imaging informatics IV

  1. PACS
    • picture archiving and communication system
    • medical imaging technology which provides economical storage and convenient access to images from multiple modalities
    • usually integrated with other healthcare information systems through database gateways to communicate and manage patient information related to imaging examinations
  2. universal format for PACS image storage is...
  3. major components of PACS
    • image and data acquisition gateway accepting images from various modalities
    • PACS server - secure network for transmission of patient information
    • archive for management and storage
    • display workstations for reviewing and interpreting images
  4. RIS
    • radiology information system
    • core system for the electronic management of imaging departments
    • tool that most technologists and radiology administrators use for their daily activities
    • complements HIS (hopstial information systems) and PACS
  5. RIS uses ____ as its core communication protocol
  6. major functions of RIS
    • patient scheduling
    • resource management
    • examination performance tracking
    • reporting
    • results distribution
    • procedure billing
  7. speech recognition
    • Speech recognition software to replace medical transcriptionist
    • Advantages in transcription costs and speed of report turnaround
    • May result in decreased radiologist efficiency
  8. DICOM
    • digital imaging and communication in medicine
    • language of imaging equipment
    • defines, formats, transmits, and manipulates information objects
    • both an image file format and a network protocol for moving images around
    • standardized syntax for radiology
  9. DICOM file format
    • header + image data
    • stored in same file
    • hundreds of pieces of metadata stored as free text
  10. HL7
    • health level seven
    • set of international standards for transfer of clinical and administrative data between various healthcare providers
  11. IHE
    • group formed to address issues of interoperability and information sharing in medical imaging
    • IHE Radiology formed in 1998
    • formed in response to lack of common language between PACS, RIS, HIS
    • PACS based on DICOM, while HIS/RIS rely on HL7
  12. AIM
    • annotation and image markup
    • standardized syntax for radiology
    • emerging XML-based standard for image metadata
    • image annoations + image markups = AIM annotation
    • uses terms from ontologies (RadLex)
  13. RDF
    • resource description framework
    • standard for data on Semantic Web
    • an XML language for representing predicate graphs on the world wide web
    • allows computer processing on the web
  14. MRRT
    • management of radiology report templates
    • developed by IHE
    • describes imaging report templates used by radiologists to guide the creation of a clinical diagnostic imaging report
    • describes methods to format imaging report templates and to manage their transport between template libraries and report creation systems
  15. RadLex
    • controlled terminology (ontology) for radiology research and practice
    • main purpose to improve communication and clinical imaging research, allow uniform indexing and retrieval of radiology information resources
  16. NLP uses in imaging
    • extracting data from radiology reports
    • querying literature
    • extracting data from DICOM image headers
  17. pre-processing in NLP (6)
    • tokenization: break up input text into tokens
    • part of speech (POS) tagging: assign POS tags to tokens (nouns, adjective and relations (predicates))
    • stemming: normalize words to their stem (root) form - useful to map input text to standard terms (ontology), improves signal
    • stop word removal: discard words that are non-informative
    • chunking: process of recognizing sequences of words that are entities in their own right
    • named entity recognition (NER): identify types of entities mentioned in text
    • sentence segmentation: break text into sentences, use rule-based methods to recognize sentence boundaries
  18. RadLex features
    • adopts existing concepts from widely accepted standards (SNOMED-CT) and fills in gaps where radiology terms are absent
    • provides mapping with related, existing terms in lexicons such as CPT, ACR Index, UMLS, allowing versatility and practical utility
    • freely availble
  19. RadLex problems
    • stored in text files and spreadsheet
    • as it grows, the hierarchy becomes difficult to browse, manage, and modify
    • lexicon is cumbersome for curation
  20. types of information extraction systems
    • rule-based (most-common)
    • co-occurrence-based
    • statistical-based
  21. information extraction: rule-based
    • most common
    • con incorporate many types of information
    • usually developed manually - rules from domain experts
    • output is usually a set of semantic annotations on text
  22. information extraction: co-occurence-based
    • two co-occurring entities are related
    • uses very little information from sentences
  23. information extraction: statistical-based
    • require training data
    • may not generalize
  24. word embeddings
    • put words into a reduced dimensionality, informative vector space
    • can discover semantic annotations from texts without need to supply domain knowlede
  25. word embedding approaches
    • Word2vec
    • Recursive neural network (RNN)
    • Convolutional neural network (CNN)
Card Set
imaging informatics IV
imaging informatics