MDClone is introducing Breach Observation Analyzer (BOA). BOA streamlines health data extraction, aggregation and analysis using any cloud-based public resource (LLMs, Gen AI, etc) without jeopardizing organization security and patient privacy.
Does your organization have a plan to leverage the growing number of LLMs and generative AI technology without compromising on privacy and security?
Each day, analytics and AI solutions are developing to become far more capable than ever before of producing insights that can transform the way we deliver healthcare. Simultaneously, healthcare organizations must grapple with growing challenges to protect both patient and organizational privacy.
How will your organization leverage the growing number of LLMs and generative AI technology without compromising on privacy and security?
How can your organization collect, aggregate and analyze data sourced from external healthcare institutions?
How can you partner with a healthcare organization to develop, test or implement your AI technology onpatient data?
The BOA product line addresses the rapidly expanding disparity between evolving AI technologies and the data requirements for retrospective medical research and healthcare data analysis. This discrepancy largely arises from patient privacy and general system security regulations governing patient data in contrast to the open, cloud-based architecture of emerging AI and machine learning technologies.
With BOA, healthcare organizations and data aggregators an end-to-end tool kit to streamline health data extraction, aggregation and analysis using any cloud-based public resource, including public LLMs and generative AI technology, without jeopardizing organization security and patient privacy.
BOA Privacy Content Firewall enables exploration of health data, including PHI, using public SaaS platforms equipped with generative AI and ML technologies.
BOA Secure Data Hub enables extraction and aggregation
of health data from multiple sources using a private
and safe channel for analysis in a central environment.
The underlying technology for MDClone BOA products combines existing MDClone synthetic data and federated learning capabilities with a unique pre-trained LLM, BOA (Breach Observation Analyzer), that detects and removes any private or sensitive data before exposing it to cloud-based open-source or proprietary analytical SaaS solutions.
MDClone’s ADAMS Platform is a big data platform that organizes patient-oriented data in a longitudinal-based model designed for clinicians and other medical or administrative staff to extract and analyze cohorts of interest. The platform includes solutions to harmonize semantics and ontologies and utilizes an internal NLP engine to extract structured data from unstructured documents.
SDE produces synthetic data on a cohort-by-cohort basis, built from data that can be extracted from a health system via the MDClone ADAMS Ask module (where previously installed) or from a pre-built tidy data set sourced from existing systems. SDL produces entire synthetic data lakes, and can include all features for entire populations of interest as a time series, including all medications administered, all labs recorded, etc.
MDClone FLS represents a specific application of federated learning in healthcare, focusing on building predictive models from retrospective patient-centric medical data. The system architecture typically involves local model training ensuring sensitive data remains within the institution, aggregation and model updates from a central hub and iterative learning allowing the entire model
development process to repeat iteratively, with each cycle improving the global model’s accuracy and generalizability.
Allows analysts in a central environment to conduct analyses on data aggregated from two or more health system sites. The RunTime Tool enables a process whereby a user can determine a single project query within a single ADAMS Platform in a central hub, built based on a planned analysis, which is then compiled into a RunTime executable agent that can run from an end-user Windows 11 machine / MacOS / Linux at an individual health system or similar organization site. The agent enables both synthetic data extraction and federated weight extraction from sites. The extracted data can then be uploaded to a central hub where the data can be unified for analysis.
BOA utilizes open-source LLM (Based on Llama 3) trained on synthetic and real data to detect possible
patient privacy breaches in both structured and unstructured data and produce non-PHI data that
preserves utility for research.
Discover the power of dynamic data exploration at your fingertips and unlock the true potential of your healthcare data.
