 |
| Figure 1: Enhancement curves for different types of tissue |
Breast cancer is the most frequently diagnosed cancer and the most common cause of cancer related death in Australian women. The most recent national figures show that:
- a woman has a 1 in 11 chance of developing breast cancer before the age of 75;
- more than 11,000 women are diagnosed with breast cancer every year;
- more than 2500 women die from breast cancer each year; and
- the risk of breast cancer increases with age.
Breast cancer screening in Australia and worldwide is based on x-ray mammography because of its low cost and moderate to high sensitivity. Nevertheless x-ray mammography has several shortcomings that have prompted the exploration of alternative imaging modalities. The most promising of these is magnetic resonance imaging (MRI).
DCE-MRI
Dynamic contrast-enhanced (DCE) MRI is becoming more widely used in the clinical setting as a supplemental imaging modality, in particular to rule out cancer in the case of suspicious lesions found using conventional x-ray mammography, and to evaluate the extent of disease within a breast affected by cancer. DCE-MRI involves imaging the breast volume as a set of two-dimensional (2D) slices, before and at several times after the injection of a contrast agent, to yield a 4D data set. Tissue of a malignant nature exhibits characteristic rapid uptake, plateau, and washout kinetics, reflected in signal vs. time curves. Benign tissue displays either no enhancement, or slowed continued enhancement with delayed washout. Examples of these curves are shown in Figure 1.
Currently the most common method of analysis of the data is the qualitative approach, which involves the manual review of the raw 4D data and derived subtraction images to identify areas of suspicious enhancement. Not only is the data set high in dimensionality, it is large in size. This makes the task of review and interpretation by the radiologist both complex and difficult. The cost of missed diagnosis is high, making the motivation to develop more advanced and effective screening methods evident.
 |
|
Figure 2: Proposed visualisation (single slice). |
The Technology
This project, a collaboration between members of the Electromagnetics and Imaging (EMI) Research Division within the School of ITEE at the University of Queensland, and Queensland X-Ray at Greenslopes Private Hospital, will develop visualisation software for dynamic contrast-enhanced breast MRI for use in the routine clinical setting to identify suspicious lesions with greater accuracy and speed than can be achieved by current practice.
The methods of colour-coding or pixel mapping in use today are generally non-linear and must be fitted pixel-wise using a non-linear least squares algorithm. These algorithms are iterative in nature and computationally very demanding - for a typical 512 × 512 slice image this can take several hours on a high-end desktop PC. They are also prone to problems such as failure to converge. The research group is investigating alternative models and fitting approaches in order to avoid the problems encountered using the more commonly used non-linear methods. The parameters of the fitted data will be used to convert the 4D data to colour-coded 3D data for visualisation and subsequent human review. An example of the data to be obtained from this is shown in Figure 2. The top image shows a clinician-selected region of interest (ROI) with the associated enhancement curve. The bottom image shows the proposed visualisation.
Preliminary results show strong correlation between "hot spots" in the visualisation and ROIs independently selected by a radiologist, however the model fitting still requires several hours per slice. Fortunately the method used to model the data is massively parallelisable, and so the group is investigating solutions based on parallel and grid-based computing (such as on QCIF-funded supercomputing facilities at UQ), which have the potential to speed the process up by several orders of magnitude. The ultimate goal is to develop software to be run on a standard desktop computer, which will be able to generate a visualisation in an amount of time that is acceptable in the clinical setting (minutes).
Participants
Dr Andrew Mehnert, Dr Stephen Wilson,
Professor Stuart
Crozier
Electromagnetics and
Imaging Research Unit, School of ITEE, University of Queensland
Industry Participants
Dr Kerry McMahon, Dominic Kennedy
Queensland X-Ray
Reports
Final Report - May
2008 (294 KB PDF)
Progress Report - November 2007
(209 KB PDF)
Progress Report - August 2007
(62 KB PDF)
Progress Report - March 2007
(249KB PDF)
Progress Report -
November 2006 (188KB PDF)