A5021730842- Machine Learning in Healthcare
- Sepsis Diagnosis and Treatment
- Time Series Analysis and Forecasting
- Topological and Geometric Data Analysis
- Cell Image Analysis Techniques
- Artificial Intelligence in Healthcare and Education
- Topic Modeling
- Natural Language Processing Techniques
- Electronic Health Records Systems
- Clinical Reasoning and Diagnostic Skills
- Multimodal Machine Learning Applications
- Radiomics and Machine Learning in Medical Imaging
- COVID-19 diagnosis using AI
- Advanced Graph Neural Networks
- Mobile Health and mHealth Applications
- Phonocardiography and Auscultation Techniques
- Congenital Heart Disease Studies
- Gaussian Processes and Bayesian Inference
- Ultrasound in Clinical Applications
- Infective Endocarditis Diagnosis and Management
- AI in cancer detection
- Image Retrieval and Classification Techniques
- Single-cell and spatial transcriptomics
- Autopsy Techniques and Outcomes
- Trauma and Emergency Care Studies
ETH Zurich
2018-2025
Stanford University
2023-2025
SIB Swiss Institute of Bioinformatics
2018-2023
Stanford Medicine
2023
Palo Alto University
2023
Board of the Swiss Federal Institutes of Technology
2021
Red Cross War Memorial Children's Hospital
1985-1991
University of Cape Town
1985-1989
Care during the COVID-19 pandemic hinges upon existence of fast, safe, and highly sensitive diagnostic tools. Considering significant practical advantages lung ultrasound (LUS) over other imaging techniques, but difficulties for doctors in pattern recognition, we aim to leverage machine learning toward guiding diagnosis from LUS. We release largest publicly available LUS dataset consisting 202 videos four classes (COVID-19, bacterial pneumonia, non-COVID-19 viral pneumonia healthy controls)....
BackgroundLarge language models (LLMs) have recently shown impressive zero-shot capabilities, whereby they can use auxiliary data, without the availability of task-specific training examples, to complete a variety natural tasks, such as summarization, dialogue generation, and question answering. However, despite many promising applications LLMs in clinical medicine, adoption these has been limited by their tendency generate incorrect sometimes even harmful statements. MethodsWe tasked panel...
When sepsis is detected, organ damage may have progressed to irreversible stages, leading poor prognosis. The use of machine learning for predicting early has shown promise, however international validations are missing.This was a retrospective, observational, multi-centre cohort study. We developed and externally validated deep system the prediction in intensive care unit (ICU). Our analysis represents first international, in-ICU study using our knowledge. dataset contains 136,478 unique...
Large-language models have recently demonstrated impressive zero-shot capabilities in a variety of natural language tasks such as summarization, dialogue generation, and question-answering. Despite many promising applications clinical medicine, adoption these real-world settings has been largely limited by their tendency to generate incorrect sometimes even toxic statements. In this study, we develop Almanac, large model framework augmented with retrieval for medical guideline treatment...
Medicine, by its nature, is a multifaceted domain that requires the synthesis of information across various modalities. Medical generative vision-language models (VLMs) make first step in this direction and promise many exciting clinical applications. However, existing typically have to be fine-tuned on sizeable down-stream datasets, which poses significant limitation as medical applications data scarce, necessitating are capable learning from few examples real-time. Here we propose...
We propose a novel approach for preserving topological structures of the input space in latent representations autoencoders. Using persistent homology, technique from data analysis, we calculate signatures both and to derive loss term. Under weak theoretical assumptions, construct this differentiable manner, such that encoding learns retain multi-scale connectivity information. show our is theoretically well-founded it exhibits favourable on synthetic manifold as well real-world image sets,...
Abstract Recently emerging large multimodal models (LMMs) utilize various types of data modalities, including text and visual inputs to generate outputs. The incorporation LMMs into clinical medicine presents unique challenges, accuracy, reliability, relevance. Here, we explore applications GPT-4V, an LMM that has been proposed for use in medicine, gastroenterology, radiology, dermatology, United States Medical Licensing Examination (USMLE) test questions. We used standardized robust...
While many approaches to make neural networks more fathomable have been proposed, they are restricted interrogating the network with input data. Measures for characterizing and monitoring structural properties, however, not developed. In this work, we propose persistence, a complexity measure architectures based on topological data analysis weighted stratified graphs. To demonstrate usefulness of our approach, show that persistence reflects best practices developed in deep learning community...
Sepsis is a life-threatening host response to infection associated with high mortality, morbidity, and health costs. Its management highly time-sensitive since each hour of delayed treatment increases mortality due irreversible organ damage. Meanwhile, despite decades clinical research, robust biomarkers for sepsis are missing. Therefore, detecting early by utilizing the affluence high-resolution intensive care records has become challenging machine learning problem. Recent advances in deep...
Clinicians spend large amounts of time on clinical documentation, and inefficiencies impact quality care increase clinician burnout. Despite the promise electronic medical records (EMR), transition from paper-based has been negatively associated with wellness, in part due to poor user experience, increased burden alert fatigue. In this study, we present Almanac Copilot, an autonomous agent capable assisting clinicians EMR-specific tasks such as information retrieval order placement. On...
Despite the eminent successes of deep neural networks, many architectures are often hard to transfer irregularly-sampled and asynchronous time series that commonly occur in real-world datasets, especially healthcare applications. This paper proposes a novel approach for classifying with unaligned measurements, focusing on high scalability data efficiency. Our method SeFT (Set Functions Time Series) is based recent advances differentiable set function learning, extremely parallelizable...
Sepsis is a leading cause of death and disability in children globally, accounting for ∼3 million childhood deaths per year. In pediatric sepsis patients, the multiple organ dysfunction syndrome (MODS) considered significant risk factor adverse clinical outcomes characterized by high mortality morbidity intensive care unit. The recent rapidly growing availability electronic health records (EHRs) has allowed researchers to vastly develop data-driven approaches like machine learning healthcare...
Most modern intensive care units record the physiological and vital signs of patients. These data can be used to extract signatures, commonly known as biomarkers, that help physicians understand biological complexity many syndromes. However, most biomarkers suffer from either poor predictive performance or weak explanatory power. Recent developments in time series classification focus on discovering shapelets, i.e. subsequences are terms class membership. Shapelets have advantage combining a...
Graph neural networks (GNNs) are a powerful architecture for tackling graph learning tasks, yet have been shown to be oblivious eminent substructures such as cycles. We present TOGL, novel layer that incorporates global topological information of using persistent homology. TOGL can easily integrated into any type GNN and is strictly more expressive (in terms the Weisfeiler--Lehman isomorphism test) than message-passing GNNs. Augmenting GNNs with leads improved predictive performance node...
Benjamin Yan, Ruochen Liu, David Kuo, Subathra Adithan, Eduardo Reis, Stephen Kwak, Vasantha Venugopal, Chloe O’Connell, Agustina Saenz, Pranav Rajpurkar, Michael Moor. Findings of the Association for Computational Linguistics: EMNLP 2023.