A5013659557- Radiomics and Machine Learning in Medical Imaging
- Cancer, Lipids, and Metabolism
- AI in cancer detection
- Metabolomics and Mass Spectrometry Studies
- Machine Learning in Healthcare
- Biomedical Text Mining and Ontologies
- Prostate Cancer Treatment and Research
- Law and Political Science
- Philosophy, Science, and History
- Probability and Statistical Research
- Artificial Intelligence in Healthcare and Education
- Genomics and Chromatin Dynamics
- Philosophical Ethics and Theory
- Cancer Genomics and Diagnostics
- Machine Learning and Data Classification
- Religion, Theology, and Education
- Free Will and Agency
- Tactile and Sensory Interactions
- Epistemology, Ethics, and Metaphysics
- Philosophy and History of Science
- Explainable Artificial Intelligence (XAI)
- Public Administration and Political Analysis
- German Social Sciences and History
- Gene expression and cancer classification
- German legal, social, and political studies
Cornell University
2021-2025
Tri-Institutional PhD Program in Chemical Biology
2024-2025
Weill Cornell Medicine
2021-2023
Harvard University
2019-2023
Dana-Farber Cancer Institute
2020-2023
University of Konstanz
2016-2023
University of Bonn
2003-2021
NeuroDevelopment Center
2019
Massachusetts General Hospital
2019
Oberlin College
2018
Abstract Medical artificial intelligence (AI) has tremendous potential to advance healthcare by supporting and contributing the evidence-based practice of medicine, personalizing patient treatment, reducing costs, improving both provider experience. Unlocking this requires systematic, quantitative evaluation performance medical AI models on large-scale, heterogeneous data capturing diverse populations. Here, meet need, we introduce MedPerf, an open platform for benchmarking in domain....
Machine learning has become an increasingly ubiquitous technology, as big data continues to inform and influence everyday life decision-making. Currently, in medicine healthcare, well most other industries, the two prevalent machine paradigms are supervised transfer learning. Both practices rely on large-scale, manually annotated datasets train complex models. However, requirement of be labeled leaves excess unused, unlabeled available both public private repositories. Self-supervised (SSL)...
Summary Genetic influences on psychiatric disorders transcend diagnostic boundaries, suggesting substantial pleiotropy of contributing loci. However, the nature and mechanisms these pleiotropic effects remain unclear. We performed a meta-analysis 232,964 cases 494,162 controls from genome-wide studies anorexia nervosa, attention-deficit/hyperactivity disorder, autism spectrum bipolar major depression, obsessive-compulsive schizophrenia, Tourette syndrome. correlation analyses revealed...
Abstract Imaging datasets in cancer research are growing exponentially both quantity and information density. These massive may enable derivation of insights for clinical care, but only if researchers equipped with the tools to leverage advanced computational analysis approaches such as machine learning artificial intelligence. In this work, we highlight three themes guide development tools: scalability, standardization, ease use. We then apply these principles develop PathML, a...
Interpretability for machine learning models in medical imaging (MLMI) is an important direction of research. However, there a general sense murkiness what interpretability means. Why does the need MLMI arise? What goals one actually seek to address when needed? To answer these questions, we identify formalize and elements MLMI. By reasoning about real-world tasks common both image analysis its intersection with learning, five core interpretability: localization, visual recognizability,...
Clear cell renal carcinoma (ccRCC) is molecularly heterogeneous, immune infiltrated, and selectively sensitive to checkpoint inhibition (ICI). However, the joint tumor-immune states that mediate ICI response remain elusive. We develop spatially aware deep-learning models of tumor features learn representations ccRCC tumors using diagnostic whole-slide images (WSIs) in untreated treated contexts (n = 1,102 patients). identify patterns grade heterogeneity WSIs not achievable through human...
Gleason score, a measure of prostate tumor differentiation, is the strongest predictor lethal cancer at time diagnosis. Metabolomic profiling and patient serum could identify biomarkers aggressive disease lead to development less-invasive assay perform active surveillance monitoring. tissue samples was performed. Metabolite levels metabolite sets were compared across scores. Machine learning algorithms trained tuned predict transformation or differentiation status from data. A total 135...
Abstract Holistic review has been widely adopted in medical education as a means of promoting equity the application process and diversity workforce. Artificial intelligence (AI) is rapidly emerging technology already having an impact on school residency students faculty alike increasingly turn to AI tools automate some steps preparation evaluation materials. While may have potential improve holistic admissions by increasing efficiency adding measure standardization among reviewers, authors...
Recently, much research in the area of haptic technologies has focused on development waist-worn vibrotactile belts as substitution or augmentation modalities for audio-visual information. Vibrotactile have been used varied applications, such navigational aids, spatial orientation display, and balance control. Researchers mostly functionality these specific applications while neglecting performance usability. Considering versatility a belt, we previously conducted study design requirements...
Machine learning has become an increasingly ubiquitous technology, as big data continues to inform and influence everyday life decision-making. Currently in healthcare, well most other industries, the two prevalent machine paradigms are supervised transfer learning. Both practices rely on large-scale, manually annotated datasets train complex models. However, requirement of be labeled leaves excess unused, unlabeled available both public private repositories. Self-supervised (SSL) is a...
Clear cell renal carcinoma (ccRCC) is molecularly heterogeneous, immune infiltrated, and selectively sensitive to checkpoint inhibition (ICI). Established histopathology paradigms like nuclear grade have baseline prognostic relevance for ccRCC, although whether existing or novel histologic features encode additional heterogeneous biological clinical states in ccRCC uncertain. Here, we developed spatially aware deep learning models of tumor- immune-related learn representations tumors using...
A human centered pragmatic approach to the design and implementation of a vibrotactile belt is presented in this paper. Based on (a) extensive usability feedback we've collected over past year, (b) thorough survey existing guidelines from literature, we propose set for development haptic belts that can span seamlessly across various applications. These cover three important aspects belts: functionality, performance usability, which are vital longitudinal use by end users. Taking...
Genes that are inherently subject to strong selective constraints tend be overretained in duplicate after polyploidy. They also continue experience similar, but somewhat relaxed, polyploidy event. We sought assess for how long the influence of is felt on these genes' pressures. analyzed two nested events Brassicaceae: At-α genome duplication most recent model plant Arabidopsis thaliana and a more hexaploidy shared by genus Brassica its relatives. By comparing strength direction natural...
Medical AI has tremendous potential to advance healthcare by supporting the evidence-based practice of medicine, personalizing patient treatment, reducing costs, and improving provider experience. We argue that unlocking this requires a systematic way measure performance medical models on large-scale heterogeneous data. To meet need, we are building MedPerf, an open framework for benchmarking machine learning in domain. MedPerf will enable federated evaluation which securely distributed...
Group instruction is the most common delivery method of motor skill training given its cost and time effectiveness. This also case during rehabilitation where therapists divide their attention among several patients. Compared to dedicated one-on-one instruction, group often suffers from reduced quality quantity feedback. Further, programs, patients struggle outside therapy sessions lack feedback found only clinic visits. We propose a wearable, low-cost motion sensing actuation system capable...