in the final scan and unusual slice orientation. These problems may cause the physician reviewing the MRI to miss or incorrectly diagnose a pathology in the fetal brain.

“There are very fine details that you need a very good volumetric picture of the brain to see,” explains Muthukrishnan. “If you have coverage gaps, you can’t see critical regions, and you need to be able to see well to detect abnormal development.”

What makes E(3)-Pose particularly good at tracking fetal motion is its focus on physics. Rather than simply training the model to learn physics from the data, the very principles of physics are baked into the neural network: left-right symmetry of the fetal head is embedded in the model architecture, making it easier for the model to accurately identify the brain’s orientation. 

When Muthukrishnan began working on the project at the start of her PhD, she had no background in physics, let alone MRI physics. Most of her experience was in computer science. “I was surprised when I had to go and implement this on the scanner,” she recalls. “I was not prepared for how many times I would crash the scanner because putting things on hardware and getting it to work on hardware is a different beast than training a model to run on data on a GPU.” 

Despite the challenges, Muthukrishnan persisted. She remembers shadowing a radiologist who performed fetal MRI scans at Boston Children’s Hospital at the start of her PhD and observing the shared frustration of the patient and radiologist: the pregnant mother felt uncomfortable after spending an hour in an MRI, only to quit the scan halfway; the radiologist was unable to procure a full, high-quality MRI to monitor the development of the fetus.

“Seeing frustration about it helped me decide what I wanted to do with my PhD, because it felt like this was a pain point,” Muthukrishnan says.

Fortunately, she had help: Co-author Borjan Gagoski, an assistant professor of radiology at Harvard Medical School and a researcher at Boston Children's Hospital, developed a sequence that acquires additional MRI scans to assist with training E(3)-Pose and worked with Muthukrishnan to help troubleshoot the scanner whenever it crashed. Co-author and Boston Children’s Research Assistant Aryn Lee recruited the pregnant volunteers and assisted with acquiring each of the MRI scans. Senior authors who helped guide the project include Elfar Adalsteinsson, MIT professor of electrical engineering and computer science; Ellen Grant, the Käthe Beutler Professor of Pediatrics and professor of radiology at Harvard Medical School; and Golland. Benjamin Billot, a researcher at Inria and a former postdoc in Golland’s group, helped Muthukrishnan get started with the initial method he developed.

“Ramya’s work elegantly combines geometric concepts like equivariance and symmetry with the power of deep learning to change how MRI scanners acquire the images, and to ultimately help the physicians detect and treat problems early in development. It’s an impressive research project that is deep and fundamental in how it models image acquisition, with clear clinical impact,” says Golland.

Grant, who directs the Fetal Neonatal Neuroimaging and Developmental Science Center (FNNDSC) at Boston Children’s, adds that Muthukrishnan’s work “will automate fetal brain MRI, replacing today's slow, manual process with faster and more comfortable scans for mothers, while also providing higher-quality images for radiologists to interpret.”

E(3)-Pose’s applications could extend beyond motion tracking for fetal MRIs. It could assist with imaging other hard-to-image anatomy, and even autonomous driving applications, where limited sensors need to capture noisy, 3D data in order to track 3D objects.

For now, Muthukrishnan hopes to improve E(3)-Pose so that it can help with imaging second-trimester or younger fetuses, as the model is primarily trained on third trimester fetuses. Second-trimester fetuses are “so small, they move around a lot more because there’s more space in the uterus,” Muthukrishnan says. “They’re more important to focus on because the earlier you detect brain abnormalities, the easier it is to intervene.” 

This work is supported by the MIT Jameel Clinic, the MIT Takeda Program, the Chou Family Transformative Research Fund, the National Institute of Biomedical Imaging and Bioengineering, and the Eunice Kennedy Shriver National Institute of Child Health and Human Development.

de MIT News https://ift.tt/sP8fRSa

Giving AI a “crystal clear” view of the materials world

In principle, artificial intelligence could make finding the right material for an application as simple as ordering dinner: Scientists would tell a large language model what they want, and it would serve them a perfectly designed crystal for the job.

Unfortunately, that approach doesn’t work well yet, and the problem comes down to data: Most materials consist of crystals, which are 3D periodic arrangements of atoms whose structures are stored in specialized formats, not the text streams that today’s large language models are trained on. The experimental literature is also heavily biased toward a few popular classes of crystals, making AI outputs skew toward those materials.

Now MIT researchers have created a multimodal AI model that tackles these problems. The model combines 3D atomic structures with natural language descriptions to create rich linguistic descriptions about the structures from experimental research papers and first-principles computational databases. The unified description outlines not just the chemical composition and structure, but also its properties, processing considerations, and potential applications, helping AI learn more about what makes materials impactful.

In a new paper in Nature Communications, the researchers showed their system could identify better material designs than models trained on text streams alone by allowing the model to “see” the atomic arrangements and bonding patterns. For example, when the researchers asked the model to identify materials suited for solid-state lithium-ion battery electrolytes, its top results included materials that had been independently identified as promising candidates. The researchers also showed their model learned geometric representations of material structures in ways text-only models could not.

“Structure is really at the heart of the materials science field,” says Ju Li, the Carl Richard Soderberg Professor in Power Engineering in the Department of Materials Science and Engineering (DMSE). “If you understand the structure, you can infer quite a lot about the properties and performance of materials. We wanted to build the connection between a material’s atomic structure and its properties and processing for the model.”

Joining Li on a paper about the model are Yang Jeong Park, first author and former MIT postdoc; Mayank Kumaran, a recent University of Illinois Urbana-Champaign computer engineering graduate; Chia-Wei Hsu SM ’26; and Elsa Olivetti, the Jerry McAfee (1940) Professor in Engineering in DMSE.

Getting models to see in 3D

Materials scientists are always on the hunt for new crystals that will lead to the products of the future. Over the last decade, AI has reshaped that process, proposing and designing new crystalline structures, which have highly ordered atomic structures and make up more than 90 percent of solid materials on Earth. Most of those models treat solid crystalline materials as 3D graphs.

The rise of large language models has given researchers another tool for innovation. But existing large language models rely heavily on text from published papers and books, limiting their knowledge of the physical world, since many potential crystal structures are stored in computational databases. Text-based models also lack understanding of visual geometry.

To overcome those challenges, the researchers decided to create a vision-language model for crystalline materials.

They trained a language model to create textual descriptions of crystal structures using an approach known as contrastive learning, in which models compare similar and dissimilar examples of structures and textual descriptions to achieve a unified understanding. This is similar to aligning the visual and language regions of the brain.

Following training, the researchers compared the predictions of two versions of the model: one trained on a dataset containing the AI-generated descriptions and the other trained on a dataset containing text from papers only. They found that the model trained on AI-generated descriptions successfully learned how to align geometric structures with text descriptions while the other did not.

“We wanted to create an AI model that works broadly across materials,” Li says. “That’s why we decided to synthesize narratives rather than rely solely on existing papers. We found when we used that narrative dataset, the models are better at predicting the function.”

Accelerating materials science

The researchers then showed their model could be given a natural language query describing a property and identify crystalline materials with that property, outperforming a baseline.

“There are sometimes these step changes within materials science around algorithmic integration toward useful tools, which the community uses to further advance the field,” Olivetti says. “This approach helps us see more deeply into three-dimensional structures and link structures to property and processing. I hope this can be one of those step changes.”

Li compares the model to AlphaFold, an AI model that links protein structures with function that has helped advance biological research and drug development.

The researchers believe their new approach can work with many crystalline materials, and they hope other material scientists will use it to accelerate their work.

“Being able to see a three-dimensional crystal and how its atoms are connected is super foundational,” Li says. “I think this will be useful to anyone who practices materials science.”

The work was supported, in part, by the U.S. Defense Threat Reduction Agency through the University Research Alliance; the U.S. Defense Advanced Research Projects Agency; and the National Research Foundation of Korea.

de MIT News https://ift.tt/maLnw0Q

Fostering community amid Kendall Square’s innovation

Kendall Square in Cambridge, Massachusetts, is often cited as the most innovative square mile in the world. But when the Kendall Square Association gathered local stakeholders for its 18th Annual Meeting Thursday, the conversation focused on innovation primarily in the context of community.

The event, which took place at the nonprofit startup incubator The Engine, attracted people from across Cambridge, including business leaders, elected officials, local residents, and students, for a night of local food and conversation about the advance of technologies like artificial intelligence and what they mean for society.

“How can we make AI not an enemy of work, but a tool for workers to expand their capabilities and build new expertise or do new things?” asked MIT Institute Professor Daron Acemoglu in a keynote talk. “The world is at a point in time where we need new innovations and new adaptations to growing challenges.”

Acemoglu’s keynote was one of several presentations at the event, which also featured music from the Kendall Square Orchestra, a vote on new KSA board members, and awards for organizations making a positive impact through innovation.

KSA is a nonprofit made up of over 150 organizations across Kendall Square, from major companies to universities, research organizations, and the independent shops and restaurants that give Kendall Square its distinct character.

“People come to Kendall Square to make the world better,” said KSA Executive Director Beth O’Neill Maloney in her opening remarks. “People come here to improve the human condition, people come here to solve big problems, and people come here to innovate for good.”

Acemoglu, who shared the Nobel Prize in economics in 2024 for his work studying how institutions affect prosperity, spoke on the subject of his new book, “What Happened to Liberal Democracy?”

Acemoglu says liberal democracy is under siege as trust in liberal institutions decline, political polarization rises, and people lose faith in elected officials. He acknowledged that the shared prosperity liberal democracy promised has not always been delivered, especially over the last 50 years, but insisted it is the best path forward and in need of renewed support.

“I believe liberal democratic institutions, together with the norms that support them, are one of the greatest human achievements,” Acemoglu said. “It would be unthinkable for people who lived 300 years ago that we could live in societies consisting of hundreds of millions of people with diverse interests, different views, doing different things, and people still participate in politics and reach compromises, and we have state institutions that, by and large, actually provide services and do things for the people, and that we build opportunities for prosperity from which most people benefit while social freedoms are largely upheld.”

Cambridge Mayor Sumbul Siddiqui also spoke at the event. Siddiqui began her talk by telling the story of Cambridge residents working with biotech labs in the 1970s to create a groundbreaking policy on recombinant DNA research, which helped turn Kendall Square into the life science hub it is today.

“All these years later I think that story captures what makes Kendall Square so special: It’s our willingness to work together even when we face big challenges,” Siddiqui said. “It isn’t just the universities or labs, it’s also the culture of collaboration among nonprofits, governments, and residents.”

Broad Institute Chief Communications Officer Lee McGuire started his talk by reviewing a 2009 report on what made Kendall Square such a powerful driver of innovation, noting that factors like talent density, education, and a collective voice remain strong in the area today.

“This report was brilliant in that it captures so many of the special things that matter for a place like Kendall,” McGuire said. “It is the job of the Kendall Square Association and all of us to help protect those things, make sure we name them, understand them, and see them as essential elements of making sure this community stays what it is no matter what happens. As long as there’s a good idea, this is the place that’s going to get that idea moving and scale it up to change the world.”

de MIT News https://ift.tt/pnDUBhw