#7 | A Surgical Robot For Space

Quote of the Week 💬 

Surgical Robot Performs Simulated Operation In Space

⌛ The Seven Second Summary: A team of researchers from the University of Nebraska successfully performed a remote surgery demo in zero gravity from Houston, Texas and other facilities on Earth.

🔬 How It Was Done:

• The robot, known as spaceMIRA, is a small, compact device. It only weighs 2 pounds (0.9 kilograms), and uses two arms to mimic the movements of a human.

• spaceMIRA was sent to the International Space Station, approximately 250 miles (400 kilometers) above Earth to perform its demo.

• The demo required surgeons to use spaceMIRA to provide tension to a simulated tissue space made of rubber bands, using one robotic hand. With the other robotic hand, the surgeons used scissors to dissect the elastic tissue.

• 6 different surgeons performed remote tests with the robot, and each demonstration focused on dissecting the correct piece of rubber band tissue.

🧮 Key Results: spaceMIRA is the first surgical robot aboard the International Space Station and it is one of the first times remote surgery tasks have been tested in space. 

💡 Why This May Matter: Today, the most common telemedicine use case connects patients with medical experts for remote discussions and consultations. However, the success observed in this remote surgical demo indicates a potential future where medical experts can perform surgeries and other complex tasks remotely to assist patients where specialists would not otherwise be available.

🔎 Elements To Consider: The locations where remote surgeries would provide the most value might also be the most challenging to transport and maintain the necessary surgical equipment. These are just a few of the many factors that need to be addressed before remote surgeries become a viable medical option of the future.

📚 Learn More: University Nebraska-Lincoln.

Stat of the Week 📊 

Neural Prosthetic Device Helps People Restore Their Memory

40%

⌛ The Seven Second Summary: Scientists from Wake Forest University and the University of Southern California successfully used a prosthetic device to help patients improve their memory recall by providing electrical pulses during visual recognition tasks.

🔬 How It Was Done:

• 14 adults with epilepsy participated in a study where they received surgically implanted sensors to map their brains in order to locate the origin of their seizures.

• The researchers also designed a computational model to illustrate the neural activity their brains use to remember different types of information as distinct patterns.

• These decoded patterns were used to send controlled, electrical pulses to specific parts of the patients’ brains as their brains attempted to store different types of information during various visual recognition tasks.

🧮 Key Results: Electrical stimulation resulted in significant changes in memory performance, with approximately 22% of cases showing a noticeable difference overall. Among participants with impaired memory function who received electrical stimulation on both sides of their brain, nearly 40% of the participants showcased significant memory recall improvements.

💡 Why This May Matter: People experience various neurological disorders, injuries and conditions that result in different types of memory, sensory and motor function issues. Computational models are showing to be valuable resources for researchers to understand how these neurological issues are distinct from one another and prescribe more precise and effective treatments for ailments in the future.

🔎 Elements To Consider: These patients were only tested on their short-term memory recall with this technology. It is unclear if a similar technique would be effective for long-term memory recall.

📚 Learn More: Wake Forest. Frontiers.

AI x Science 🤖

Meta’s New AI Model Learns By Watching Videos

Last week Meta released its latest AI model, V-JEPA. This model is trained to understand and predict the context of events in a video through a method that is similar to how humans interpret videos.

What makes this model unique is the way it was trained. First, the model did not receive any labeled data to identify objects or events in the videos it received. More importantly, the researchers used a masking methodology to block out large portions of a video and/or clip out significant parts of a video, so the model was forced to predict the context of what was taking place in the video based on the limited information it had available to it.

For example, imagine there was a training video of a child who ate a bowl of chocolate. The model would only receive the first and last portions of the video and had to predict the events that led to the child getting chocolate smeared across its mouth. Another type of training video could be a clip of a cat sleeping, then waking up and getting food. Everything in the video would be blurred or blacked out, except for a view of the cat’s face and legs, and the model had to predict what was taking place.

These types of video abstraction exercises are easy for humans to interpret and understand, but difficult for AI models because these systems are known to be extremely pixel sensitive. Nonetheless, V-JEPA became capable at this task and outperformed similar model methodologies. Furthermore, V-JEPA had a 1.5x - 6x improvement with its computational efficiency during pre-training. This marks another example of researchers who have improved model performance and efficiency by experimenting with innovative machine reasoning techniques. Meta AI. Github.

Our Full AI Index

• Research: Scientists from the U.S. Department of Energy used AI systems to identify and simulate the performance of 6 new materials with high potential to absorb CO₂, all accomplished with just 12 hours of compute time. Without the use of AI, it would have taken weeks, if not months to accomplish this feat. For our long-time readers, this may sound similar to the story we highlighted in early January about Microsoft and PNNL researchers who identified 23 promising new battery materials with just 80 hours. Argonne. Nature.

• Business: Google announced its most capable AI model yet, Gemini 1.5. The previous version of Gemini had a context window of 32,000 tokens. Google claims this new version is capable of running up to 1 million context tokens, allowing the model to process, contextualize, and remember vast amounts of information at once. For context, 1 million context tokens roughly translates to accommodating 1 hour of video, 11 hours of audio, code bases with over 30,000 lines of code, or over 700,000 words. Google. 

• Multilingual AI: The research team at Cohere For AI released an open-source large language model supporting 101 languages — more than double the number of languages covered by existing open-source models. Cohere. arXiv.

• Cultural Events: An artist from Barcelona designed a hologram using AI. This summer, she plans to marry her holographic companion as part of a social experiment. Euronews.

• Policy: State legislatures throughout the U.S. are proposing AI-related bills left and right in an attempt to keep up with the technology’s rapid development. According to Axios, as of February 7th, there were 407 total AI-related bills across more than 40 states, which is up from 67 bills just a year ago. Axios.

Other Observations 📰

The American Heart Association Turns 100

The American Heart Association is celebrating its 100th anniversary this year. As a celebration of its longevity and impact, the organization released a report of the progress that has been made to combat heart disease and stroke over the past century, as well as underscore the challenges that need to be overcome to continue steady progress in these areas in the future.

The top line highlights of the report are encouraging. From 1950 to 2021, heart disease death rates has fallen by 70%. Furthermore, since 1988, stroke deaths rates have fallen by nearly 33%. American Heart Association. AHA Journal.

Our Full Science Index

• Energy: Brazil’s Ministry of Mines and Energy announced the country powered 92.4% of its electricity from renewables in 2023. The Rio Times.

• Space: Southwest Research Institute scientists have discovered, for the first time, water molecules on the surface of an asteroid. They detected approximately 12-ounces of water trapped in a cubic meter of soil spread across its surface. Southwest Research Institute. Planetary Science Journal.

• Cancer: A Phase III clinical trial found a combination treatment of traditional chemotherapy with a new drug nearly quadrupled the median survival of participants with an aggressive form of lung cancer 36 months after treatment, compared to the placebo-chemotherapy (11.9% vs 3.3%). Queen Mary. JAMA Network.

• Alzheimer’s: In the largest study of its kind, the profiles of 11 different proteins in the blood were shown to predict dementia up to 15 years before a clinical diagnosis with 90% accuracy. Warwick. Nature.

• Parkinson’s: After posting successful clinical trail results, nearly 1,000 people with advanced Parkinson's disease in England will receive a new treatment to manage and improve their symptoms. BBC.

• Medical Approvals: Last but not least, two noteworthy medical approvals worth learning about.

  • The FDA approved the first cellular therapy to treat patients with unresectable or metastatic melanoma, which is an aggressive form of skin cancer that cannot be removed through surgery. FDA.

  • Biogen announced the European Commission approved its treatment for Friedreich’s ataxia, which is an inherited gene disorder that progressively debilitates some of the body's nerves. According to Biogen, its therapy is the first treatment approved within the European Union for this disease. Biogen.

Media of the Week 📸 

OpenAI Unveils Sora, Its New Text-To-Video Generator

Open AI announced its forthcoming release of a text-to-video model, Sora. The model can generate videos up to a minute long from text prompts alone. This video is a demo of several examples the company is showcasing, and the quality is a significant leap forward from similar tools that were released last year. Open AI.

Many Stars Sparkle In Hubble’s Latest Image

NASA scientists captured a photo of the star cluster NGC 2298. This particular cluster is a collection of thousands of stars held closely together by a shared gravitational attraction. NASA.

A Mushroom Is Found Sprouting Out Of A Living Frog

Researchers from the World Wildlife Fund-India discovered a frog with a mushroom growing from its body. The frog seemed completely unbothered by this, and is leading to many questions about how this is even possible. New York Times. Reptiles & Amphibians.

This Week In The Cosmos 🪐

Feb 24: A full moon.

Until Next Time 💭