NCRI demonstrator radiology surgery pathology - PowerPoint PPT Presentation

1 / 19
About This Presentation
Title:

NCRI demonstrator radiology surgery pathology

Description:

NCRI demonstrator radiology surgery pathology – PowerPoint PPT presentation

Number of Views:27
Avg rating:3.0/5.0
Slides: 20
Provided by: cancerinfo8
Category:

less

Transcript and Presenter's Notes

Title: NCRI demonstrator radiology surgery pathology


1
NCRI demonstrator radiology surgerypathology
2
MRI images
Resected specimen
Microscopic slices
Macro slices
  • Objectives of NCRI informatics program
  • technical development of a common informatics
    platform
  • facilitation cultural change, so that the
    platform can be understood and used
  • implementation by the UK cancer research
    community for greater impact of research, better
    cost effectiveness, and generation of new
    knowledge
  • We have the opportunity to implement this
    demonstrator in new planned clinical trials

3
Progress digital slide database
  • Slides collected from 11 international centres
  • 68 Phase II Core rectal cancer cases (1500
    slides) digitised

4
Core - surgery on all cases
5
(No Transcript)
6
Radiology open
7
CRM analysis
MRI pre-resection
histology
Since the measurement is taken in plane (!!), it
depends on the imaging plane. It should be 3D
Is there sufficient circumferential resection
margin?
We have developed a system for re-sampling the
geometry, to compensate for the imaging angle.
On 10 cases, difference between 2D and 3D CRM
assessment is 2.1mm, and 3D is more closely in
accordance with pathology assessment.
Petroudi, Brown, Bond, Brady, CRM assessment ,
IEEE EMBC, 2006
8
Reconstruction From Macro Slices
Resected specimen
9
Registration is Needed
  • Animation of slices before registration

10
Location of colorectum
Slice 3, 4, 5 (3mm thick)
11
Alignment of successive slices
  • The Grey contour is the floating contour in its
    original position.
  • The Red contour is the reference contour
  • The white one is in the resultant position.

White
12
Enhanced segmentation
Human-assisted segmentation
Automatic segmentation
  • Segmentation is complex and may give incorrect
    results
  • Registration is simpler but relies on correct
    segmentation
  • There is currently still a requirement for human
    intervention for best segmentation results

13
3-D reconstruction
  • A stack of images can be animated and
    interrogated
  • Volume can be meshed for modelling
  • Data can be interpolated between slices
  • Multi-modal (MR) data can be merged

14
Security
  • Certificate based security
  • User - Server and Server - Server mutual
    authentication
  • Developing automated methods of performing the
    complex initial setup to minimise impact on user
  • Extending standard web services
  • Modifying default security handler
  • Discovered bug allowing security to be bypassed
  • Discovered bug preventing authentication under
    certain circumstances

15
Data
  • Schema for structured storage of DICOM Data
    refined from the eDiaMoND schema
  • Generate DICOM style study, series and instance
    ID's for non-DICOM image data
  • Data load process allowing simple references
    between images of different types to be entered

16
Data Federation
  • Obtaining data from several sources and combine
    to answer user query
  • Registering servers with each other to allow
    cross server querying of data sources
  • Allow query to be specified simply in an XML
    document
  • Easily understandable
  • Provision of parameterised predefined queries

17
Challenges
  • Accessing grid nodes that are located behind
    firewalls particularly when the nodes are located
    inside the NHS
  • Provision of an understandable flexible
    federation system
  • Provision of a fine grained access control system
    with an associated administration tool
  • Developing database schemas to allow the
    representation of complex relationships between
    differing data sources.

18
Demonstration
  • Secure data access retrieval
  • Subsequent analysis (segmentation and
    registration) of histology images

19
Taking the work forward
  • Image analysis
  • Non-rigid registration of macro slices, 3D
    reconstruction
  • Non-rigid registration of macro to MRI volume (no
    cost to NCRI)
  • Integration with lymph detection/analysis (no
    cost to NCRI)
  • Informatics infrastructure
  • Extend federation service resilience
  • Extend schema to support complex queries over
    multiple modalities
  • Fine-grained access control (no cost to NCRI)
  • Develop production-level implementation
  • Testing in clinical trials
  • Aristotle
  • Advanced rectal cancer trial
  • RT/5FU vs RT/5FU/Oxaliplatin /- biological
  • Needs accurate MRI staging pre and post Rx,
    quality of surgery and pathological response
  • Low rectal cancer study
  • Difficult surgery, frequent use of RT/CT and new
    cylindrical operation
  • Neo-adjuvant chemotherapy in colon cancer study
  • Staging, quality of surgery and response to
    therapy needed
Write a Comment
User Comments (0)
About PowerShow.com