RESEARCH

What are you interested in?

All topics
AI and Society
Causal Inference
Neuroscience
NLP
Vision
Urban Planning and Smart Cities
Neuroscience and Psychology
AI and Education
Health-Biomedicine
Fundamentals of AI and DS
Environment
Economy and Finance
Digital Humanities
AI, Ethics & Law
AI in the Physical Sciences
AI for social good
Real time Augmented Reality Registration of Stretchable Structures for Needle Navigation through the Gluteus Muscles

Research

Feb 24th, 2022
Real time Augmented Reality Registration of Stretchable Structures for Needle

Researchers: Dr. Dan Raviv, (Electrical Engineering), Shai Tejman-Yarden, M.D., Edmond and Lily Safra Children's Hospital, Chaim Sheba Medical Center & Sackler Faculty of Medicine

  • Vision
  • Health-Biomedicine
  • AI for social good

Virtual Reality (VR) and Augmented Reality (AR) are two popular platforms to enhance human vision by providing additional layers of information not visible to the naked eye. The gaming industry is blooming using this technology, and the tech giants have recently manufactured high accuracy VR/AR glasses.

 

In this proposal, we claim that we can use the same hardware but add improved AI algorithms of alignment between stretchable domains and project pre-scanned CT images on top of the soft tissues. The current alignment done in a virtual scenario is not sufficient for surgical procedures, especially where there is a movement of the skin and fat tissues between the CT scans and the 3D positioning of the body during the medical procedure. While brain surgery navigation is accurate enough for computer assistance, it is built for hard tissues (bones) and requires heavy preparation per patient. Here we focus on a quick solution for soft tissues with minimal resource requirements.

 

Using state-of-the-art rigid and non-rigid geometric deep learning alignment algorithms developed recently in Dr. Raviv’s lab and the medical expertise from Dr. Tejman-Yarden’s lab, we believe we can solve a fundamental medical task. Using superior algorithms, we plan to provide a robust real-time data fusion tool of soft tissues for surgical needs. We plan to focus first on needle insertion through the Gluteus Muscles, but the same approach can be used for multiple organs and multiple treatments.

 

 

Research

Aug 11th, 2021
The phylogenetic tree reconstruction game: developing reinforcement-learning

Researchers: Tal Pupko (Life Sciences), Yishay Mansour (Computer Science), Itay Mayrose (Life Sciences) 

  • Health-Biomedicine
  • Fundamentals of AI and DS
Causal Effects of Antibiotic Use on Resistance. Photo by Nastya Dulhiier on Unsplash

Research

Jul 25th, 2021
Causal Effects of Antibiotic Use on Resistance

Researchers: Dr. Daniel Nevo (Statistics), Dr. Uri Obolski (Public Health)

  • Health-Biomedicine

Antibiotic resistance of bacterial infections is a major public health threat. When a patient is presented with a bacterial infection, there is a built-in uncertainty in a doctor’s decision of prescribing antibiotics before laboratory confirmed results are available.

 

A key piece of information is missing in this decision-making process - a quantitative estimate of the causal effect of antibiotics on future antibiotic resistance of bacterial infections.

 

We apply advanced causal inference and machine learning methods to a highly detailed dataset of electronic medical records of over 40,000 Israeli hospitalized patients with bacterial infections. 

 

Antibiotic resistance frequencies are dynamic and constantly changing due to the rapid evolutionary response of bacteria to environmental and individual-level changes.

 

Our first step towards understanding this time-varying system is to identify predictors with constant effect on antibiotic resistance across time, despite the dynamic nature of the system. These predictors are themselves causal, or are surrogates of stable causal relationships.

 
 
Prior hospitalization days - for DS center
 
 
 
Admission to ER - for DS center
 
 
​Time-varying associations of the predictors along with 95% confidence intervals. Local estimation of time-varying coefficient models is used to identify predictors with time-fixed effects. The left panel shows a constant association (up to noise) while the right panel shows substantial changes in the association through time.  

 

Research

Jul 12th, 2021
Dynamic risk prediction (DRP) model for post trauma psychopathology

Researchers:  Talma Hendler (Psychology & Medicine), Malka Gorfine (Statistics)

 

  • Health-Biomedicine

 

Hippocrates, the father of medicine, put forth the importance of prognosis over diagnosis stating that; “there is no such thing as a disease; there are individuals who fall ill”. Modern medicine, however, has made diagnosis the venerable element of medical practice. Mental disorders lack indefinite anatomical and or pathogenic indicators, and therefore could benefit from an evidence-based prognostic approach. Such models built on neurobiological and psychological measurements will allow reliably predicting the risk of future illness, likelihood of remission vs chronicity as well as treatment response in an individual patient (i.e. personalized manner). Our work aims to develop a prediction model for developing psychopathology after an exposure to a traumatic stressor. 

 

Stress is ever-present in our lives, significantly impacting the onset and aggravation of mental and physical health and resulting in an economic cost of ~ 300€ billion annually. Critically, even a single life-threatening stressful event can trigger an onset of a debilitating mental disorder like Post-Traumatic Stress Disorder (PTSD), the most common trauma-related psychopathology. However, the trajectories of response to a potentially traumatic event vary immensely between individuals, ranging from full remission to lifelong debilitating disorder. We assume that the trajectory of traumatic stress response could be depicted by repeated measurements at multi–domains, enabling us to capture the nature and dynamics of the response which might result in post-traumatic psychopathology. It is well established that knowing the clinical trajectory early on is essential for efficient prevention and/or treatment of PTSD as well as for unveiling the underlying mechanisms of the disorder.

 

The main goal of this project is to apply advanced statistical and machine learning methods for developing Dynamic Risk Prediction (DRP) models for individual post-traumatic psychopathology. We believe that such a model could serve for precise and personalized monitoring, prevention and treatment of PTSD. The computation of the DRP model will be based on three unique independent datasets recently obtained at PI Hendler’s lab, covering post-traumatic psychopathology development from one-month post-trauma and up to many years later on (i.e., acute and chronic PTSD populations). 

 

Research

Jul 12th, 2021
Artificial Intelligence and Machine learning models for assessing physiological

Researchers: Yftach Gepner (Public health), Noam Ben-Eliezer (Biomedical Engineering), Hayit Greenspan (Biomedical Engineering)

 

  • Health-Biomedicine

 

Physical activity is one of the strongest beneficial behaviors for human health decreasing the risk for hypertension, diabetes and cardiovascular pathologies, and for improvement of physical function and mental health. In addition, exercise training increases whole body metabolism, alleviates inflammatory response, and particularly, can improve muscle microstructure which is strongly associated with muscle strength and physical capacity. However, the quality of evidence on the effect of exercise training across different age groups, on muscle microstructural state, and its association with blood markers concentration remains unclear. An extensive amount of data is typically collected in clinical setting. This consists of mainly blood markers for cardiometabolic risk and imaging data, and might allow to identify variety of medical states and response patterns. Yet, the amounts of collected data are so high that it remains mostly unused and underutilized. This underutilization is indeed detrimental in clinical setting, and contributes to diagnostic bias and inter-observer variability, which influence patients’ healthcare.

 

In our study we develop a new artificial intelligence (AI) and machine learning (ML) platform that will be able to quantify and predict muscular response and recovery curves following physical training. Using human trial, recreationally active young and middle-aged males completed downhill running protocol, while undergoing MRI scans, blood tests for cardiometabolic and inflammation markers and physical performance testes. Overall, we collected a total of ~33,600 images, including tracking muscle fibers, quantifying MRI relaxation times to assess inflammatory processes, edema or temporary swelling, and quantifying fat content. This data is processed using an AI/ML model to determine subjects’ response following the exercise. These neural networks process a mixture of all collected data and mainly new quantitative MRI biomarkers that can be accurately tracked over time, compared between subjects, and generalizable to other studies. The ensuing platform will use to design personalized training programs, but also will have the potential to improve the diagnosis and prognosis of patients suffering from pathologies that affect muscle state including heart failure, diabetes, and general muscle dystrophies and neuromuscular diseases.

 

Figure 1: In our research we establish a new AI platform that is built on a deep learning neural network in order to (1) process large amounts of data including ca. 33,600 MRI images; (2) integrate data from three different sources: MRI scans, numeric blood markers, and exercise performance tests; and (3) learn and predict post-training muscle recovery curves.

 

The ensuing framework will help in identifying general muscle response patterns and the tailoring of personalized training programs – a critical aspect for both professional and un-professional athletes, and for patients suffering from pathologies that affect muscle state such as heart failure, diabetes, and general muscle dystrophies and neuromuscular diseases.