SNOMED CT Implementation Approaches and Use Cases

Part 1. 충북대학교 성수미

Contents

• SNOMED CT Concept model & Description Logic
• Implementation Approaches
• Implemenation Use Cases
• Query and Expression Constraint Language

Concept model / Description logic

purpose represent clinical information retrieve and anlyze health information share and integrate data

Concept + Description + Relationship

• Concept: clinical idea with a unique identifier (enable meaning based queries)
  • Code <-> Concept <-> Clnical idea
• Description Fully specified name (FSN) + synonyms
• Relationships: association

  • is a	relationship: subtype “polyhierarchy”

    • Hierarchy is formed from

      is a

      relationships

  • attribute relationship: represent the meaing of the association betwe4en the source and destination concepts
    • characteristics of the meaning of a concept (e.g. finding site)

Concept model

Domain and Range Rule: Domain (top level hierarchy) -> Attribute -> Range ex) clinical finding attributes: finding site, associated morphology, associated with, due to, etc…

SNOMED CT expression

• structured combination of one or more concept identifiers used to represent a clinical idea in a logical manner, which is automatically processable
• precoordinated expression + postcoordinated expression
  • Right hip joint - example

    • precoordinated: 362906008

      Right hip joint

    • post-: 182201002

      Hip joint

      : 272741003

      Laterality

      = 24028007

      Right

• creating postcordinated expressions: refinement + qualification
  • refinement: refine the value of one or more of the defining attributes of the concept
  • qualification: applying permitted values to the attributes. attributes applied need not be present in the definition of the concept that is being qualified.
• compositional grammar

Implementation Approaches

• is not all or nothing!
  • simple code system or a powerful terminological resource
  • simpler uses have benefits
  • complete use delivers even more benefits
  • stepwise approach is possible
    • a code system to store clinical information
    • interface terminology
    • indexing system
    • common terminology
    • dictionary
    • extensible foundation

ex 1. Common terminology for communication (with external system)

• health records mapped to SNOMED CT for exchange
• no changes required to core clinical system ex 2. Common terminology for data integration
• used to integrate multiple data sources, which use different code systems or free text
• provides a consistent representation for combined use ex 3. Indexing system for retrieval (patient health records -> reporting and analysis OR data extract to DW)
• records may be stored using text or other code systems
• records indexed using SNOMED CT ex 4. code system for clinical data storage
• SNOMED CT codes stored within clinical data ex 5. interface terminology for data capture
• terms shown to user come directly from SNOMED CT
• uses local dialect preferred and acceptable synonyms ex 6. Dictionary for analytics, Dictionary for knowledge base, extensible foundation for clinical data
• uses SNOMED CT’s hierarchies & logic definitions
• support querying and aggregation over healt records
• may use …

Terminology binding

Implementation Use Cases

• Data collection / capture
• Constraining searches
• Rationalizing display of search results
• Data analysis with SNOMED CT
  • enables analysis by any hierarchy and subsumption
  • supports analysis based on defining relationships
  • supports meaningful health records analysis (meaning based data extraction)
• Customization: Reference Set
  • different reference set types suitable to meet specific requirements for use

Query and Expression Constraint Language

• Executing expression constraints (expression constraint operatiors!)
  • Expression Constraint Language (ECL)

SNOMED CT browser Practice

Part 2. 박현애 Health O&T

Introduction to Snowstorm

Introduction

• snowstorm
  • SNOMED international open-source terminology server
  • https://github.com/IHTSDO/snowstorm

컴퓨터 비전 인공지능의 이해와 실습! SAM을 활용하는 자동 레이블링!

김영곤, 김민경

컴퓨터 비전 인공지능의 이해와 실습 (김영곤/서울대학교병원)

• 핵심내용
• part 1 이론 및 실습
  • 디지털 영상의 기초
  • 기계학습 기초 이해
  • 딥러닝 원리 이해

디지털 영상의 기초 이해

• 컴퓨터 비전
  • classification, detection, segmentation, style transfer, super resolution
  • 응용 예: self-driving (with radar) / CAD for healthcare / Inventory management / Body camera / Crop and livestock monitoring / Warehouse management
• 디지털 영상의 기초
  • digital image processing
    • image processing, sampling/quantization, resizing, image enhancement, spatial domain filtering 해당 개념소개와 실습진행
  • Pixel: 디지털 이미지를 구성하는 최소단위
  • Pixel level: Add, Inverse, Threshold

  • Transformation: Flip, Resize, Translate, Rotate, Shear

  • Filtering
    • convolution (2D convolution)
• 기계학습 기초 이해
  • Artificial Neural Networks (ANNs) & Perceptron
  • multi layer perceptron (MLP), backpropagation algorithm
  • ReLU
  • Keras
• 파이썬 기본 라이브러리