AI Spotlight: Mapping Out Links Between Drugs and Birth Defects

Updated on Nov 19, 2025 | Artificial Intelligence, Featured, Research

Avi Ma’ayan, PhD, Director of the Mount Sinai Center for Bioinformatics at the Icahn School of Medicine at Mount Sinai

Birth defects can be linked to many factors—genetic, environmental, even pure chance. Characterizing the links of any factor to congenital abnormalities is a daunting task, given the vastness of the problem.

In the face of this challenge, a team of researchers at the Icahn School of Medicine at Mount Sinai tapped artificial intelligence (AI) methods to shed light on associations between existing medications and their potential to induce specific birth abnormalities.

“We wanted to improve our understanding of reproductive health and fetal development, and importantly, warn about the potential of new drugs to cause birth defects before these drugs are widely marketed and distributed,” says Avi Ma’ayan, PhD, Professor of Pharmacological Sciences and Director of the Mount Sinai Center for Bioinformatics at Icahn Mount Sinai.

The team developed a knowledge graph—a descriptive model that maps out the relationships between entities and concepts—called ReproTox-KG to integrate data about small-molecule drugs, birth defects, and genes. In addition to constructing the knowledge graph, the team also used machine learning, specifically semi-supervised learning, to illuminate unexplored links between some drugs and birth defects.

Here’s how ReproTox-KG works as a knowledge graph to predict birth defects.

The study examined more than 30,000 preclinical small-molecule drugs for their potential to cross the placenta and induce birth defects, and identified more than 500 “cliques”—interlinked clusters between birth defects, genes, and drugs—that can be used to explain molecular mechanisms for drug-induced birth defects. Findings were published in Communications Medicine on July 17, and the platform has been made available on a web-based user interface.

In this Q&A, Dr. Ma’ayan, senior author of the paper, discusses ReproTox-KG and its potential impacts.

What was the motivation for your study?

The motivation for the study was to find a use case that combines several datasets produced by National Institutes of Health (NIH) Common Fund programs to demonstrate how integrating data from these resources can lead to synergistic discoveries, particularly in the context of reproductive health.

The study identifies some relationships between approved drugs and birth defects to identify existing drugs that are currently not classified as harmful but which may pose risks to the development of a fetus. It also provides a new global framework to assess potential toxicity for new drugs and explain the biological mechanisms by which some drugs known to cause birth defects may operate.

What are the implications?

Identifying the causes of birth defects is complicated and difficult. But we hope that through complex data analysis integrating evidence from multiple sources, we can improve our understanding of reproductive health and fetal development, and also warn about the potential of new drugs to cause birth defects before these drugs are widely marketed and distributed.

What are the limitations of the study?

We have not yet experimentally validated any of the predictions. There are currently no considerations of tissue and cell type, and the knowledge graph representation omits some detail from the original datasets for the sake of standardization. The website that supports the study may not be appealing to a large audience.

How might these findings be put to use?

Regulatory agencies such as the U.S. Environmental Protection Agency or the Food and Drug Administration may use the approach to evaluate the risk of new drug or other chemical applications. Manufacturers of drugs, cosmetics, supplements, and foods may consider the approach to evaluate the compounds they include in products.

What is your plan for following up on this study?

We plan to use a similar graph-based approach for other projects focusing on the relationship between genes, drugs, and diseases. We also aim to use the processed dataset as training materials for courses and workshops on bioinformatics analysis. Additionally, we plan to extend the study to consider more complex data, such as gene expression from specific tissues and cell types collected at multiple stages of development.

Learn more about how Mount Sinai researchers and clinicians are leveraging machine learning to improve patient lives

AI Spotlight: Guiding Heart Disease Diagnosis Through Transformer Models

AI Spotlight: Forecasting ICU Patient States for Improved Outcomes

Mount Sinai’s Role in Hemodialysis: From the First Treatment in the United States to Continuing Innovations

Updated on Sep 14, 2023 | Featured, Research

It has been more than 75 years since Mount Sinai conducted the first hemodialysis treatment in the United States in 1948, a monumental accomplishment, and Mount Sinai continues to play a leading role in research to help patients in need of this lifesaving treatment.

The first type of dialyzer, called the artificial kidney, was built in 1943 by a Dutch physician, Willem Kolff, MD, PhD, working in the Netherlands during World War II. He attempted to treat more than a dozen patients with acute kidney failure over the next two years and continuously improved his machine design.

In 1947, Dr. Kolff came to the United States to demonstrate his model artificial kidney at The Mount Sinai Hospital. On January 1948, Alfred Fishman, MD, and Irving Kroop, MD, who had been trained by Dr. Kolff, used his machine for the first time to treat a patient with acute renal failure who eventually recovered completely. That first dialysis took place at 11 pm on January 26, 1948. It lasted for six hours, and it represents the first clinical use of the artificial kidney in the United States.

As this showcased the latest technology, many visitors came to the operating room galleries to see the machine in action. It was the only one in New York City at the time. Patients from other hospitals were transferred to Mount Sinai to receive treatment. When a patient was too ill to travel, they packed up the machine and drove it over.

Dr. Kolff later shared his machines with other hospitals. When he returned to Holland, one of his machines stayed at Mount Sinai.

Over the next two years, the same team continued using this machine for dialytic therapy in patients with acute renal failure.  As a result of this work, The Mount Sinai Hospital opened the first artificial kidney center in New York in 1957, which included new designs that were engineered by staff. This was led by Sherman Kupfer, MD, who spent his career at Mount Sinai and made several contributions to the study of kidney disease.

Maintenance hemodialysis therapy for patients with advanced chronic kidney disease would not start until several years later in 1960 by Belding Scribner, MD, at the University of Washington in Seattle. The main problem for chronic maintenance hemodialysis was, and still is, maintenance of an open vascular access to perform the dialysis. Long-term use of native veins for dialysis blood access leads to eventual fibrosis and disappearance of veins; therefore the need for a special vascular access to secure long-term hemodialysis (HD).

The first meaningful breakthrough for vascular access came from Dr. Scribner’s group with the advent of the externalized Quinton-Scribner shunt. This access, first described in 1960, used the newly available material Teflon in an externalized circuit with cannulas placed in the radial artery and a peripheral vein in the arm that could subsequently be attached to the dialysis circuit. Although this method proved the first reliable, longer-term access for hemodialysis, it was still prone to the many infectious and hemorrhagic sequelae of its forebears.

The real answer to this problem also came from the Mount Sinai family when in 1966 Michael J. Brescia, MD, and colleagues at the Bronx VA Hospital published their seminal paper on how to perform hemodialysis using venipuncture of a surgically created arteriovenous fistula. The implementation of the surgically created arteriovenous fistula (AVF) allowed the development of modern chronic hemodialysis with about half a million of patients undergoing in-center hemodialysis three times weekly in the United States by 2020.

Over the past 75 years, many technological improvements have been done in hemodialysis machines, but the essence of the process remains unchanged since the early 1960s. Home hemodialysis very prevalent in the 1960s, became very uncommon afterwards, but there has been a recent surge of interest in it again. During this period of time we have also seen significant development of chronic peritoneal dialysis as another modality to provide long-term dialysis as well as the establishment and improvement of kidney transplantation as another therapy for chronic kidney disease.

“During all these years, The Mount Sinai Hospital has been at the forefront of all of these changes in the area of dialysis, making sure we offer all modalities of therapy as well as best level of care to all our patients,” says Jaime Uribarri, MD, Professor of Medicine (Nephrology). In this Q&A, Dr. Uribarri talks about the future of dialysis.

Jaime Uribarri, MD,

What research is currently being done at the Icahn School of Medicine at Mount Sinai in hemodialysis? Why is it important?

Several areas of research in hemodialysis are currently being performed at Mount Sinai. For example, Evren Azeloglu, PhD, Associate Professor of Medicine (Nephrology), and Pharmacological Sciences, and a team of researchers have invented a new implantable vascular access port that diminishes the risk of bleeding and infection in preclinical studies. The port also reduces pain and discomfort and allows easy self-cannulation, which enables safe home hemodialysis. This device is currently being perfected for future clinical use.

In addition, on a different front, Lili Chan, MD, Associate Professor of Medicine (Nephrology and Internal Medicine) has been working trying to use artificial intelligence to identify symptoms and social determinants of health from the electronic health records of patients on dialysis. This would potentially allow for measures to improve treatment and management of symptoms and other unmet social determinant of health, which are associated with adverse clinical outcomes in these patients.

Hemodialysis is needed because of the inexorable progression of some forms of chronic kidney disease. The Renal Division has been intensively studying and assessing potential therapies to slow progression to end stage renal disease and therefore delaying or avoiding the eventual need for dialysis.

What challenges remain in the delivery of hemodialysis, and how is Mount Sinai addressing those?

Many challenges remain in this arena. One challenge is there is limited access to dialysis centers in the community. Mount Sinai is addressing this by expanding our outpatient dialysis units to the outer boroughs. Also, limited access to home dialysis therapies, especially for minority populations remains a concern. Overall, the United States does not have enough home hemodialysis patients, and in-center hemodialysis is burdensome. Mount Sinai is addressing this by growing its home program and bringing a significant equity lens into it. Despite the great success with arteriovenous fistulas, vascular dialysis access patency, maintaining a way to access a patient’s veins, remains a problem. Mount Sinai is addressing this with its new implantable vascular access port.

How does the future of hemodialysis look like? 

The future of hemodialysis can be seen in several developments:

  • Technological advances of the machines should make the procedure easier.
  • An increasing proportion of patients are using home dialysis instead of in-center dialysis and are using peritoneal dialysis. Mount Sinai is positioned to help response to these changes.
  • Advances in pharmacological therapies are helping to slow the progression of chronic kidney disease as well as to increase the long-term survival of kidney transplants. This should decrease the need for hemodialysis in the future.

September 14, 2023: This post has been updated to include corrections regarding the history and development of the dialysis device at Mount Sinai.

 

Learn more about dialysis services at Mount Sinai

Mount Sinai Researchers Play a Prominent Role in a Wide Probe Into Long COVID

Updated on Jul 25, 2023 | Featured, Research

RECOVER—Researching COVID to Enhance Recovery—is a nationwide initiative dedicated to understanding why some people develop long-term symptoms following COVID-19 infection. Recently, for the first time, the project yielded outcomes that are expected to help standardize the definition of long COVID toward these goals.

The researchers created a scoring system based on the symptoms that most clearly distinguished patients previously infected with COVID-19 from those who had not. Their work was published online May 25 in the Journal of the American Medical Association.

“This initiative is driven by our shared vision to deepen our understanding of one of the most perplexing maladies of our times and to inform recovery and treatment of individuals who grapple with persistent symptoms following COVID-19 infection and learn why they do, so that we can find the most targeted solutions to help patients,” says Alexander Charney, MD, PhD, Associate Professor of Psychiatry, Genetics and Genomic Sciences, Neuroscience, and Neurosurgery, and Director of The Charles Bronfman Institute of Personalized Medicine at the Icahn School of Medicine at Mount Sinai.

Click here to watch an episode of Mount Sinai’s Road to Resilience podcast titled “The Long Battle With Long COVID.”

Of the 37 symptoms studied, 12 were identified that set those with long COVID apart: post-exertional malaise, fatigue, brain fog, dizziness, gastrointestinal issues, heart palpitations, loss of sexual desire or performance, loss of smell or taste, thirst, chronic cough, chest pain, and abnormal movements. Based on those symptoms, the investigators found that 23 percent of participants with a prior COVID-19 infection crossed the study threshold for long COVID.

“This is the first step of a series of studies that will be published using RECOVER data that will provide critical insights about the incidence, risk factors and pathophysiology of long COVID. This information will be vital for managing the large number of patients afflicted by this emerging condition,” says Juan Wisnivesky, MD, DrPH, Professor of Medicine at Icahn Mount Sinai, a clinical epidemiologist who is one of the primary leads of the RECOVER adult cohort study at the Mount Sinai site.  Dr. Wisnivesky is Chief of the Division of General Internal Medicine for the Mount Sinai Health System.

“Our overarching goal for RECOVER is to continue refining the definition of long COVID and to understand the biological causes of the condition. As a major hub for RECOVER enrollment and given the richly diverse communities that we serve—along with our grit—we are uniquely positioned to do just that.” — Alexander Charney, MD, PhD

The results were based on a survey in which 9,764 patients self-reported their symptoms. Next, data from the survey will be compared against lab and imaging results to validate the current findings.

“Until we establish a unifying framework to define it, it’s like we’re flying a plane without navigation,” Dr. Charney says when describing recent developments to come out of the Long COVID RECOVER multi-center study on which Mount Sinai is a co-author. “Defining long COVID in different ways will similarly get us varied answers. Only by creating a clear and uniform definition of long COVID can we accurately diagnose and begin the road to effective treatments. This study brings us closer to that goal of uniformity.”

On the hotly debated vaccination question—in another RECOVER study development—co-author Girish Nadkarni, MD, MPH, noted a preliminary signal in the data showing that being vaccinated was associated with a lower risk of long COVID. Dr. Nadkarni is the Irene and Dr. Arthur M. Fishberg Professor of Medicine, Director of The Charles Bronfman Institute of Personalized Medicine, and Chief, Division of Data Driven and Digital Medicine (D3M), Department of Medicine at Icahn Mount Sinai.

Mount Sinai colleagues in the Department of Psychiatry are leveraging RECOVER biospecimens to gain deeper insights into the cognitive and behavioral manifestations seen in long COVID.

Led by Scott Russo, PhD, Professor of Neuroscience and Director of the Center for Affective Neuroscience and Brain Body Research Center at Icahn Mount Sinai, the research will explore immune-mediated mechanisms underlying the emergence of depression and anxiety observed in this cohort. The work could inform the identification of biomarkers and enhance understanding of psychiatric symptoms associated with the disorder.

Dr. Russo and his team will also use RECOVER MRI data to measure changes in the penetrability of the blood-brain barrier associated with immune alterations and depression in long COVID.

Additional RECOVER brain initiatives, led by a team including Dr. Wisnivesky, Alex D. Federman, MD, MPH, and Jacqueline H. Becker, PhD, will conduct one of the first randomized clinical trials to investigate a potential therapeutic intervention for brain fog, a condition that affects many long COVID patients. Dr. Federman is Professor of Medicine and Director of Research for the Division of General Internal Medicine. Dr. Becker is a neuroscientist and Assistant Professor of Medicine.

Notably, Judith Aberg, MD, FIDA, FACP, the George Baehr Professor of Medicine, Dean of System Operations for Clinical Sciences, and Chief of Infectious Diseases, is a member of the RECOVER Pathobiology and Interventions Task Force. The task force recently published a white paper emphasizing the need for therapeutic interventions for individuals with long COVID. As the Principal Investigator, Dr. Aberg oversees the master contract for upcoming interventional trials sponsored by RECOVER.

“One of the exciting aspects of RECOVER is the collaboration of multi-disciplinary teams across Mount Sinai Health System engaged in interventional trials to improve the quality of life of those with long COVID,” says Dr. Aberg.

Sean Liu, MD, PhD, Medical Director of the COVID-19 Clinical Trials Unit, will lead a study evaluating the use of antivirals to treat long COVID based on the hypothesis that SARS-CoV-2 may persist in certain parts of the body and that a prolonged antiviral course may eliminate these hidden virus reservoirs.

Sarah Humphreys, MD, Assistant Professor of Medicine, will be the local Principal Investigator at Mount Sinai’s Center for Post-COVID Care. Other upcoming trials expected include interventions to improve cognitive function (the RECOVER-NEURO trial) and to address sleep disturbances (the RECOVER-SLEEP trial).

Mariana G. Figueiro, PhD, Professor at the Department of Population Health Science and Policy and Director of the Light and Health Research Center, is one of the Co-Principal Investigators nationally for RECOVER-SLEEP. The upcoming clinical trial will investigate the impact of light, alone and in combination with melatonin, on sleep in long COVID. The Center will serve as the core hub for this research protocol.

“Our overarching goal for RECOVER is to continue refining the definition of long COVID and to understand the biological causes of the condition.  As a major hub for RECOVER enrollment and given the richly diverse communities that we serve—along with our grit—we are uniquely positioned to do just that,” says Dr. Charney.

Read more about Research at Mount Sinai

AI Spotlight: Guiding Heart Disease Diagnosis Through Transformer Models

Updated on Nov 19, 2025 | Artificial Intelligence, Featured, Research

Akhil Vaid, MD, left, and Girish Nadkarni, MD, MPH, right, are working to make artificial intelligence models more feasible for reading electrocardiograms, using a novel transformer neural network approach.

Electrocardiograms (ECGs) are often used by health providers to diagnose heart disease. At times, irregularities in the recordings are too subtle to be detected by human eyes but can be identified by artificial intelligence (AI).

However, most AI models for ECG analysis use a particular deep learning method called convolutional neural networks (CNNs). CNNs require large training datasets to make diagnoses, which spell limitations when it comes to rare heart diseases that do not have a wealth of data.

Researchers at the Icahn School of Medicine at Mount Sinai have developed an AI model, called HeartBEiT, for ECG analysis, which works by interpreting ECGs as language.

The model uses a transformer-based neural network, a class of network that is unlike conventional networks but does serve as a basis for popular generative language models, such as ChatGPT.

Here’s how HeartBEiT works as an artificial intelligence deep-learning model, and how it compares to CNNs.

HeartBEiT outperformed conventional approaches in terms of diagnostic accuracy, especially at lower sample sizes. Study findings were published in npj Digital Medicine on June 6. Akhil Vaid, MD, Instructor of Data-Driven and Digital Medicine, was lead author, and Girish Nadkarni, MD, MPH, Irene and Dr. Arthur Fishberg Professor of Medicine, was senior author.

In this Q&A, Dr. Vaid discusses the impact of this new AI model on reading ECGs.

What was the motivation for your study?

Deep learning as applied to ECGs has had much success, but most deep learning studies for ECGs use convolutional neural networks, which have limitations.

Recently, the transformer class of models has assumed a position of importance. These models function by establishing relationships between parts of the data they see. Generative transformer models such as the popular ChatGPT utilize this understanding to generate plain-language text.

By using another generative image model, HeartBEiT creates representations of the ECG that may be considered “words,” and the whole ECG may be considered a single “document.” HeartBEiT understands the relationship between these words within the context of the document, and uses this understanding to perform diagnostic tasks better.

What are the implications?

Our model forms a universal starting point for any ECG-based study. When comparing our model to popular CNN architectures on diagnostic tasks, HeartBEiT ended up with equivalent performance and better explanations for the model’s thinking and choices using as little as a tenth of the data required by other approaches.

Additionally, HeartBEiT generates very specific explanations of which parts of an ECG were most responsible for pushing a model towards making a diagnosis.

What are the limitations of the study?

Pre-training the model takes a fair amount of time. However, fine-tuning it for a specific diagnosis is a very quick process that can be accomplished in a few minutes.

HeartBEiT was compared against other conventional AI methods on diagnostic measures, including left ventricular ejection fraction ≤40%, hypertrophic myopathy, and ST-elevation myocardial infarction, and was found to perform better.
How might these findings be put to use?

Deployment of this model and its derivatives into clinical practice can greatly enhance the manner in which clinicians interact with ECGs. We are no longer limited to models for commonly seen conditions, since the paradigm can be extended to nearly any pathology.

What is your plan for following up on this study?

We intend to scale up the model so that it can capture even more detail. We also intend to validate this approach externally, in places outside Mount Sinai.

Learn more about how Mount Sinai researchers and clinicians are leveraging machine learning to improve patient lives

AI Spotlight: Forecasting ICU Patient States for Improved Outcomes

When Can a Patient Come Off a Ventilator? This AI Can Help Decide

Delivering the Future of Vaccines With mRNA Technology

Updated on Jun 15, 2023 | Featured, Research, School, Vaccines

From left to right, Peter Palese, PhD, Horace W. Goldsmith Professor of Medicine; Miriam Merad, MD, PhD, Mount Sinai Professor in Cancer Immunology; Özlem Türeci, MD, Chief Medical Officer of BioNTech; Uğur Şahin, MD, Chief Executive Officer of BioNTech; Dennis Charney, MD, Anne and Joel Ehrenkranz Dean of Icahn School of Medicine at Mount Sinai

One of the great tools that helped turn the tide of the COVID-19 pandemic was the use of vaccines, which prevented millions of deaths and hospitalizations in the U.S. and around the world. Key vaccines were those based on messenger RNA (mRNA) technology, which provide information for the molecules that teach the cells in the body to generate proteins used by viruses or cancers, allowing the body’s immune system to recognize and fight off future infections or transformed cancer cells.

The Icahn School of Medicine at Mount Sinai honored the efforts of executives of German biotechnology firm BioNTech, which partnered with Pfizer to develop and make available one of the most widely used COVID-19 vaccines in the country, during its 54th Commencement on Thursday, May 11. Uğur Şahin, MD, Chief Executive Officer of BioNTech, and Özlem Türeci, MD, its Chief Medical Officer, received honorary Doctor of Science degrees.

Research into mRNA technology for vaccines goes back to the 1990s, and has grown in leaps and bounds since, said Dr. Türeci in a guest lecture hosted by the Marc and Jennifer Lipschultz Precision Immunology Institute, held separately from the Commencement.

The COVID-19 pandemic provided an opportunity for the technology to be adapted at a large scale, and the momentum gained and lessons learned was only the starting point to pave the way for greater heights for the development of mRNA vaccines, she said.

In this Q&A, Drs. Şahin and Türeci spoke about what the future of mRNA vaccines could look like.

After two years of COVID-19 vaccines:

  • An estimated 18 million hospitalizations were prevented
  • More than 3 million deaths were avoided
    Source: New York City-based foundation The Commonwealth Fund

Percentage vaccinated in United States by manufacturer:

  • Pfizer/BioNTech: 60%
  • Moderna: 37%
  • Johnson & Johnson: 3%
    Source: Centers for Disease Control and Prevention

What are some active areas of research in which mRNA technology is being worked on?

Dr. Şahin: There are investigational cancer vaccines in which mRNA technology is being used to deliver instructions to generate antibodies or cytokines. This technology can theoretically be used to deliver any bioactive molecule.

Our focus at the moment is the development of cancer vaccines, and one special application of cancer vaccines we’re working on is the so-called “personalized cancer vaccines.” mRNA technology is particularly well suited to deliver a vaccine that consists of mutations of the tumor identified from the patient.

Dr. Türeci presenting to members of the Marc and Jennifer Lipschultz Precision Immunology Institute.

What is it about mRNA technology that makes it so well suited for cancer vaccines?

Dr. Türeci: We have been interested in cancer vaccines all along, and tried different technologies, and mRNA is the delivery technology that comes with its own edge. Its immunogenicity is very versatile and its transience has the potential to lead to a favorable safety profile. These characteristics are the reasons why we chose mRNA to deliver cancer antigens.

Any solid cancer could be appropriate for application. We have ongoing clinical trials in melanoma, head-and-neck cancer, pancreatic cancer, and non-small cell lung cancer.

Beyond cancer vaccines, we believe any bioactive cancer immunotherapy that is based on protein could be delivered by mRNA.

What about non-cancer diseases? Is mRNA technology suitable there?

Dr. Türeci: There are other areas, such as infectious diseases, in which mRNA could have an advantage. As long as you have the right protein structure to stimulate an immune response, you can theoretically also use mRNA here.

There are clinical trials in infectious diseases: COVID-19, for example, but also malaria or shingles.

What are some current limitations of mRNA technology? And how are researchers working to overcome those?

Dr. Türeci: We are very far advanced in the delivery component of the technology, and these advancements have made COVID-19 vaccines, as well as cancer vaccines in clinical testing, feasible. However, if you want to target specific organs, you need specialized, targeted delivery technologies.

For example, if you want to address something in the brain, you need a delivery technology that brings the mRNA into the brain. There may be monogenetic diseases in which the sample protein is deficient in the organ, and so limits how the mRNA can be expressed there.

So the lipid nanoparticle used to contain the COVID-19 vaccine, for example, might not be applicable for any other organs?

Dr. Türeci: This delivery technology was specifically designed and developed to deliver mRNA to the lymphatic system. If the mRNA needs to be delivered to different organs, it required new formulation.

What is a lipid nanoparticle? Read more about it here

When the public first became aware of mRNA technology through COVID-19 vaccines, there was skepticism. Do you envision similar skepticism as new mRNA vaccines roll out, and if so, how can we dispel such skepticism?

Dr. Türeci: Skepticism can only be addressed by transparent communication, through the disclosure of data, and proper education. I think there is a zeitgeist of skepticism. That skepticism isn’t necessarily specific to mRNA technology. But once they start to understand the mechanisms behind the technology, and the rationale of why we’re working on it, we can start to dispel it.

Do you foresee mRNA technology to grow exponentially into the future?

Dr. Şahin: Yes, mRNA vaccines could be really big, but it will happen slowly. It will take a few more years, but we are starting to see really promising candidates using this technology.

Read more about Research at Mount Sinai

AI Spotlight: Forecasting ICU Patient States for Improved Outcomes

Updated on Nov 19, 2025 | Artificial Intelligence, Featured, Research

AI Spotlight: Forecasting ICU Patient States for Improved Outcomes

Girish Nadkarni, MD, MPH, and Faris Gulamali

Artificial intelligence (AI) and machine learning (ML) have seen increasing use in health care, from guiding clinicians in diagnosis to helping them decide the best course of treatment. However, AI still has much unrealized potential in various health care settings.

Mount Sinai researchers are exploring bringing AI into intensive care, and developed Spatial Resolved Temporal Networks (SpaRTeN), a model to assess high-frequency patient data and generate representations of their state in real time.

The work was presented at the Time Series Representation Learning for Health workshop on Friday, May 5, hosted by the International Conference for Learned Representations, a premier gathering dedicated to machine learning.

Hear from Girish Nadkarni, MD, MPH, Irene and Dr. Arthur Fishberg Professor of Medicine at the Icahn School of Medicine at Mount Sinai and the leader of the SpaRTeN research, and Faris Gulamali, medical student at Icahn Mount Sinai and member of the Augmented Intelligence in Medicine and Science lab, on what lay behind creating the model and what it could achieve for patients.

What was the motivation for your study?

A growing amount of research is indicating the need to redefine critical illness by biological state rather than a non-specific illness syndrome. Advances in genomics, data science, and machine learning have generated evidence of different underlying etiologies for common ICU syndromes. As a result, patients with the exact same diagnosis can have entirely different outcomes.

What are the implications?

In the ICU, representations of a patient can be used to guide personalized treatments based on personalized diagnoses rather than generic treatments with empirical diagnoses.

What are the limitations of the study?

In this study, we only looked at using one type of data at a time in real time. For example, we looked primarily at measures of intracranial pressure. However, the ICU has many types of data being output simultaneously. Future work hopes to integrate all the different types of data such as electrocardiograms, blood pressure, and imaging to improve patient representations.

How might these findings be put to use?

These patient representations are being combined with data on medications and procedures to determine how to optimize patient treatment based on underlying state rather than common illness syndromes.

What is your plan for following up on this study?

In this study, we focused primarily on creating the algorithm and showing that it works for the case of intracranial hypertension. In future studies, we would like to integrate multiple data modalities such as imaging, electrocardiograms, and blood pressure as well as intervention-based data such as medications and procedures to determine precise empirical interventions that lead to improvements in short-term and long-term patient outcomes.

Learn more about how Mount Sinai researchers and clinicians are leveraging machine learning to improve patient lives

Computational Neuroscientist Opens Doors for New Ideas and Talent to Thrive

When Can a Patient Come Off a Ventilator? This AI Can Help Decide