COST B11

Quantitation of Magnetic Resonance Image Texture

GLASGOW WORKING GROUP - 20th - 21st November, 1998

Progress with Texture Methods and Applications in MRI

R E P O R T

Download MS-Word document: rbl3626.doc


Present:
 
Dr R A Lerski, Dundee  Dr Y Rolland, Rennes 
Prof Dr L R Schad, Heidelberg  Dr M Bock, Heidelberg 
Prof Dr A Materka, Lódz Dr M Strzelecki, Lódz 
Prof A Spisni, Parma  Dr B Michie, Dundee 
Prof J Chambron, Strasbourg  Ms S Allein, Brussels 
Dr H Stødkilde-Jørgensen, Aarhus Ass Prof A Santos, Madrid 
Dr A Bharath, London  Dr J F Goudemant, Mons 
Mr H Langenberger, Vienna  Dr M Hajek, Prague 
Ass Prof A Lundervold, Bergen  Mrs A Barclay, Dundee 
 

 Friday 20th November, 1998

1) Welcome and Introduction

The proceedings commenced with R Lerski (Chairman, Cost B11) welcoming everyone to Glasgow for the second Working Group meeting of COST Action B11, and explaining the purpose of the meeting.   2) Presentation by Dr B Michie - Spatial scale of tissue textures A full and detailed description was given of the visual appearances and types of pathologies that are commonly seen in clinical practice, and that might be of interest to the project. There existed a wide range, from macroscopic to those that need microscopic examination to diagnose. Typical resolution in MRI corresponded to the lowest magnification normally used by pathologists.   3) Comments by Dr Y Rolland Dr Rolland pointed out that clinical diagnoses were arrived at by assessing all the information available. Texture analysis would be a part of the decision process and we should find methodology to add all the information together and find the overall correlation.
 
4) Presentation by Dr R A Lerski - Textures of test materials including foams 5) Presentation by Prof A Materka - Statistical features and modelling of MR foam images 6) Presentation by Prof J Chambron - Distinct patterns of MRI active and non-active MS plaques by using texture analysis Professor Chambron described studies to investigate the use of texture analysis in MS. Gadolinium contrast MRI was used as the ?gold-standard? and comparison made with texture analysis for classification. With a limited sample, good accuracy of classification was obtained.   7) Presentation by A Lundervold - Functional MRI using texture analysis Preliminary studies had been made of the value of texture analysis in functional brain measurements.   8) Presentation by A Bharath - Software - Use of MATLAB A full suite of texture analysis software had been developed at Imperial College, London, in MATLAB. This was well documented and tested.   9) Presentation by L Schad - Report on NMRWin changes In collaboration with the Lødz group, checks had been made of the NMRWin software and comparison made with MAZDA. Reticulated foam test images had been examined.  

10) Presentation by A Santos - Analysis of texture parameter data for classification

Texture parameter data collected in Dundee had been extensively analysed with respect to its ability to classify foams, liver, brain and plain gel. Generally the classification was successful except for some failures in the separation of the finest foams and of brain and liver tissue.

General Comments:

  Q. R Lerski asked B Michie if it was easy to distinguish between liver and brain pathology.

A. Yes (although this is not a question commonly asked) - resolution may be the key.

Y Rolland commented that you can?t just use texture in isolation and that micro- and macro-texture is most important.

 
11) A Spisni - Use of MRI in the food industry (to discuss for the 5th Framework):   Professor Spisni suggested the possibility of setting up a small Working Group meantime, with a view to putting forward an application for the 5th Framework which, it was thought, may have a good chance of being financed if industry was included. Mons are also starting a project of this nature using phantoms at present.   Phase I      - Selection of products
                    - Definition of criteria to evaluate quality
                    - Definition of criteria for shelf-life
                    - Necessary parameters
                    - Definition of measurement protocols
Phase II     - Acquisition of set of images in various conditions
                    - Comparison
                    - Integration of parameters
Phase III    - Validation of analytical method
                     - Development of analysis
                            i.e. Development of equipment hardware
                     - Commercialisation
 
The parties interested in this Working Group are as follows:
Prof A Spisni - Italy
Prof R Muller/Dr J-F Goudemant - Mons
Prof A Santos - Madrid
Dr Y Rolland - Rennes
Dr H Stødkilde-Jørgensen - Aarhus
Dr M Hajek - Prague
Dr R Lerski - Dundee
 

Saturday 21st November, 1998
 

CONCLUSIONS AND PLANNING SESSION
 

1) Test Objects - DUNDEE

R Lerski to check with A Mandenoff if a subsidy could be applied for to have these tubes made.
 
2) Test Object Trials - HEIDELBERG AND DUNDEE
 
50mm tubes to be used

Heidelberg  -->  4 foams
                            Multislice (5 slices)
                            256 and 512 pixels
                            FOV = 100 mm  -->  250 mm
                            SL = 2  -->  5 mm
                            Gradient Echo/Spin Echo
                            NSA = 1 to 64

Relaxation Time Maps should be calculated as well

T1 ~ 200 ms (gel) - Dysprosium gel should be considered

Analysis could be done in Dundee

Data distribution - CD-ROM/FTP

Aim:        SNR Dependence of texture parameters
                Resolution dependence

Timescale:

3) Software Requirements  
Routine Development
   
MAZDA (Lodz) MATLAB
QC - Rennes/Madrid 

STSM - Rennes/Lodz

(ROI Definition could be a problem) 

(C Code NMRWin --> IC, London + Bergen) 

  • IC to standardise implementation
  • Multi-license (RAL to investigate)
  TXT Processing 
  • Excel
 

4) Meetings 1999

 

5) Short Term Missions

 

6) Web-site

 

7) Normalisation

  
COST B11 Homepage