| STAGING OF PROSTATIC CANCER
USING MAGNETIC RESONANCE IMAGING: COMPUTER-ASSISTED IMAGE ANALYSIS
Stuart Williams MA MRCP FRCR, Consultant Radiologist at the Norfolk & Norwich University Hospital. Reyer Zwiggelaar PhD, Lecturer, School of Computing Sciences, University of East Anglia, Norwich |
Aims of Project
We wished to develop a software tool that will benefit patients with prostate cancer by aiding accurate interpretation of their staging examinations.
During the project, we have been investigating computer-aided detection and staging of prostate cancer from MR images [1]. The achievement so far is promising.
To segment the prostate from MR images, Zwiggelaar et al. developed a semi-automatic method based on polar-transform [2]. Simple techniques, such as line detection and non-maximum suppression, are used to track the boundary of the prostate in a polar-transformed image. The initial results, based on a small set of data, indicate a good correlation with expert based manual segmentation. Zhu et al. applied Active Shape Modelling (ASM) to prostate segmentation from MR images [3]. The results were encouraging when compared to the semi-automatic segmentation based on polar-transform, and manual annotations. To further improve the 2D segmentation results, Zhu et al. applied Gaussian Mixture Models to the modelling of intensity information in ASM method and obtained reasonable results [4,5]. Zhu et al. went on to develop a hybrid 2D+3D shape modelling approach and extract 3D prostate surfaces from volumetric data [6,7]. The results, based on twenty-six MRI volumes, showed a high correlation with the manual annotations. When compared to 2D ASM, the hybrid ASM approach shows equivalent precision on individual slices, but higher consistency with regarding to capturing the 3D surface of the prostate.
The latest developments by Zwiggelaar et al. are related to the staging of prostate cancer based on MRI data [8]. Information local to the prostate boundary was used to determine if cancer was confined to the prostate gland. Initial evaluation indicated promising classification results. However, at the same time there are clear limitations covering data normalisation and generalisation.
Overall, our research resulted several approaches to segment the prostate gland from both 2D images and 3D volumes with considerable precision, and initial work on prostate cancer staging has shown promising prospect of this research work.
References
[1] Zhu, Y., Williams, S., Zwiggelaar, R., “Staging of Prostate
Cancer Using Magnetic Resonance Imaging: Computer-Assisted Image Analysis”,
UK Radiological Congress 2004, Manchester.
[2] Zwiggelaar, R., Zhu, Y., Williams, S., “Semi-automatic segmentation
of prostate MRI”, Lecture Notes in Computer Science 2652, pp. 1108-1116,
2003
[3] Zhu, Y., Zwiggelaar, R., Williams, S., “Prostate segmentation:
A comparative study”, In: Proceedings of Medical Image Understanding
and Analysis, pp. 129-132, 2003
[4] Zhu, Y., Williams, S., Zwiggelaar, R., “An Improved ASM Approach
Using Mixture Models for to Prostate Segmentation from MR Images”,
In: Proceedings of Medical Image Understanding and Analysis, pp. 248-251,
2004.
[5] Zhu, Y., Williams, S., Zwiggelaar, R., “Improving ASM search
using mixture Models for grey level profiles”, In: Proceedings
of 7th International Conference on Pattern Recognition and Image Analysis,
to appear, 2004.
[6] Zhu, Y., Williams, S., Zwiggelaar, R., “Segmentation of volumetric
prostate MRI data using hybrid 2D+3D shape modelling”, In: Proceedings
of Medical Image Understanding and Analysis, pp. 61-64, 2004
[7] Zhu, Y., Williams, S., Zwiggelaar, R., “A hybrid ASM approach
for sparse volumetric data segmentation”, In: Proceedings of 7th
International Conference on Pattern Recognition and Image Analysis, to
appear, 2004.
[8] Zwiggelaar, R., Zhu, Y., Williams, S., “Towards classification
of prostate MRI”, In: Proceedings of Medical Image Understanding
and Analysis, pp. 204-207, 2004.
Research summary, 25 October 2004
Project G2002/18.