British Journal of Radiology (2006) 79, 455-463
© 2006 British Institute of Radiology
doi: 10.1259/bjr/56511504
Three-dimensional MRI of the male urethrae with implanted artificial sphincters: initial results
J Deng, MMed, PhD1,4,
M A Hall-Craggs, FRCR2,
M D Craggs, PhD3,
R Richards, PhD1,
S L Knight, MSc, PhD3,
A D Linney, PhD1 and
A R Mundy, FRCS3
1 Department of Medical Physics and Bioengineering, 2 Department of Radiology and, 3 Institute of Urology, 4 Department of Obstetrics and Gynaecology, University College London, Gower Street, London WC1E 6BT, UK
 |
Abstract
|
|---|
The aim of this study was to develop a method for simultaneous 3D visualization of a new type of artificial urethral sphincter (AUS) and adjacent urinary structures. Serial MR tomograms were acquired from seven men after AUS implantation. 3D reconstruction was performed by thresholding original (positive) and inverted (negative) image intensity and by subsequently fusing positive and negative images. Results show that the bladder, cuff and balloons of the AUS of originally high intensity were imaged in 3D by thresholding the positive datasets. The urethrae and corpora cavernosa penis of originally low intensity were displayed in 3D by thresholding the negative datasets. Fusion of the positive and negative datasets allowed simultaneous visualization of the AUS complex and adjacent urinary structures. All the structures of interest were also clearly seen by interactive multiplanar reformatting. Coronal tomographic datasets provided better 3D and reformatted 2D images than sagittal and transverse datasets. This technique offers a simple means for evaluating the complex urethral anatomy and the AUS, and has potential for improved 3D visualization of many other complex morphological and pathological conditions.
|
Introduction
|
|---|
Artificial urinary sphincters (AUS) have been used for many years to treat male stress urinary incontinence. The most widely-used type is the American Medical Systems AS800 AUS, but this is not without its problems, the most serious one being urethral erosion caused by the constant pressure of the inflated AUS on the urethra [1]. Post-operative imaging is important for assessing the position, configuration and function of an implanted AUS. The AS800 contains radio-opaque substances to make it visible on X-ray CT, but this exposes the patient to ionizing radiation. In addition, the contrast agents can degrade the device.
To reduce the erosion risk, a new type of AUS with conditional occlusion has been developed by Craggs et al [2]. The Craggs AUS consists of four main parts (Figure 1
). The perineal part is a cuff implanted around the bladder neck or urethra. The next two, intrapelvic or lower abdominal parts, are made up of two balloons, usually placed extraperitoneally close to the bladder. The fourth part is a scrotal pump connected to the balloons and cuff by tubing. The whole implant is made from silicone rubber and contains normal saline. This lends itself to MRI, a modality using no ionizing radiation. In addition, MRI can embrace a large body volume without a need, unlike ultrasound, for a particular imaging window and direct contact with the body parts being examined. This is necessary for visualizing the Craggs AUS because it occupies both intrapelvic and perineal spaces, and its cuff is implanted around the delicate urethral tissue that should not be deformed for functional assessment. Consequently, MRI was chosen to visualize the AUS in this study.
View larger version (23K):
[in this window]
[in a new window]
|
Figure 1. Diagram of Craggs artificial urinary sphincter. 1: urethral cuff; 2: primary reservoir/balloon; 3: additional reservoir (stress-relief balloon); 4: scrotal pump. The design allows a temporarily increased intra-abdominal pressure (caused by a stress such as coughing) to be transmitted from 3 into 1. This prevents stress incontinence, as well as urethral erosion that could be caused by a constant high pressure on the urethra even during rest when a conventional artificial sphincter was used (Artwork by Martin Knight).
|
|
Ultrasound, CT and MR imaging have all been used for visualizing the pelvis in incontinent patients [3–5]. Conventional two-dimensional (2D) imaging requires expertise to compose a mental three-dimensional (3D) picture from a series of 2D images. Studies have demonstrated that 3D ultrasound [6, 7], CT [8] and MRI [9, 10] improve comprehension of the spatially complex morphology in the pelvis.
To the best of our knowledge, there are no previous reports of the use of MRI for the assessment of AUS. Initially, we used cross-sectional images, directly available from standard MR scanners, for this assessment, and 2D MR appearances of the Craggs AUS will be described in a separate report. In brief, it is easy to comprehend the structures and positions of different parts (except the connecting tubes) of the AUS and their relationships to corresponding local body parts through serial 2D images. However, it is difficult to form a reliable 3D impression of all the AUS parts, the complex connections between them and the global spatial relationship to the relevant anatomies. This is clinically desirable, particularly when cross-sectional imaging planes are not perfectly aligned with ideal anatomical axes or imaginary axes of the AUS parts.
Recently, 3D reconstruction has become available on some more advanced MR scanners. Three 3D display methods are often used on these systems. The first method is multiplanar reformatting. This requires further mental work from the operator to construct a 3D object. The second is volume 3D display, in which an object is rendered somehow transparent so that all the structures inside it become more or less visible, even when views to one structure may in reality be blocked by other structures in the front. The third, surface 3D display, probably offers the most realistic 3D images as structures of interest are represented as solid objects. Surface display is often achieved by first applying a threshold so that structures with signal intensities below it will not be rendered. This is sufficient for some clinical applications, such as for a CT skeletal examination; high-intensity bone is visualized while low-intensity soft tissues are "removed". However, this approach is not sufficient for the imaging of complex biological structures where a simultaneous display of different intensity structures is necessary. As in this study, there is a need to show the AUS (mainly of high intensity) simultaneously with the bladder and urethra (mainly of high and low intensities, respectively). When a threshold appropriate for displaying the high-intensity structures (the bladder and AUS) is applied, the low-intensity structures (the urethra) are not seen. If a lower threshold appropriate for displaying the urethra is applied, irrelevant structures of high (fat) or intermediate (other soft-tissues) intensity are also shown, obscuring both low- and high-intensity structures of interest. In other words, simple thresholding could not be used to present the two complexes simultaneously and distinguishably.
In this paper, we describe the "inversion-fusion" method we have developed for processing sequential 2D MR datasets of patients with implanted Craggs AUSs [2]. This facilitates the simultaneous 3D visualization of the implant and the lower urinary tract. We then examine whether this 3D method could bring out information additional to 2D imaging. Finally, we discuss the feasibility of using this relatively simple, generic methodology with 3D facilities available on advanced commercial imaging modalities to improve 3D data reconstruction.
|
Materials and methods
|
|---|
Data acquisition
With approval from the Local Ethics Committee, nine patients suffering persistent post-prostatectomy incontinence received the Craggs AUS implantation after giving informed consent. Excluding one patient with extensive surgical metalwork in his pelvis, all other eight underwent MR scans about 2 months after implantation. They were scanned on a 1.5 T Siemens Vision system (Erlangen, Germany) using a phase-array body coil. Images were acquired using T2 weighted 2D Tru-FISP sequence (repetition time (TR): 6.32 ms; echo time (TE): 3 ms; field of view (FOV): 230 mm; rectangular field of view (RFOV): 50%; flip angle (FA): 70°; time of aquisition (TA): 12–21 s) in the coronal, transverse and sagittal planes.
In this study, the term "series" is being used to describe a set of anatomically consecutive slices acquired during one MR scan. Because of the difficulty comprehending global relationships between all AUS parts and relevant urinary structures using individual series acquired from the first MR patient, several anatomically consecutive series were acquired from each of the remaining seven patients (aged from 59 years to 76 years). This allowed the entire AUS and the neighbouring anatomy to be sampled by these series at regular intervals.
A total of 86 series were recorded from the 7 patients (4 to 20 from each). Each series consisted of 7 to 12 contiguous slices, all with a slice thickness of 5 mm (minimum achievable by the scanner at the time). A serial 3D dataset was then formed by one series or by combining two to six consecutive series, consisting of 7 to 42 slices (Table 1).
In five patients, some serial datasets were acquired in pairs, offset by 50% of the slice thickness (Figure 2a,b), and each of the pairs was later used for creating an interlaced 3D dataset (Figure 2c).
View larger version (35K):
[in this window]
[in a new window]
|
Figure 2. Schematic drawings of data acquisition and post-processing. Pixels (voxels) of high, intermediate and low intensities on original (positive) images are numbered 1, 2, and 3, respectively. (a,b) Two series are acquired, each with a 5 mm slice interval. The second series is scanned with a 2.5 mm offset from the first one. (c) After interweaving the slices of the two series, a (positive) dataset (Slices 0–5) is created, with the slice interval halved to 2.5 mm and the number of slices doubled. A negative dataset (Slices 0'–5') is also created, changing pixels of low-intensity (numbered 3) into high-intensity. (d) Two thresholds appropriate for displaying originally high intensity pixels in Slices 0–5 and negatively high intensity pixels in Slices 0'–5', correspondingly, remove intermediate and low intensity pixels. Background pixels in both positive and negative datasets are not displayed. The remaining pixels (voxels) are areas of interest, which can then be fused into a single 3D image (not shown here), restoring the original spatial relationship between pixels (voxels) 1 s and 3 s.
|
|
The pixel matrix in each slice was 256x256, with a pixel size of 0.8984x0.8984 (mm). The voxel matrix (a 3D dataset) was therefore 256x256 for x and y dimensions and from 7 to 42 for z dimension. The resultant voxel size is 0.8984x0.8984x5.0000 (mm) for serial datasets and 0.8984x0.8984x2.5 (mm) for interlaced datasets.
Data processing
The datasets were transferred to an MGI 3D Workstation (Medical Graphics & Imaging Group, University College London, www.medphys.ucl.ac.uk/mgi/workstat.htm), and converted into MGI format using DispImage program [11]. Interlaced 3D datasets were generated by interweaving and then renumbering paired slices from paired serial datasets (Figure 2c
).
There were two complexes of interest: the urinary complex consisting of the bladder, urethra and penis, and the AUS complex consisting of the cuff, the balloons and the tubes connecting them. On T2 weighted images, the urinary complex mainly showed low intensity except for the urine within the bladder and urinary tract, which was high signal. The AUS complex showed high intensity. As mentioned earlier, simple thresholding could not be used to present the two complexes together clearly.
To address this problem, a negative dataset was created by inverting the intensity of the native, positive dataset (Figure 2c,d and Figure 3). The low-intensity structures (urethra and penis) became high-intensity ones in the negative dataset. The positive and negative datasets were then combined into a new, positive+negative dataset (Figure 2c).
3D reconstruction
This was also carried out on the MGI Workstation. The system provides various options for the reconstruction and manipulation of 3D datasets from serial tomographic imaging modalities [12–16]. 3D surface display and multiplanar reformatting were used in this study to present the structures of interest for analysis.
3D surface display for structures in the overlapping datasets was achieved by four main steps. First, two blocks containing structures of interest in a positive+negative sequence were defined under the guidance of multiplanar reformatting. This avoided segmenting most of the fat in the abdomen, skeletal muscles in the pelvic floor and thighs (in a negative dataset which had more or less the same intensity as that of the urethra) and the image background area (which became the highest intensity area in a negative dataset). Second, two thresholds appropriate for rendering the originally high-intensity structures (the bladder, the balloon and the cuff) and the inversely high-intensity structures (the corpora cavernosa) were applied to the two blocks separately. This resulted in the structures of interest being roughly visualized (Figure 2d). Third, some "image surgery" was performed to remove remaining unwanted parts that were still above thresholds and therefore might obscure the viewing of structures of interest. Having done these, the two complexes could be displayed for inspection separately (Figure 2d and Figure 4a). Finally, by putting the two objects into the same co-ordinate system, a fused image is created for assessing their spatial relationships (Figure 4b and Figure 5).
View larger version (44K):
[in this window]
[in a new window]
|
Figure 4. 3D reconstruction with surface displays.(a) Structures above corresponding thresholds in the positive and negative sequences are separately visualized. In order not to obscure the smaller structures (in grey), the positive sequence is placed closer to the readers, so the bladder is placed in front of the lower abdomen. The threshold in the negative sequence is set to intermediate low so that the separations of the three spongy structures are displayed as solid while their inner regions as (artificially) hollow. (b) The structures from both positive and negative sequences are fused together to reveal the relationship between the cuff and the urethra; in this case, the cuff is fully sealed (also see movies on our website for a better 3D perception).
|
|
View larger version (78K):
[in this window]
[in a new window]
|
Figure 5. A 3D surface display showing that the cuff is not fully wrapped around the corporus cavernosum urethrae. Again, it is an image after fusing the cuff in the positive and the penis in the negative sequences. The threshold in the negative sequence is set very low so that the entire spongy structures are displayed as solid(compare with Figure 4). Also see a movie on our website.
|
|
Data analysis
The level of analysis was at the dataset level (Table 1) and between various scan settings (Table 2). The quality of the final reconstructed 3D images in terms of depicting the spatial relationship between the structures of interest was scored by two observers (JD, RR) as poor, adequate and good. "Poor" means that in a single fused 3D object, the spatial relationship between the structures of interest cannot be visualized, and subsequently, the AUS functioning position cannot be diagnostically assessed. "Adequate" means that in a single fused 3D object, the spatial relationship of the AUS cuff to the corpora cavernosa urethra was visualized. Some other parts of the urinary tract or of the AUS were not clearly visualized, but this did not affect our judgement of the completeness of the AUS cuff wrapping the urethra, the most important diagnostic issue to be assessed. "Good" means that in addition to the cuff-urethral relationship, the entire AUS complex and the entire urinary complex were well visualized.
View this table:
[in this window]
[in a new window]
|
Table 2. Initial comparison of the number and percentage of useful 3D series/datasets between various scan settings and post-processings
|
|
|
Results
|
|---|
A total of 35 datasets containing the region of interest for 2D and 3D analysis were obtained directly or combined from 86 series acquired. Eight interlaced datasets were further created from the 35 datasets (Table 1). Depending on data quality and suitability, it took about 10–30 min to process one sequence and attain fused 3D images. The longer time was caused by processing few, but graphically more difficult, sagittal and transverse datasets using manual segmentation of regions of non-interest (i.e. unwanted abdominal fat or other tissues with complex greyscale composition, see Discussion on Image processing). The "inversion-fusion" method itself, together with series-combination and slice-interlacing, only took a few seconds.
The 3D results were listed in Table 1. The comparison of the 3D images acquired from coronal, sagittal and transverse base data is shown in Table 2. The coronal datasets offered the best 3D results with 65.2% good and 34.8% adequate images. None of the sagittal and transverse datasets produced good 3D images, but six of 10 sagittal datasets were adequate. There were only two transverse datasets, precluding assessment.
Interlaced datasets provided the best 3D results with good 3D images obtained in seven (87.5%) of eight datasets (all derived from coronal datasets). Motion artefact was noticeable in all eight datasets, but this degraded the bladder more than the cuff (Figure 6).
View larger version (63K):
[in this window]
[in a new window]
|
Figure 6. Multiplanar reformatting of the same 3D dataset inFigure 4a: Three orthogonal images are obtained by cross-sectioning the bladder, showing the motion artefact (arrowheads) in the interlaced data, which appears more severe in the upper side of the bladder. (b) Another three orthogonal images are obtained by cross-sectioning the cuff, showing the well-sealed cuff, although not so apparent as in 3D images. Note the depiction of the tubing (arrows) connecting the cuff and the balloons. Its entire course may not be visualized by 3D surface display due to its heterogeneous signal intensity, but can be traced by interactively reformatting sequential 2D images.
|
|
The spatial relationships of all the structures of interest were best assessed when the images from both positive and negative datasets were fused. Urinary structures and the AUS were clearly demonstrated using 3D surface display, and this was particularly well seen when viewed as a 3D movie (available at: www.medphys.ucl.ac.uk/mgi/jdeng/ follow the links "Dynamic Face/Body Parts" on the top panel, then "Erecting Penis" on the left panel). The 3D images have depicted various cuff conditions between the well-implanted, fully wrapped (Figure 4) to the failed, fully open (Figure 5), and from un-inflated to fully inflated.
2D views of the region of interest can be interactively obtained using multiplanar reformatting under the guidance of 3D objects (Figure 6), offering a means for detailed examination of structures that may not be well visualized by simple 3D reconstruction. This is particularly useful for following up structures of heterogeneous intensity such as the tubing between the balloons and cuffs and for following the thin urine residues along the urethra that were not acquired in single original imaging planes. All these have been found difficult when only viewing original 2D slices.
|
Discussion
|
|---|
2D versus 3D imaging
This initial study has shown that, by using our "inversion-fusion" method, 3D images simultaneously displaying both AUS and urinary complexes can be obtained from serial MR datasets. Although subject to further large-scale studies, visualizing the datasets as fused 3D objects and through reformatted 2D images appears to have at least three advantages over viewing them as original individual 2D slices.
First, images reconstructed by 3D surface rendering can offer more straightforward, panoramic views of all four main parts of the AUS and their relationships to each other and to the urinary tract. This is important for post-operative assessment as the four parts are implanted in three very different anatomical places. It is possible for an experienced radiologist to form such a 3D impression from individual slices, but it requires greater training while still subject to human errors.
Second, multiplanar images reformatted from a 3D dataset in any desirable orientation allow tracking of the entire course of the tubing connection between the AUS parts. This has proved to be very difficult when tracking through the original 2D slices. It is crucial to check this connection when blockage and/or entanglement are suspected.
Third, the full 3D visualization of the AUS complex has paved the way for its volumetric quantification. This will be helpful for evaluation of possible AUS leakage by comparing the given amount of saline pumped in to the increment in volume before and after the pumping. It is impossible to work out such a change mentally unless there is serious leak.
More generically speaking, this study has shown that clinically useful 3D images can be attained by combined use of commercially available modality and acquisition methods (MR interlaced acquisition with coronal sequences) with relatively simple post-processing and display methods (intensity inversion and image fusion).
Admittedly, 3D techniques have not been sophisticatedly developed such that they can render texture details. In this study, these features, such as the intricate structures of the AUS were seen better on the 2D slices (either from original cross-sectional data or from multiplanar reformatting).
Imaging modality
Ultrasound examination has a narrow FOV for superficial structures, and soft-tissue deformation results from the direct contact between transducer and the skin [15]. There is also insufficient definition of the AUS structures, as seen in our unpublished 3D ultrasound study of some of the patients. These prevent ultrasound from being our first choice for structural assessment. However, the latest real-time 3D colour Doppler ultrasound is likely to play a role in evaluation of penile haemodynamics under different AUS hydraulic and other physiological circumstances. It may also enable real-time checking of leakage during pumping [17–19].
Imaging resolution
The in-plane image quality obtained from this MR study is much higher than that obtained from our unpublished 3D ultrasound study. This is because, compared with ultrasound, MRI has higher intrinsic tissue contrast and no restriction on imaging window [20, 21]. However, the out-of-plane (z-dimension) spatial resolution is far from ideal. For instance, while the voxel sizes in the x and y dimensions were equally small (about 1 mm), the size in the z-dimension was 5 mm. To address this unbalanced match, interlaced acquisitions were performed in the last five patients and this halved the z voxel size to 2.5 mm (Figure 2b). Thinner slices and true 3D data sets can now be acquired in our centre after this study. Although this will reduce the degree of anisotropy, it is unlikely to be eliminated. In addition, no matter how thin a slice may become available from a newer scanner, radiologists will immediately expect to see tissue details at an even smaller scale. Consequently, interlaced acquisition may still be helpful in future studies.
Imaging orientation
In this study, coronal datasets provided the best 3D results (Table 2). This is because the slices are obtained perpendicular to the long axis of the penile urethra and the cuff, offering multiple serial slices through the structures of interest. In contrast a sagittal or transverse series acquires only three to four slices through the cuff in the penis, simply not offering sufficient data for 3D reconstruction. Though, recently available thinner slice acquisition may change this.
Motion artefacts
There are several sources of movement and misregistration artefacts that cause image degradation: respiration, peristalsis, urine and blood flow, gross body and penile movement. Because the sequence acquisition time was up to 21 s, periodic motion was likely to cause artefacts. However, obvious motion artefact was mainly due to misregistration between the added image series. This is mainly because it is difficult to breath-hold at exactly the same point of respiration for each image series.
Fortunately, the impact of motion artefact appeared minimal in the region of the cuff (Figure 6), primarily because it was extraperitoneal and stabilized by the urethra and local ligaments. In the future, faster acquisition and motion tracking [22] may minimize motion artefacts.
Image processing
With a few computer mouse clicks, the interlacing and inverting techniques have improved resolution and speeded up segmentation of the region of interest, respectively. Nonetheless, cumbersome manual segmentation is still needed for removal of regions of non-interest but of a wide-range of mixed signal intensities.
Clinicians are frequently caught in a dilemma in 3D segmentation of graphically complicated structures. On one hand, simple algorithms such as thresholding are often unable to depict all the structures of interest that have heterogeneous intensities (even when they are made up of the same biological tissues), or have variable intensity gradients between different tissue borders. This often results in incomprehensible or even misleading 3D images. On the other hand, complex algorithms often demand extra expertise and are time-consuming. This study has tried to strike a balance between the simplicity of the image processing and the quality of the resulting images. Some of the generic tricks developed in this study such as "inversion-fusion", can be easily adapted by commercial systems to speed up and simplify 3D reconstruction. In fact this has been implemented in the latest MGI 3D Workstation for commercial release. In addition, automatic contouring with operator interactive correction may allow rapid segmentation with sufficient accuracy. Automatic contouring may also be used for fast removal of extraneous, but homogeneous, tissues, such as high-intensity fat from the abdominal wall (Figure 6), which would otherwise obscure the region of interest in 3D images. Hence, the entire post-processing and display time are expected to be reduced to about 10 min.
A final point to make is that for visual reality and clarity, the "inversion-fusion" method was only used for 3D surface rendering in this study, but in principle, it can also benefit volume rendering.
|
Conclusion
|
|---|
This MR study has demonstrated the feasibility of simultaneous visualization of all relevant parts of the AUS and the lower urinary tract in three dimensions using these novel image processing methods. This has potential for reducing difficulty in comprehending the spatial position, geometry and function (and malfunction) of implanted sphincters and the impact on the local anatomy. The inversion-fusion approach can be developed as a general application for simplifying and improving 3D visualization of many other complex morphological and pathological conditions.
JD is supported by an MRC Clinician Scientist Fellowship. We acknowledge some of the technical and equipment support from EPSRC-MRC's MIAS IRC and RCR's Pump Priming grant. The prototype artificial urinary sphincters were developed in collaboration with Bibby-Sterilin Limited and Isotron plc in a British Government Link project funded by the Department of Health and the Department of Trade & Industry.
|
Acknowledgments
|
|---|
We thank the MR staff of Middlesex Hospital for data acquisition, D Plummer and D McDonald of Medical Physics Department for developing DispImage and other image conversion programmes. We are also grateful to A Todd-Pokropek for advice on MR data processing and C H Rodeck of Obstetric Department for research planning.
Received for publication October 1, 2004.
Revision received April 19, 2005.
Accepted for publication October 13, 2005.
|
References
|
|---|
- Venn SN, Greenwell TJ, Mundy AR. The long-term outcome of artificial urinary sphincters. J Urol 2000;164:702–6.[Medline]
- Craggs MD, Mundy AR, Dwyer PL, Susser J, Knight SL. Long-term clinical outcome of a new AUS with conditional occlusion for genuine stress incontinence: optimal cuff pressures for continence. BJU Int 2002;90:18
- McFarlane IP, Foley SJ, Shah PJ. Long-term outcome of permanent urethral stents in the treatment of detrusor-sphincter dyssynergia. Br J Urol 1996;78:729–32.[Medline]
- Maki DD, et al. Injected periurethral collagen for postprostatectomy urinary incontinence: MR and CT appearance. Abdom Imaging 2000;25:658–62.[CrossRef][Medline]
- Mostwin JL, Genadry R, Saunders R, Yang A. Stress incontinence observed with real time sonography and dynamic fastscan magnetic resonance imaging--insights into pathophysiology. Scand J Urol Nephrol Suppl 2001;94–9.
- Klein HM, Kirschner-Hermanns R, Lagunilla J, Gunther RW. Assessment of incontinence with intraurethral US: preliminary results. Radiology 1993;187:141–3.[Abstract/Free Full Text]
- Umek W, et al. The urethra during pelvic floor contraction: observations on three-dimensional ultrasound. Obstet Gynecol 2002;100:796[Abstract/Free Full Text]
- Stenzl A, et al. 3-dimensional computerized tomography and virtual reality of the lower urinary tract in women. Int Urogynecol J Pelvic Floor Dysfunct 1999;10:248–53.[Medline]
- Myers RP, et al. Puboperineales: muscular boundaries of the male urogenital hiatus in 3D from magnetic resonance imaging. J Urol 2000;164:1412–5.[Medline]
- Ueno S, et al. Three-dimensional display of the pelvic structure of anorectal malformations based on CT and MR images. J Pediatr Surg 1995;30:682–6.[Medline]
- Plummer DL. A display and analysis tool for medical images. Rivista di Neuroradiologia 1992;5:489–95.
- Linney AD, et al. Three-dimensional visualization of data on human anatomy: diagnosis and surgical planning. J Audiov Media Med 1993;16:4–10.[Medline]
- Brookes JA, Deng J, Wilkinson ID, Lees WR. Three-dimensional imaging of the postmortem fetus by MRI: early experience. Fetal Diagn Ther 1999;14:166–71.[CrossRef][Medline]
- Deng J, Brookes JA, Gardener JE, Rodeck CH, Lees WR. Three-dimensional magnetic resonance imaging of the postmortem fetal heart. Fetal Diagn Ther 1996;11:417–21.[Medline]
- Deng J, et al. Novel technique for three-dimensional visualisation and quantification of deformable, moving soft-tissue body parts. Lancet 2000;356:127–31.[CrossRef][Medline]
- Deng J, et al. Dynamic three-dimensional colour Doppler ultrasound of human fetal intracardiac flow. Ultrasound Obstet Gynecol 2002;20:131–6.[CrossRef][Medline]
- Deng J. Terminology of three-dimensional and four-dimensional ultrasound imaging of the fetal heart and other moving body parts. Ultrasound Obstet Gynecol 2003;22:336–44.[Medline]
- Deng J. Minimally compressed imaging of deformable moving body parts using dynamic 3D ultrasonography and colour Doppler - with facial and genital parts as exemplars. Euroson School on 3D Ultrasound Imaging. London: Hammersmith Hospital, UK, May 4–6, 2005. p81–93
- Deng J, et al. Real time three-dimensional ultrasound visualization of erection and artificial coitus. Int J Androl 2005; Online early (doi: 10.1111/j.1365-2605.2005.00617.x)
- Coakley FV, et al. Urinary continence after radical retropubic prostatectomy: relationship with membranous urethral length on preoperative endorectal magnetic resonance imaging. J Urol 2002;168:1032–5.[CrossRef][Medline]
- Mikuma N, Tamagawa M, Morita K, Tsukamoto T. Magnetic resonance imaging of the male pelvic floor: the anatomical configuration and dynamic movement in healthy men. Neurourol Urodyn 1998;17:591–7.[Medline]
- Keegan J, Gatehouse P, Yang GZ, Firmin D. Coronary artery motion with the respiratory cycle during breath-holding and free-breathing: implications for slice-followed coronary artery imaging. Magn Reson Med 2002;47:476–81.[Medline]
Top
Abstract
Introduction
