Computers Math

A new study led by investigators from Mass General Brigham has found that ChatGPT was about 72 percent accurate in overall clinical decision making, from coming up with possible diagnoses to making final diagnoses and care management decisions. The large-language model (LLM) artificial intelligence chatbot performed equally well in both primary care and emergency settings across all medical specialties. The research team’s results are published in the Journal of Medical Internet Research.
“Our paper comprehensively assesses decision support via ChatGPT from the very beginning of working with a patient through the entire care scenario, from differential diagnosis all the way through testing, diagnosis, and management,” said corresponding author Marc Succi, MD, associate chair of innovation and commercialization and strategic innovation leader at Mass General Brigham and executive director of the MESH Incubator. “No real benchmarks exists, but we estimate this performance to be at the level of someone who has just graduated from medical school, such as an intern or resident. This tells us that LLMs in general have the potential to be an augmenting tool for the practice of medicine and support clinical decision making with impressive accuracy.”
Changes in artificial intelligence technology are occurring at a fast pace and transforming many industries, including health care. But the capacity of LLMs to assist in the full scope of clinical care has not yet been studied. In this comprehensive, cross-specialty study of how LLMs could be used in clinical advisement and decision making, Succi and his team tested the hypothesis that ChatGPT would be able to work through an entire clinical encounter with a patient and recommend a diagnostic workup, decide the clinical management course, and ultimately make the final diagnosis.
The study was done by pasting successive portions of 36 standardized, published clinical vignettes into ChatGPT. The tool first was asked to come up with a set of possible, or differential, diagnoses based on the patient’s initial information, which included age, gender, symptoms, and whether the case was an emergency. ChatGPT was then given additional pieces of information and asked to make management decisions as well as give a final diagnosis — simulating the entire process of seeing a real patient. The team compared ChatGPT’s accuracy on differential diagnosis, diagnostic testing, final diagnosis, and management in a structured blinded process, awarding points for correct answers and using linear regressions to assess the relationship between ChatGPT’s performance and the vignette’s demographic information.
The researchers found that overall, ChatGPT was about 72 percent accurate and that it was best in making a final diagnosis, where it was 77 percent accurate. It was lowest-performing in making differential diagnoses, where it was only 60 percent accurate. And it was only 68 percent accurate in clinical management decisions, such as figuring out what medications to treat the patient with after arriving at the correct diagnosis. Other notable findings from the study included that ChatGPT’s answers did not show gender bias and that its overall performance was steady across both primary and emergency care.
“ChatGPT struggled with differential diagnosis, which is the meat and potatoes of medicine when a physician has to figure out what to do,” said Succi. “That is important because it tells us where physicians are truly experts and adding the most value — in the early stages of patient care with little presenting information, when a list of possible diagnoses is needed.”
The authors note that before tools like ChatGPT can be considered for integration into clinical care, more benchmark research and regulatory guidance is needed. Next, Succi’s team is looking at whether AI tools can improve patient care and outcomes in hospitals’ resource-constrained areas.
The emergence of artificial intelligence tools in health has been groundbreaking and has the potential to positively reshape the continuum of care. Mass General Brigham, as one of the nation’s top integrated academic health systems and largest innovation enterprises, is leading the way in conducting rigorous research on new and emerging technologies to inform the responsible incorporation of AI into care delivery, workforce support, and administrative processes.
“Mass General Brigham sees great promise for LLMs to help improve care delivery and clinician experience,” said co-author Adam Landman, MD, MS, MIS, MHS, chief information officer and senior vice president of digital at Mass General Brigham. “We are currently evaluating LLM solutions that assist with clinical documentation and draft responses to patient messages with focus on understanding their accuracy, reliability, safety, and equity. Rigorous studies like this one are needed before we integrate LLM tools into clinical care.” More

Social risk factors such as financial instability and housing insecurity are increasingly recognized as influencing health. But unlike diagnosis codes, prescription information, lab or other test reports, social risk factors do not adhere to standardized, controlled terminology in a patient’s electronic medical record, making this information difficult to extract from the clinical notes where they typically are found.
A new study has found that a natural language processing (NLP) system developed by Regenstrief Institute and Indiana University Richard M. Fairbanks School of Public Health informaticians showed excellent performance when ported to a new health system and tested on more than six million clinical notes of patients seen in Florida. Performance was evaluated for generalizability and portability, defined as ease and accuracy when deploying the software in a new environment and of updating its use to meet the needs of new data.
“Social factors have a great impact on our health. It’s not just the medical care that we receive, but it’s also the places where we live, the places where we work and our access to food and transportation and other resources that have a major influence on our health,” said Chris Harle, PhD, the Regenstrief and IU Fairbanks School faculty member who is senior author on the study. “It’s important for the clinicians and health systems providing medical care to know about people’s social risk factors so when prescribing medications, ordering tests or planning to perform a procedure, they can better treat the whole person — perhaps with lower cost drugs or alternative sources for tests — and can also link them to services that help address their needs for a safe place to live and healthy food to eat.”
In this study, the researchers’ NLP rule-based model searched through text that physicians or other clinicians had written in the clinical notes of patients’ electronic health records, looking for key words or phrases that were likely to indicate difficulty with housing (for example: lack of permanent address) or financial needs (for example: inability to afford follow-up care) of patients at a healthcare system in a new and quite different geographic area. In spite of challenges (for example: name of a homeless shelter without indication of the facility’s function or regional variation or local nuances in language), the research scientists verified that the NLP models, with relatively simple modifications, could deliver highly accurate performance as compared to the gold standard of human review.
“Is a patient diagnosed with diabetes? It’s relatively easy to find that information in an electronic health record because the same words and codes are more likely to be used in health systems in central Indiana as are used in Florida or elsewhere in the U.S. But social risk factors don’t have nearly as established and widely used words, phrases or codes to identify them. Therefore, it’s harder to search through and determine a patient has a financial need than it is to say a patient has diabetes,” said Dr. Harle. “Our work is important for patients because ultimately their health is related to a variety of factors in their life, including social factors. For example, are clinicians incorporating in their decision making a patient’s ability to recover from a surgery as it’s going to be different if they have stable housing versus unstable housing?
“The more that we can disseminate and adapt natural language processing and other artificial intelligence methods that fully describe a patient to give clinicians a full 360 understanding of patients’ needs, the better. If we can extract social information more efficiently, it’s less costly. Then we can start to take what we’d call a population health perspective. So, if a health system can efficiently identify the patients who have housing instability — the population of patients who have this need — then the healthcare system may be able to employ a more proactive population-based intervention to serve that whole group of people, connecting them, for example, to the housing services in the community or financial resources that might be available.”
Dr. Harle, an information scientist and health services researcher who focuses on the design, adoption, use and value of health information systems, notes that this study was a team effort across multiple institutions of professionals who work in the clinical arena (including individuals who study how patients access and use care), public health, population health and healthcare administration as well as technically knowledgeable and skilled systems specialists. “Bringing people together who have that diversity of understanding leads to pragmatically useful studies like this one,” he said. More

As children head back to school, two issues have climbed higher on their parents’ list of concerns: the role of social media and the internet in kids’ lives.
Over half of parents also rate mental health issues as leading health concerns for children and teens, according to the University of Michigan Health C.S. Mott Children’s Hospital National Poll on Children’s Health.
Overall, emotional health and technology use dominated this year’s top 10 list of parent concerns about health-related issues for kids in the U.S.- surpassing childhood obesity, which parents rated the number one children’s health issue a decade ago.
“Parents still view problems directly impacting physical health, including unhealthy eating and obesity, as important children’s health issues. But these have been overtaken by concerns about mental health, social media and screen time,” said Mott Poll co-director and Mott pediatrician Susan Woolford, M.D., M.P.H.
Two-thirds of parents are worried about children’s increased time on devices, including overall screen time and use of social media, taking the No.1 and No.2 spots on the list of children’s health concerns this year, according to the nationally representative poll.
“Children are using digital devices and social media at younger ages, and parents may struggle with how to appropriately monitor use to prevent negative impacts on safety, self-esteem, social connections and habits that may interfere with sleep and other areas of health,” Woolford said.
Screen time became a growing concern for parents during the pandemic, previous reports suggest. Woolford encourages parents to regularly evaluate their children’s use of technology and consider limiting use if they notice signs of unhealthy interactions or behaviors. Certain social media and device settings can also help protect kids. More

In recent years, pneumatic artificial muscles (PAMs) have emerged as promising actuators for simulating human-like movements, with prominent applications in various industries including robotics, rehabilitation, and prosthetics. PAMs are usually composed of rubber and covered with braided yarn and can mimic the mechanics of human muscles. They can stiffen and contract on being supplied with pressurized air and soften and lengthen upon releasing the air. However, PAM is a nonlinear system and experiences huge latency, making it important to have control systems that can regulate their performance.
While determining a nonlinear mathematical model for PAM is challenging, researchers in the past have proposed many control methods to solve the problems associated with PAM. However, while these traditional control methods exhibit decent performance, they are not able to deal with PAM’s nonlinearity and hysteresis. Moreover, while learning control algorithms have been theoretically effective in improving PAM-based system’s performance, their implementation is practice is quite difficult.
To overcome these limitations and address this open problem, a group of researchers led by Associate Professor Ngoc-Tam BUI of the Innovative Global Program, College of Engineering, Shibaura Institute of Technology in Japan, along with Dr. Quy-Thinh Dao of Hanoi University of Science and Technology, has proposed a novel solution. In their study published in the journal Scientific Reports on 22 May 2023, they propose a control approach called “adaptive fuzzy sliding mode controller (or AFSMC)” that uses fuzzy logic (a type of computational thinking) for estimating control parameters of PAM-based systems.
“The proposed innovative control strategy leverages the Takagi-Sugeno fuzzy algorithm to estimate the disturbance component and automatically update the output variable values, demonstrating enhanced tracking accuracy and adaptability compared to traditional sliding mode control methods,” explains Associate Professor BUI.
The researchers first developed a sliding mode controller with a control signal that incorporates a special variable to estimate the disturbances and improve the control performance. Next, they designed an adaptive fuzzy algorithm, wherein parameter vectors of the component rules are automatically updated by an adaptive law, to compute the disturbance variable. The stability of the developed ASFMC algorithm was then analyzed using the Lyapunov stability condition (used to study the stability of a nonlinear system). Furthermore, the researchers conducted a series of experiments to assess the performance of their controller by comparing it with traditional sliding mode control methods.
Remarkably, the AFSMC approach exhibited improved tracking accuracy, with a root mean square error value of 2.68° at a frequency of 0.5 Hz under load, while the sliding mode controller approach displayed a higher value of 4.21°. Moreover, it showed exceptional adaptability to abrupt external disturbances. Explaining these results further, Associate Professor BUI says, “In a comparative evaluation against the well-known commercial rehabilitation system, LOKOMAT, the AFSMC controller delivered similar performance. It also exhibited superior adaptability to sudden load changes, swiftly returning to the desired trajectory by manipulating its control output.”
These findings thus point to the potential of the novel AFSMC approach for integration into robotic rehabilitation devices, assistive devices, and physical therapy equipment for precise and personalized therapy. Moreover, this approach can aid in the design and development of advanced prosthetic limbs for enhanced functionality and rehabilitation outcomes.
Talking about the long-term implications of this study, Associate Professor BUI says: “With the outcomes of this research, the emergence of a commercial rehabilitation system actuated by PAM can be anticipated within the next 5 to 10 years. This innovative system will provide significant benefits to patients, including those with spinal cord injuries and stroke and others requiring rehabilitation.”
While this research has laid the groundwork for advancing trajectory tracking control in PAM systems, we hope that it ignites further exploration and development in the field of rehabilitation technology. More

In many cases, cells are very active in their movement and serve as power generators. The ability of cells to produce physical forces is one of the basic functions of the body. When running, for example, the forces generated in the cells cause the muscles to contract and the breath to work. It has been possible to measure even the forces experienced by individual proteins by force sensors developed in the past, but previously intracellular forces and mechanical strains could not have been measured.
Together with the scientists from The Ohio State University OSU, cell biology researchers at Tampere University have developed a force sensor that can be attached to the side of a mechanically responding protein, allowing it to sense forces and strain on the protein within the cell.
The development of the micro-sized sensor began on a conference travel in December 2019.
“The power-sensing part is like a rubber band that changes colour when stretched. This part is attached to the antibodies at both ends of the rubber band, which bind to the cellular target protein under study. The force or elongation of the studied protein can then be detected under a microscope by following the elongation of the rubber band, i.e. the colour it produces,” says Teemu Ihalainen,a Senior Research Fellow from BioMediTech at Tampere University.
According to Ihalainen, the force sensor, which is only about twenty nanometres in size, can be easily generalised to a wide range of cell biology research and various target proteins. With the help of the protein biosensor, forces can be measured, for example, in the nuclear membrane, between different proteins, or generally in the cytoskeleton of the cell. It allows the mechanics of the cell to be transformed into visible form for the first time. There has already been great interest in this technology in various laboratories in Japan, India, Norway and the United States.
Internal forces of the cell provide information on the mechanics of cancer
Cells are subject to forces all the time, both in normal bodily functions and diseases. More

Researchers at the National University of Singapore (NUS) have developed a novel aero-elastic pressure sensor, called ‘eAir’. This technology can be applied to minimally-invasive surgeries and implantable sensors by directly addressing the challenges associated with existing pressure sensors.
The eAir sensor promises increased precision and reliability across medical applications. It can potentially transform laparoscopic surgeries by enabling tactile feedback for surgeons, allowing more precise manipulation of patient tissues. In addition, the sensor can improve patient experiences by offering a less invasive means of monitoring intracranial pressure (ICP), a key health metric for individuals with neurological conditions.
Led by Associate Professor Benjamin TEE from the NUS College of Design and Engineering and NUS Institute for Health Innovation & Technology, the research team’s findings were recently published in scientific journal Nature Materials on 17 August 2023.
From lotus leaf to laboratory: Harnessing nature’s design
Conventional pressure sensors frequently struggle with accuracy. They have trouble delivering consistent readings, often returning varying results when the same pressure is applied repeatedly and can overlook subtle changes in pressure — all of which can lead to significant errors. They are also typically made from stiff and mechanically inflexible materials.
To address these challenges in pressure sensing, the NUS team drew inspiration from a phenomenon known as the ‘lotus leaf effect’ — a unique natural phenomenon where water droplets effortlessly roll off the leaf’s surface, made possible by its minuscule, water-repelling structures. Mimicking this effect, the team has engineered a pressure sensor designed to significantly improve the sensing performance.
“The sensor, akin to a miniature ‘capacity meter’, can detect minute pressure changes — mirroring the sensitivity of a lotus leaf to the extremely light touch of a water droplet,” explained Assoc Prof Tee. More

Environmental regulators and other organizations should do more scientific experimentation to inform natural resource policy, according to an international group of economists that includes University of Wyoming researchers.
In a new paper in the  journal Science, the economists say more frequent use of up-front experiments would result in more effective environmental policymaking in areas ranging from pollution control to timber harvesting across the world.
“Although formal experimentation is a cornerstone of science and is increasingly embedded in nonenvironmental social programs, it is virtually absent in environmental programs,” the researchers wrote. “Strengthening the culture of experimentation in the environmental community will require changes in norms and incentives.”
The paper acknowledges that scientists and practitioners can legitimately argue about how much time and effort should be given to experiments in environmental policy, but it contends that the current allocation of roughly zero percent is suboptimal.
The paper was produced by The Teton Group, an initiative led by Professor Todd Cherry, the John S. Bugas Chair in UW’s Department of Economics. The prominent group of economists meets every fall in Wyoming to discuss critical ideas that impact environmental policy and economic development. Members include colleagues from UW and scholars in behavioral environmental policy from Carnegie Mellon University, Johns Hopkins University, Purdue University, the University of Texas-Austin, the University of Wisconsin-Madison and several key European universities. The group of UW economists include Todd Cherry, Jacob Hochard, Stephen Newbold, Jason Shogren, Linda Thunström and Klaas van ‘t veld.
“Guesswork is expensive, so we need to apply tools that reduce uncertainty about what works and what doesn’t,” Cherry says. “Lessons learned can improve current and future policy.”
According to the new paper, environmental scientists and practitioners typically rely on field experience, case studies and retrospective evaluations of programs that were not designed to generate evidence about cause and effect. The result can be ineffective or even counterproductive programs. More

Osaka Metropolitan University scientists have developed an AI model that accurately estimates a patient’s age, using chest radiographs of healthy individuals collected from multiple facilities. Furthermore, they found a positive relationship between differences in the AI-estimated and chronological ages and a variety of chronic diseases, such as hypertension, hyperuricemia, and chronic obstructive pulmonary disease. In the future, it is expected that AI biomarkers will be developed to predict life expectancy, estimate the severity of chronic diseases, and forecast surgery-related risks.
What if “looking your age” refers not to your face, but to your chest? Osaka Metropolitan University scientists have developed an advanced artificial intelligence (AI) model that utilizes chest radiographs to accurately estimate a patient’s chronological age. More importantly, when there is a disparity, it can signal a correlation with chronic disease. These findings mark a leap in medical imaging, paving the way for improved early disease detection and intervention. The results are set to be published in The Lancet Healthy Longevity.
The research team, led by graduate student Yasuhito Mitsuyama and Dr. Daiju Ueda from the Department of Diagnostic and Interventional Radiology at the Graduate School of Medicine, Osaka Metropolitan University, first constructed a deep learning-based AI model to estimate age from chest radiographs of healthy individuals. They then applied the model to radiographs of patients with known diseases to analyze the relationship between AI-estimated age and each disease. Given that AI trained on a single dataset is prone to overfitting, the researchers collected data from multiple institutions.
For the development, training, internal and external testing of the AI model for age estimation, a total of 67,099 chest radiographs were obtained between 2008 and 2021 from 36,051 healthy individuals who underwent health check-ups at three facilities. The developed model showed a correlation coefficient of 0.95 between the AI-estimated age and chronological age. Generally, a correlation coefficient of 0.9 or higher is considered to be very strong.
To validate the usefulness of AI-estimated age using chest radiographs as a biomarker, an additional 34,197 chest radiographs were compiled from 34,197 patients with known diseases from two other institutions. The results revealed that the difference between AI-estimated age and the patient’s chronological age was positively correlated with a variety of chronic diseases, such as hypertension, hyperuricemia, and chronic obstructive pulmonary disease. In other words, the higher the AI-estimated age compared to the chronological age, the more likely individuals were to have these diseases.
“Chronological age is one of the most critical factors in medicine,” stated Mr. Mitsuyama. “Our results suggest that chest radiography-based apparent age may accurately reflect health conditions beyond chronological age. We aim to further develop this research and apply it to estimate the severity of chronic diseases, to predict life expectancy, and to forecast possible surgical complications.” More