Quantitation of Magnetic Resonance Image Texture
PARMA WORKING GROUP - 12th to 14th February, 1999
Application of Magnetic Resonance Texture Analysis
to the Food Industry
R E P O R T
Present: Dr R A Lerski, Dundee, UK
Prof A Spisni, Parma, IT
Prof R N Muller, Mons, BE
Dr H Stødkilde-Jørgensen, Aarhus, DK
Ass Prof A Santos, Madrid, SP
Dr M Hajek, Prague, CZ
Dr A Karlsson, Tjele, DK
Dr F Rodler, Aarhus, DK
Dr J Bezy-Wendling, Rennes, FR
Dr G Collewet, Rennes, FR
Mrs A Barclay, Dundee, UK
Friday 12th February, 1999
A Spisni opened the meeting by welcoming everyone to Parma. R Lerski then went on to explain, for the purpose of first-time participants, that this EC COST project started in May 1998 and would carry on until 2002, with, hopefully, the possibility of continuing thereafter.
R Lerski also stated that the purpose of the meeting was to discuss applications of NMR and MRI in the study of food, and also to present and discuss the techniques of MRI texture analysis. He stated that the following questions, in particular, would be addressed:
The following presentations were given:
1) R A Lerski - Previous Work in MRI Texture Analysis
A full description was given of the meaning of MRI texture analysis, the results so far achieved in reticulated foam test objects and the future direction of research planned within COST B11 for medical applications.
2) A Spisni - Introduction to NMR Applications in Food
A review was presented of the wide range of MRI applications with respect to (Italian) food.
Flour : Images of dried grain were shown - germination can be defined in this way, i.e. the difference can be seen between wet/dry areas.
Pasta : The process of drying pasta is very important. Images of wheat pasta were shown - the signal intensity detected is simply water.
Parmesan Cheese : What can be done to help in this area? Is it possible to follow the ageing of parmesan through images (T2 weighted). Some images were then shown, and the difference between a one month and a 21 month cheese was evident - there was a process of diffusion of fatty acids and water. T1 weighted is different from T2 weighted. Therefore, can help with the assessment of the ageing process. The present method of quality control is with the use of a small hammer to judge the acoustic resonance (with an expert ear).
Parma Ham : Images were shown and it was found that a large part of the meat could be discriminated. The major problems were to know how much fat was in the ham, and the ageing process (has to be aged properly or very bad results are obtained).
In all cases it was felt that the image analysis could be improved by texture analysis, but this had yet to be demonstrated.
Saturday 13th February, 1999
The following presentations were given:
1) R N Muller - Analysis of post-harvest changes in apples by Magnetic Resonance and Relaxometry
Regarding the aim of the studies, R Muller said that apples were probably not the best fruit to study as there is not much difference between fresh and 3 month old apples. However, a series of images were shown indicating the changes that can be observed.
Results were shown from magnetisation transfer - good parameters to use.
2) MR methods in the evaluation of meat quality. Can parameters from texture analysis be indicative for the first 12 hours post-slaughtering changes in meat from the pig?
The use of magnetic resonance and imaging in meat quality was discussed. In two cases, NMR phosphorous spectroscopy was used, viz.
1) Non de-bleeded pig in-situ - slaughtered in MRI machine.
2) Debleeded pig - in-vivo - slaughtered first.
In the first case the ATP peak was maintained over 13 hours, but in the second it disappeared very quickly.
Image results from 3 pigs were shown. Simple texture analysis (wavelet transform) had been applied, but the MAZDA program had not yet been used since the images available could not be read. The results were encouraging.
The two main chemical components of porcine skeletal muscle and pork are water and protein. The water content is approximately 75%. This water is held by the protein fraction, which is 20% of the tissue. This means that a small change in the ability of the protein fraction to bind water has a huge effect on the water-holding capacity of the meat. Therefore, a very important quality trait for pig meat is its water-holding capacity. Water-holding is measured gravimetrically as drip loss on a standardised piece of meat placed in a standardised environment during a standardised period of time. The drawback of this method is that it is invasive and destructive as the meat has to be cut. Today there are no possibilities to estimate or predict the final water-holding capacity of the meat non-invasively or non-destructively. As changes in the muscle fibres and fascia cannot be detected by the human eye, we are therefore interested in investigating the possibilities of using texture and wavelet analyses of NMR images produced during the post mortem process.
There are two parts of interest:
a) Can texture analysis be used to estimate the water-holding in meat at a specific time point post slaughter (1-3 days)?
b) Is it possible, at an early stage (say within 6 hours) after slaughter, to predict the water-holding of the meat at 1-3 days post-slaughter by studying the time course non-invasively by NMR?
3) M Hajek - What we can expect from MR Quality Control of food products
- QC of food projects
- Most important parameters
- NMR methodology available in food analysis
4) M Hajek - Fruits and vegetable models suitable for texture analysis. The comparison with human tissue.
A wide range of images of different fruits and vegetables were shown. Several examples indicated texture properties very similar to human tissue or to the reticulated foam test objects now used.
5) G Collewet - Cheese analysis with MRI
48 soft cheeses of 6 different types were analysed by a MRI system (0.2T Open Siemens). 5 types of sequence were used : 3 spin echo (proton density, T2 and T1 weighted), 1 gradient echo (magnetic susceptibility weighted) and 1 multi-echo (24) for T2 measurements. Several characteristics were extracted from the images : T2 histograms, openings features (surface, perimeter, compacity, distance to the next neighbour), parameters from texture analysis using co-occurrence matrix and grey-level histograms. For each of these three families, a PCA was computed. Using the PCA results, a FDA was computed too. The texture analysis parameters were found to have the best results.
6) J Bezy-Wendling - Skeletal muscle MRI texture analysis
Concluded that automatic texture analysis in muscular dystrophies gave better results than visual analysis by a radiologist.
J Bezy-Wendling also commented that J de Certaines of Rennes was hoping to have a project called BIOSPIN up and running in two years time (budget of 8M EUROS). This project would set up a French network of groups interested in the use of NMR in food applications. She also mentioned that the report from this meeting could be considered for publication in the Magnetic Resonance Analysis Journal.
1) What are the questions in Food Analysis?
- Process Control: Product at beginning of process
3) NMR METHODS
The relationship between the measurements possible is shown in the diagram:
4) Texture Analysis
- Partners - any other suggestions to Dundee?
- Industrial partners - any other suggestions to Dundee?
- Food experts to be found (search for more people who have
The diagram summarises the overall aim of an ongoing project.
Dr R A Lerski
Chairman COST B11
3 March, 1999