2.gif){kind=small}
friday 1999
Medical Image Processing, Visualisation and Analysis
St. Edmund Hall, Oxford 12th to 16th July 1999
Friday 16th
1. Image Processing in Magnetic Resonance, presented by Professor Jo Hajnal (Hammersmith Hospital). This was an extremely interesting and highly relevant lecture which complemented the earlier presentation by Professor Mark Henkelman. The two closing remarks in this lecture were that:
- it was important to have an appreciation of the properties of MR acquisition since this leads to the productive use of specially tailored methods, particularly for image production and extraction of specific physical parameters
- image processing methods are increasingly being integrated with image acquisition systems in order to produce methods that reduce artefacts, speed up acquisition, or increase data content or throughput.
2. Medical Image Processing in Nuclear Medicine, presented by Professor Andrew Todd-Pokropek (University College London). The aim of this presentation was to provide an understanding of the special nature of medical image processing in the area of nuclear medicine applications. It covered the following topics:
- Acquisition of data in both SPECT and PET systems
- Tomographic reconstruction methods
- Processing and the extraction of physiological data
- Multi-modalities
- Medical Images networks. This was an excellent PowerPoint presentation, which was accompanied by a comprehensive set of notes.
3. Mammographic Image Analysis: Analysis of Breast Cancer Images Based on the Physics of Image Formation, presented by Professor Mike Brady (Oxford University). The final presentation of the summer school was given by Professor Mike Brady, who gave an interesting and detailed overview of the problems and challenges faced in the analysis of mammograms and contrast enhanced breast MRI, and the current image analysis techniques employed. The first half of the presentation covered x-ray mammography and included the following sections:
Clinical Challenges This section gave a number of sad statistics and emphasised the need to improve diagnosis of mammograms since:
- 25% of cancers are missed by the radiologists
- 70-80% of biopsies turn out to be benign
- inter- and intra-radiologist variability is typically 30%.
Computing Challenge The types of problems encountered when performing automatic analysis of Mammograms:
- Image smoothing can remove micro calcifications
- Smoothing a mass changes the appearance of benign tissue to malignant
- Sharpening removes scatter but malignant tissue then appears to be benign.
Diagnosing Breast Disease Using X-ray Mammography
An outline was given of the types of structures which radiologists look for on normal and abnormal mammograms. This highlighted the range of variability of normal mammograms, due to age differences, previous surgery, and effects of HRT.
Having given the problems associated with the interpretation of Mammograms, Professor Brady then went on to outline his approach to image enhancement and identifying micro calcifications, using the h int representation. This is a normalised mammographic representation depicting the thickness of interesting (non- fat) tissue in the compressed breast. Further details can be found in Mammographic Image Analysis: R Highnam, M Brady, Kluwer Series on Medical Imaging, March 1999, ISBN: 0-7923-5620-9
Analysis of Contrast Enhanced Breast MRI
The second half of the presentation covered the analysis of contrast enhanced breast MRI, and included the following points:
- Modelling contrast agent take up
- Modelling of breast tissue motion
- Image segmentation
- Detection of areas of focal enhancement
- The need for motion correction using image registration. This was a very well presented and informative lecture.
Concluding remarks
The main take home message from these lectures and in fact from the whole course was the importance of understanding the physics behind the image formation process, as this holds the key to improving the processing and analysis of medical images. The summer school was extremely interesting, relevant, and highly enjoyable. Also, I thought that St. Edmund Hall was a good location for the School. I would like to thank the Mayneord-Philips Trust and its sponsors, (BIR, IOP and IPEM) for supporting my attendance through a £200 bursary.
Fraser Hatfield
Taken from IPEM SCOPE 8(4) December 1999
Friday's Lecture Notes Online
Image Processing in MR(pdf 28kb). Prof. Jo Hajnal
Image Processing in Nuclear Medicine(pdf 58kb) Prof. Andrew Todd-Pokropek
Mammographic Image Analysis. Prof. Mike Brady
Back to the Start
1999 Summer School
To read the PDF document with a screen reader please link to the Adobe Accessibility Resource Center