Skill UI

This utility separates clinical entity extraction from clinical coding. It processes de-identified text, utilizing OpenMed to identify key clinical spans. These spans are then programmatically mapped into structured FHIR R4 resources (like Condition, Observation, and MedicationStatement) and assembled into a Bundle. This process ensures deterministic, standardized data output suitable for interoperability and medical record structuring.

Utilizes OpenMed's analyze_text to perform sophisticated Named-Entity Recognition (NER) on clinical and biomedical text. This tool extracts structured entities such as diseases, drugs, anatomy, genes, and proteins. It provides precise character spans and model confidence scores, making it essential for building robust, privacy-first clinical Natural Language Processing (NLP) pipelines.

This skill extracts critical patient metadata (PHI) from DICOM headers using pydicom, and recursively pulls structured report narrative text from DICOM-SR content trees. It is essential for ingesting medical images into NLP pipelines, enabling subsequent de-identification and clinical text analysis.

Use openmed.extract_pii to detect and retrieve specific types of Protected Health Information (PHI) or Personally Identifiable Information (PII) spans (names, dates, MRNs, etc.) from clinical text. Unlike deidentification, this function returns the original text and the detected spans, allowing users to audit, preview redactions, normalize labels, or route spans to a custom anonymizer without altering the source data. It is ideal for pre-processing and policy enforcement.

This skill extracts crucial Social Determinants of Health (SDOH) — such as housing instability, food insecurity, and transportation barriers — from unstructured clinical notes. It maps the identified text spans to standardized ICD-10-CM Z-codes (Z55–Z65). It is essential for health-equity analytics, building risk stratification tools, and identifying coding gaps by converting narrative text into queryable, discrete data points.

This skill facilitates the ingestion of unstructured clinical notes and reports from FHIR R4 EHR APIs (e.g., Epic, Cerner). It handles complex tasks including paginating large search results, decoding base64 attachments (PDF, text), and extracting narrative content. The extracted text is then passed to the OpenMed pipeline for HIPAA-compliant de-identification and subsequent NLP analysis, enabling structured data insights from raw clinical records.

Synthea generates statistically realistic, fully synthetic patient records in multiple standards (FHIR R4, C-CDA, CSV). This tool is essential for developers and researchers needing safe, shareable test data for OpenMed pipelines, CI fixtures, or specialized de-identification leakage gate test sets. It ensures zero risk of exposing real Protected Health Information (PHI).

This skill replaces detected Protected Health Information (PHI) in clinical notes with realistic, type-matched synthetic surrogates. Unlike simple redaction, surrogates maintain the natural readability and structure of the text, ensuring the data remains parseable for downstream NLP tasks, QA, or human review. It supports consistent, reproducible fake identities and allows custom identifier generation for proprietary formats.

This module converts diverse clinical inputs—including scanned images, faxes, CSV/TSV exports, and C-CDA XML—into a normalized, clean text format (`ExtractedDocument`). It supports on-device OCR (Tesseract/PaddleOCR) and structured redaction for tables. It is the crucial first stage for feeding de-identification and Named Entity Recognition (NER) pipelines.

A comprehensive, systematic workflow for reverse engineering client-side JavaScript. It guides users through five stages: Observation, Capture, Rebuilding, Patching, and Deep Dive. Applicable for locating API signatures, analyzing encryption parameters, tracking runtime function calls, and reconstructing complex web environments locally for evidence collection and reproduction. Utilizes advanced tools like jshookmcp and various js-reverse_* utilities.

This skill manages the loading and efficient caching of OpenMed clinical/biomedical Named Entity Recognition (NER) models. Models are downloaded once from the Hugging Face Hub, enabling full offline operation afterward. It supports model reuse across multiple calls, configuring cache directories, specifying devices (CPU/GPU), and handling various model identifiers (registry keys, full HF IDs, or local paths).

This skill maps free-text laboratory and clinical observation names to standardized LOINC codes. It resolves ambiguous terms by considering the full six-axis model: Component, Property, Time, System (specimen), Scale, and Method. It integrates with FHIR terminology services to ensure accuracy, supporting US Core Laboratory Results and UCUM units.