Skip to content

Impossible they said

Gordicknot

The diagnostic process worldwide consists of semantic descriptions of disease and various standards for coding symptoms and disease. However, these approaches have significant limitations:

Coding Systems Limitations

Coding systems like ICPC, ICD, SNOMED, and UMLS are used to describe disease, primarily for administrative purposes. Few, if any, doctors use them for clinical work. Instead of evolving for clinical use, these systems have become increasingly ambiguous and unusable.

Semantic Description Challenges

Semantic descriptions of disease vary depending on the source. Few descriptions of disease include the same symptom descriptions. Some descriptions are unintelligible even for doctors, who must make assumptions about implied meaning or alternatively trust their own experience.

The Ambiguity Problem

To use these descriptions of disease effectively, one must be a doctor. The level of ambiguity is much too high for other people, let alone computers. This ambiguity is itself a major hurdle for effective treatment of disease.

Our Solution: Empirica

We rewrite the documentation for symptoms and disease to remove the ambiguity. This is a vast undertaking requiring clinical expertise, semantic understanding, data analysis, and programming expertise. Empirica:

  • Enables patients and doctors to arrive at diagnosis and treatment with higher quality, faster
  • Helps doctors using Empirica for diagnosis to typically arrive at more solid diagnosis while using a fraction of the time, compared to traditional methods
  • Saves patients typically hours for each incident, while still being able to provide more precise information to the doctor

Technical Implementation

Platform

Empirica runs in Microsoft Azure.

Core Components

The system consists of a database with structured information about symptoms and diagnoses and their relationships.

Artificial Intelligence Integration

We have spent thousands of hours using AI to understand potential and limitations, make analysis and code, retrofit data and as expert systems. We work with OpenAI, Anthropic, Groq, Azure AI and more.

Survey Systems

We have built several generations of survey systems in Empirica:

  • Gen 1: Survey systems integrated as part of patient dialog (Typeform and others)
  • Gen 2: Survey system created as part of the patient dialog, without use of an external provider
  • Gen 3: Survey system entirely in the database
  • Gen 4: Integrated dialog system driven by symptoms and diagnosis

Health Vocabularies and Standards

We've extensively studied ICPC, ICPC2, ICD10, ICD11, SNOMED, UMLS, and developed systems for automatically diagnosing according to these standards.

Symptoms and Diagnosis Collection

We gather and analyze information from authoritative medical sources including Mayo Clinic, Medline, WebMD, Healthline, Medscape, and research sites.

Language Processing

We've invested significant resources in investigating how medical language can be broken down to data, making it accessible and actionable.

Multi-language Support

Empirica is inherently in English language and translated to other languages. It is currently available in English, Norwegian, and Hungarian. We plan to make Empirica available in all major languages, initially in the Western hemisphere, including Italian, German, Spanish, Polish, French, and Dutch.

The benefits of Empirica are immediately visible for doctors and patients within the specialties and countries where the service is available.