Lifeboat Foundation: Safeguarding Humanity
Toggle light / dark theme

Blog

Category: robotics/AI – Page 96

Predicting cardiovascular events from routine mammograms using machine learning

Background Cardiovascular risk is underassessed in women. Many women undergo screening mammography in midlife when the risk of cardiovascular disease rises. Mammographic features such as breast arterial calcification and tissue density are associated with cardiovascular risk. We developed and tested a deep learning algorithm for cardiovascular risk prediction based on routine mammography images.

Methods Lifepool is a cohort of women with at least one screening mammogram linked to hospitalisation and death databases. A deep learning model based on DeepSurv architecture was developed to predict major cardiovascular events from mammography images. Model performance was compared against standard risk prediction models using the concordance index, comparative to the Harrells C-statistic.

Results There were 49 196 women included, with a median follow-up of 8.8 years (IQR 7.7–10.6), among whom 3,392 experienced a first major cardiovascular event. The DeepSurv model using mammography features and participant age had a concordance index of 0.72 (95% CI 0.71 to 0.73), with similar performance to modern models containing age and clinical variables including the New Zealand ‘PREDICT’ tool and the American Heart Association ‘PREVENT’ equations.

The scientist using AI to hunt for antibiotics just about everywhere

When he was just a teenager trying to decide what to do with his life, César de la Fuente compiled a list of the world’s biggest problems. He ranked them inversely by how much money governments were spending to solve them. Antimicrobial resistance topped the list.

Twenty years on, the problem has not gone away. If anything, it’s gotten worse. Infections caused by bacteria, fungi, and viruses that have evolved ways to evade treatments are now associated with more than 4 million deaths per year, and a recent analysis, published in the Lancet, predicts that number could surge past 8 million by 2050. In a July 2025 essay in Physical Review Letters, de la Fuente, now a bioengineer and computational biologist, and synthetic biologist James Collins warned of a looming “postantibiotic” era in which infections from drug-resistant strains of common bacteria like Escherichia coli or Staphylococcus aureus, which can often still be treated by our current arsenal of medications, become fatal. “The antibiotic discovery pipeline remains perilously thin,” they wrote, “impeded by high development costs, lengthy timelines, and low returns on investment.”

New AI method flags fluid flow tipping points before simulations break down

David J. Silvester, a mathematics professor at the University of Manchester, has developed a novel machine-learning method to detect sudden changes in fluid behavior, improving speed and the cost of identifying these instabilities and overcoming one of the major obstacles faced when using machine learning to simulate physical systems. The findings are published in the Journal of Computational Physics.

Computational simulations of mathematical models of fluid flow are essential for everyday applications ranging from predicting the weather to the assessment of nuclear reactor safety. The advent of this simulation capability over the past 50 years has revolutionized the development of fuel-efficient airplanes, and sail configurations on racing yachts can now be optimized in real time, providing the marginal gains needed to win races in the America’s Cup.

Optimized aerodynamics means that modern day cyclists can ride faster, golf balls fly further and Olympic swimmers consistently set world records. Computational fluid dynamics also enables the modeling of the flow of blood in the human heart, making the provision of patient-specific surgery possible.

Compression technique makes AI models leaner and faster while they’re still learning

Training a large artificial intelligence model is expensive, not just in dollars, but in time, energy, and computational resources. Traditionally, obtaining a smaller, faster model either requires training a massive one first and then trimming it down, or training a small one from scratch and accepting weaker performance.

Researchers at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL), Max Planck Institute for Intelligent Systems, European Laboratory for Learning and Intelligent Systems, ETH, and Liquid AI have now developed a new method that sidesteps this trade-off entirely, compressing models during training, rather than after.

Electrofluidic fiber muscles could enable silent robotic systems

Muscles are remarkably effective systems for generating controlled force, and engineers developing hardware for robots or prosthetics have long struggled to create analogs that can approach their unique combination of strength, rapid response, scalability, and control. But now, researchers at the MIT Media Lab and Politecnico di Bari in Italy have developed artificial muscle fibers that come closer to matching many of these qualities.

Like the fibers that bundle together to form biological muscles, these fibers can be arranged in different configurations to meet the demands of a given task. Unlike conventional robotic actuation systems, they are compliant enough to interface comfortably with the human body and operate silently without motors, external pumps, or other bulky supporting hardware.

The new electrofluidic fiber muscles—electrically driven actuators built in fiber format—are described in a recent paper published in Science Robotics. The work is led by Media Lab Ph.D. candidate Ozgun Kilic Afsar; Vito Cacucciolo, a professor at the Politecnico di Bari; and four co-authors.

AI diffusion models tailor drug molecules to custom-fit protein targets, speeding drug development and evaluation

University of Virginia School of Medicine scientists have developed a bold new approach to drug development and discovery that could dramatically accelerate the creation of new medicines. UVA’s Nikolay V. Dokholyan, Ph.D., and colleagues have developed a suite of artificial intelligence-powered tools, called YuelDesign, YuelPocket and YuelBond, that work together to transform how new drugs are created. The centerpiece, YuelDesign, uses a cutting-edge form of AI called diffusion models to design new drug molecules tailored to fit their protein targets exactly, even accounting for the way proteins flex and shift shape during binding.

A companion tool, YuelPocket, identifies exactly where on a protein a drug can attach, while YuelBond ensures the chemical bonds in designed molecules are accurate. Together, the approach is poised to improve both how new drugs are designed and how quickly and efficiently existing drugs can be evaluated for new purposes.

“Think of it this way: Other methods try to design a key for a lock that’s sitting perfectly still, but in your body, that lock is constantly jiggling and changing shape. Our AI designs the key while the lock is moving, so the fit is much more realistic,” said Dokholyan, of UVA’s Department of Neurology. “This could make a real difference for patients with cancer, neurological disorders and many other conditions where we desperately need better drugs targeting these wiggly proteins but keep hitting dead ends.”

/* */