Tabular Data
Tabular Data
Tabular data—organized in rows and columns—forms a cornerstone of biomedical datasets. Each row typically represents a distinct sample (e.g., a patient visit or biological measurement), while each column corresponds to a specific feature or attribute associated with that sample.
Electronic Health Records (EHRs)
Electronic Health Records (EHRs) encompass clinical records and patient demographics, forming a vital source of information within healthcare. These datasets contain a wide range of details, including:
Age, sex, and other demographics
Medical history and diagnoses
Prescribed treatments
Laboratory test results
Procedural records and clinical notes
Data Support
Our platform provides a secure and privacy-preserving marketplace for EHR data, enabling encrypted encoding, sharing, and collaborative analysis across institutions.
EHR data is typically stored in:
CSV or TSV files
Relational databases
Standardized EHR platforms
Equivariant Encryption (EE) ensures that all data remains encrypted during processing. This allows for collaborative use without ever compromising patient privacy.
Model Support
Foundation models applied to EHRs unlock powerful capabilities such as:
Predicting disease onset or progression
Stratifying patients by risk
Modeling treatment outcomes across populations
Our platform enables secure fine-tuning and evaluation of models on encrypted EHRs using EE.
MOTOR: A foundation model for time-to-event prediction in healthcare, trained on over 55 million patient records.
Clinical Trial Data
Clinical trial data is a specialized form of tabular data collected during the evaluation of drugs, devices, and therapeutic protocols. Each record may include:
Patient demographics and medical history
Treatment assignment (e.g., dosage, duration)
Measured outcomes, efficacy endpoints, and adverse events
Data Support
Clinical trial data is often stored in:
CSV and Excel spreadsheets
Structured clinical databases
Our platform offers a secure and auditable system for managing and analyzing clinical trial data. By applying EE, we ensure encrypted storage, sharing, and model training—while maintaining trial integrity and patient confidentiality.
EE also enables secure coordination across multi-site trials.
Model Support
Foundation models can be used to:
Predict trial outcomes
Stratify patients for personalized trial arms
Optimize trial design and summarization
However, risks like hallucination and overfitting require careful validation—especially when operating on sensitive datasets.
Rexis enables distributed evaluation and fine-tuning of these models using encrypted clinical trial data.
Panacea: A foundation model for clinical trial summarization, recruitment, and design. It uses transformer-based architectures to extract key information and improve relevance matching across large-scale trial data.
Last updated