MR imaging and MR spectroscopic imaging of prostate cancer

Transcription

1 Magn Reson Imaging Clin N Am 12 (2004) MR imaging and MR spectroscopic imaging of prostate cancer Arumugam Rajesh, MB, BS a, Fergus V. Coakley, MD b, * a University Hospitals of Leicester, Leicester General Hospital, Gwendolen Road, Leicester LE 5 4PN, United Kingdom b Department of Radiology, University of California, San Francisco, Box 0628, M-372, 505 Parnassus Avenue, San Francisco, CA , USA Prostate cancer incidence and mortality rates vary worldwide. In the United States, prostate cancer is the most common noncutaneous malignancy affecting men and is the second-leading cause of cancer death. Approximately 8% of American men will be diagnosed with prostate cancer during their lifetime, and 20% of these men will die of the disease [1]. Epidemiology Risk factors for developing prostate cancer include advancing age, African-American ethnicity, and a positive family history. The role of diet and other factors also has been investigated. During the 1990s, the incidence of prostate cancer increased and later decreased, probably as a result of the emergence of widespread serum prostaticspecific antigen (PSA) level testing that changed diagnostic sensitivity. The age-adjusted mortality increased over the same period, however, suggesting an increase in disease prevalence or aggressiveness. More recently, the age-adjusted mortality has decreased, which could reflect earlier diagnosis that allows more effective treatment (Fig. 1). Alternatively, these changes in mortality simply may be a result of attribution bias, particularly because the changes have paralleled those of incidence. Given the indolent natural history of prostate cancer, a true improvement in treatment likely would take several years to affect mortality. The absence of a lag between the * Corresponding author. address: Fergus.Coakley@radiology.ucsf.edu (F.V. Coakley). incidence and mortality changes may support this latter explanation [2]. Despite the sizeable mortality from prostate cancer, many cases are subclinical, and small foci of incidental prostate cancer can be detected in up to 40% of men at autopsy [3]. The autopsy incidence of histologic prostate cancer seems constant among countries and races, but the clinical incidence of prostate cancer is higher in Western countries and in African Americans. The management of early-stage prostate cancer is controversial because patients whose disease is indolent and incidental cannot be distinguished reliably from patients whose disease is progressive and life-threatening. Current methods of prostate cancer evaluation by digital rectal examination (DRE), transrectal ultrasound (TRUS), Gleason score, sextant biopsy, and serum PSA assay generally can predict only behavior for indolent or aggressive cancers. Most patients fall between these extremes, when the techniques are of limited accuracy [4 6]. Because of the prevalence and mortality of prostate cancer and the limitations of current evaluation methods, prostate cancer is a major medical and socioeconomic problem. Staging The tumor, node, metastasis (TNM) and Jewett-Whitmore staging systems are used commonly, and are based on the local, nodal, and distant extent of disease [7,8]. Table 1 summarizes the staging systems. Staging of prostate cancer is crucial to predicting prognosis and planning treatment. A general understanding of prognosis and treatment strategies by stage /04/$ - see front matter Ó 2004 Elsevier Inc. All rights reserved. doi: /j.mric

2 558 A. Rajesh, F.V. Coakley / Magn Reson Imaging Clin N Am 12 (2004) Fig. 1. Epidemiologic trends in prostate cancer incidence and mortality in the United States from 1992 to facilitates clinically relevant radiologic interpretation of prostate MR imaging and MR spectroscopic imaging. Prognosis is related closely to stage. Despite the prevalence of prostate cancer, published studies on prognosis are relatively sparse. Good prognostic studies are lacking because the mortality from unrelated causes is high in elderly men with prostate cancer, and long-term follow-up of 10 to 15 years is required for meaningful evaluation in lower-stage disease. Table 2 summarizes the available data on prognosis of prostate cancer by stage [9 13]. Treatment options are related to stage, and are summarized in Table 3 [14]. There are many Table 1 Staging systems for prostate cancer Jewett-Whitmore TNM Description A I (T1N0M0) Organ-confined tumor. Clinically and radiologically inapparent. B II (T2N0M0) Organ-confined tumor. Clinically or radiologically apparent. T2A: Localized to a quadrant T2B: Localized to one side T2C: Bilateral C III (T3N0M0) Extracapsular extension, or seminal vesicle invasion. a T3A: Unilateral or bilateral extracapsular extension T3B: Seminal vesicle invasion D1 IV (N1-2) Locoregional adenopathy. N1: Microscopic nodal metastases N2: Macroscopic nodal metastases D2 IV (T4 or N3 or M1-2) Distant spread. T4: Invasion of the bladder, external sphincter, or rectum N3: Extraregional nodal metastases M1: Elevated acid phosphatase M2: Distant visceral or bony metastases a In the fourth edition of the American Joint Committee on Cancer staging system, unilateral and bilateral extracapsular extension were classified separately as T3A and T3B. This distinction was dropped from the fifth and subsequent editions.

3 A. Rajesh, F.V. Coakley / Magn Reson Imaging Clin N Am 12 (2004) Table 2 Prognosis in prostate cancer by stage Jewett-Whitmore TNM 2-y mortality 5-y diseasespecific mortality A and B I/II 10% 9 22% C III 18% 40% D1 IV 34% D2 IV 42% 10-y diseasespecific mortality controversies and unanswered questions in the management of prostate cancer, primarily because of the inability to predict accurately the natural history of localized disease [15]. Some recent studies have questioned the validity of watchful waiting or surveillance as an option for those with localized disease and a reasonable life expectancy. A Danish Cancer Registry study showed that patients with clinically localized prostate cancer who are candidates for curative therapy at diagnosis have significant excess mortality when treated expectantly and followed for 10 or more years [16]. A recent Scandinavian, randomized, controlled trial demonstrated improved outcome at 8 years or more of follow-up when radical prostatectomy was compared with watchful waiting [17]. These results may not apply directly to the North American population of men with newly diagnosed prostate cancer, however, in whom disease is impalpable more frequently and detected by PSA testing alone. Watchful waiting may be an appropriate option for some of these patients with early low-grade or small-volume tumors. Anatomy of the prostate Overview The prostate is a pale, firm exocrine gland shaped like an inverted pyramid that surrounds the urethra between the bladder neck and genitourinary membrane. The base lies superiorly (just below the bladder) and the apex lies inferiorly (just above the urogenital diaphragm). The ejaculatory ducts pass obliquely through the gland to enter the prostatic urethra at the verumontanum. The normal adult gland measures approximately 4 cm (transverse) 3 cm (anteroposterior) 3 cm (craniocaudad), and weighs 15 to 20 g. It functions as an accessory sex gland, and contributes approximately 0.5 ml to the normal ejaculate volume of 3.5 ml. The secretions of the prostate are thought to help liquefy semen. The contemporary approach to prostate anatomy describes the internal structure in terms of zones. Zonal anatomy The simplest conceptual approach to the zonal anatomy of the prostate is the two-compartment model, where the prostate is likened to a cone containing a scoop of ice cream [18]. The cone is the peripheral zone, and makes up 70% of the prostate gland by volume in young men. The ducts of the peripheral zone glands drain to the distal prostatic urethra. The scoop of ice cream is the central zone, and makes up 25% of the prostate gland volume in young men (Fig. 2). The ejaculatory ducts traverse the central zone, Table 3 Treatment in prostate cancer by stage Jewett-Whitmore TNM TMN Conventional treatment options A and B I/II T1 and T2 < 10-y life expectancy: radiotherapy a ; watchful waiting > 10-y life expectancy: prostatectomy; radiotherapy C III T3 Radiotherapy; adjuvant hormonal therapy in high-risk patients D1 IV N1-2 Hormonal therapy b D2 IV N3/M1-2 Hormonal therapy; palliative radiotherapy for bony metastases; second-line agents for hormone refractory cancer a Includes standard external beam radiotherapy, three-dimensional conformal radiotherapy, intensity-modulated radiotherapy, and brachytherapy. b Includes subcapsular orchidectomy, estrogens, antiandrogens, and luteinizing hormone-releasing hormone agonists.

4 560 A. Rajesh, F.V. Coakley / Magn Reson Imaging Clin N Am 12 (2004) Fig. 2. (A) Schematic coronal view of the prostate in a young man, illustrating the contemporary zonal description of prostatic anatomy. In a younger man, most of the central gland is composed of central zone tissue. (B) Schematic coronal view of the prostate in an older man, illustrating the progressive enlargement of the transition zone that occurs with age as a result of benign prostatic hyperplasia. Most of the central gland is composed of transition zone tissue. and the ducts of the central zone drain to the region of the verumontanum clustered around the entry of the ejaculatory ducts. The remaining 5% of the prostate consists of the transition zone, which is composed of two small bulges of tissue that surround the anterior and lateral parts of the proximal urethra in a horseshoe-like fashion. This two-compartment model is deficient anteriorly, where the peripheral zone is interrupted by the anterior fibromuscular stroma, a band of smooth muscle mixed with fibrous tissue that forms a thick shield over the anterior aspect of the gland. As a result, the peripheral zone lies predominantly lateral and posterior to the central zone. The zonal anatomy of the prostate changes with age (Figs. 2 and 3). The transition zone becomes bigger as a result of benign prostatic hyperplasia and compresses the surrounding central zone. The latter becomes the surgical pseudocapsule and is radiologically noteworthy because on axial T2-weighted images through the base of the prostate, the compressed central zone may mimic low T2-signal tumor in the peripheral zone around the transition zone (Fig. 4). One approach to incorporate and simplify the age-related changes in the prostate zonal anatomy is to refer to the transition and central zones collectively as the central gland. Using this terminology, the prostate is composed of the peripheral zone and central gland, such that the central gland is composed mainly of central zone tissue in young men and mainly of transition zone tissue in older men. Because prostate cancer is largely a disease of older men, in the population that comes to MR imaging, it is reasonable to regard central gland and transition zone as nearly synonymous. Prostate capsule The anatomic or true capsule of the prostate is a 2- to 3-mm thick layer of fibromuscular tissue, indistinct from the surrounding fascial tissue [19]. The anatomic capsule should not be confused with the surgical or pseudocapsule that develops around the central gland in the aging hyperplastic prostate. The prostatic capsule comprises the fibromuscular stroma that lies between the glandular components of the prostate and the periprostatic loose connective tissue. The histology of the capsule is particularly complex at the apex, because the fibromuscular band is demarcated less well and the glandular elements are not confined as clearly. Occasionally, glandular elements are found loose in the apical fibromuscular stroma. This is particularly relevant to the pathologist attempting to assess extracapsular extension at the prostatic apex because the apex does not have clearly defined histologic landmarks. At the prostatic base, the periphery of the prostate is composed predominantly of prostatic stroma, which merges imperceptibly with the bladder musculature and the stroma of the seminal vesicles. Glandular elements are sparse in this region and usually form epithelial islands surrounded by thick bundles of fibromuscular stroma. Neurovascular bundle Sympathetic nerve fibers from the lumbar sympathetic chain pass inferiorly into the pelvis alongside the aorta and iliac arteries [20]. Parasympathetic fibers enter the pelvis as direct branches of S2 to S4. Both sets of fibers intermix as a mesh of nerves posterior to the bladder, seminal vesicles, and prostate. This mesh is known as the pelvic plexus. The cavernous nerve arises as many

5 A. Rajesh, F.V. Coakley / Magn Reson Imaging Clin N Am 12 (2004) Fig. 3. (A) Coronal T2-weighted image of the prostate in a young man. The prostate is small, and zonal differentiation is not appreciable. (B) Coronal T2-weighted image of the prostate in a middle-aged man. The transition zone (asterisk) is visible and moderately enlarged from benign prostatic hyperplasia. The compressed central zone (arrows) is seen at the periphery of the transition zone. (C) Coronal T2-weighted image of the prostate in an older man. The transition zone (asterisk) is enlarged markedly by benign prostatic hyperplasia. The central zone (thick arrows) is compressed and forms the pseudocapsule. The peripheral zone (thin arrows) also is compressed. fine fibers from the pelvic plexus, containing sympathetic and parasympathetic nerves. The cavernous nerve then runs inferiorly, as one or several large bundles, along the posterolateral aspect of the prostate. Arterial and venous prostatic vessels in this location accompany the cavernous nerve, and together these structures form the neurovascular bundles. Prostatic anatomy as seen on MR imaging The zonal anatomy of the prostate cannot be distinguished on T1-weighted images because of uniformly intermediate signal intensity [20]. The prostatic zones are well demonstrated on T2- weighted images (Fig. 5). The anterior fibromuscular stroma is of low T1 and T2 signal intensity [21]. The peripheral zone is of high T2 signal intensity, similar to or greater than the signal of adjacent periprostatic fat. The anatomic or true capsule surrounding the peripheral zone appears as a thin rim of low signal intensity on T2- weighted images. The central and transition zones are of lower T2 signal intensity than the peripheral zone, possibly because of more compact smooth muscle and sparser glandular elements [21]. There is also an age-related increase in the T2 signal intensity of the peripheral zone [22].

6 562 A. Rajesh, F.V. Coakley / Magn Reson Imaging Clin N Am 12 (2004) Fig. 4. (A) Axial T2-weighted image of the prostate shows symmetric low T2 signal intensity (arrows) at the base of the gland, which could be interpreted as tumor. (B) Coronal T2-weighted image of the prostate showing the axial level (line) through which (A) was obtained, demonstrating that the low T2 signal at this level is caused by transverse sectioning through the pseudocapsule and does not represent cancer. The proximal urethra is rarely identifiable, unless a Foley catheter is present or a transurethral resection has been performed. The verumontanum can be visualized as a high T2 signal intensity structure. The distal prostatic urethra can be seen as a low T2 signal intensity ring in the lower prostate because it is enclosed by an additional layer of muscle [20]. The ability of MR imaging to provide multiplanar images of the prostate, with separation of the gland from adjacent structures, makes it the ideal modality to perform volumetric measurements. The prostate volume is half the product of the maximum craniocaudad, anteroposterior, and transverse dimensions [23]. The vas deferens and seminal vesicles are seen particularly well on axial and coronal images, whereas the neurovascular bundles can be seen best on axial images. The penile root can be seen inferiorly, separated from the prostatic apex by the urogenital membrane. For descriptive purposes, the prostate is described conventionally in terms of sextants, based on division of the gland into thirds in the craniocaudad direction (base, midgland, and apex), and then into left and right sides. Accordingly, the six sextants are the left base, left midgland, left apex, right base, right midgland, and right apex. Technique of prostate MR imaging Fig. 5. Axial T2-weighed MR image of the prostate, illustrating the zonal anatomy and neurovascular bundles (arrows). MR imaging and MR spectroscopic imaging allow a one-stop shop for detailed anatomic and metabolic evaluation of the prostate gland. Neither TRUS nor CT can offer this simultaneous coverage and tissue detail. MR spectroscopic imaging is a method of demonstrating normal and altered tissue metabolism, and is therefore fundamentally different from other imaging modalities that only assess abnormalities of structure. MR imaging, and especially MR spectroscopic imaging, are relatively new technologies in continued evolution. It is too early to judge their true role, which may not be established for years. For example, an early multiinstitutional trial showing no difference between

7 A. Rajesh, F.V. Coakley / Magn Reson Imaging Clin N Am 12 (2004) Fig. 6. (A) Photograph of the inflatable balloon-covered endorectal coil used for MR and MR spectroscopic imaging of the prostate. Use of this coil improves the quality of the images and is crucial for spectral acquisition. (B) Coronal T2- weighted image of the prostate obtained without an endorectal coil. (C) Coronal T2-weighted image of the prostate obtained with an endorectal coil, showing the improvement in spatial resolution and noise reduction. TRUS and MR imaging in staging accuracy is cited frequently as evidence that MR imaging has no role in the assessment of prostate cancer [24]. This study was conducted without the incorporation of MR spectroscopic imaging, with sequences and equipment that are obsolete by current standards, and without the use of endorectal coils (Fig. 6). MR imaging and MR spectroscopic imaging can be performed as one combined examination that takes approximately 60 minutes. An endorectal coil is essential for performance of MR spectroscopic imaging, and significantly improves the staging accuracy of MR imaging [25]. It is helpful, but not crucial, to postprocess to compensate for the reception profile of the endorectal coil [26]. T1-weighted MR imaging T1-weighted images of the pelvis aid in the detection of postbiopsy hemorrhage, lymphadenopathy, and bone metastases (Figs. 7 and 8). Approaches to T1-weighted images vary, with some centers only obtaining relatively thick sections from the iliac crests to the symphysis pubis, and other centers also obtaining thin section T1- weighted images through the prostate that correspond to the thin section T2-weighted images. Spin-echo imaging is a reasonable sequence choice because of the low susceptibility to artifact as compared with gradient-echo sequences. The authors protocol uses axial spin-echo T1-weighted images obtained from the aortic bifurcation to

8 564 A. Rajesh, F.V. Coakley / Magn Reson Imaging Clin N Am 12 (2004) Fig. 7. Axial T1-weighted MR image of the prostate in a patient who had undergone transrectal biopsy recently. A large area of hemorrhage is visible as a region of increased T1 signal intensity (arrow) in the left side of the gland. the symphysis pubis, using the following parameters: repetition time (TR)/echo time (TE) = 700/ 8 milliseconds, slice thickness = 5 mm, interslice gap =1 mm, field of view (FOV) = 24 cm, matrix , frequency direction transverse (to prevent obscuration of pelvic nodes by endorectal coil motion artifact), and one excitation. T2-weighted MR imaging T2-weighted images in the axial and coronal planes are the cornerstone of MR imaging in prostate-cancer evaluation. These images clearly depict the distinction between the peripheral zone and the central gland, allow the detection of tumor as low T2 signal intensity foci in the peripheral zone (Fig. 9), and facilitate staging by demonstration of extracapsular extension and seminal-vesicle invasion. These features usually are assessed best on thin-section axial images, although the coronal plane is helpful for the sextant localization of tumor in the craniocaudad direction as being in the base, midgland, or apex. The coronal plane also is useful in the assessment of seminal-vesicle invasion. The authors protocol uses thin-section high spatial resolution axial, and coronal T2-weighted fast spin-echo images of the prostate and seminal vesicles are obtained using the following parameters: TR/effective TE = 5000/96 milliseconds, echo train length = 16, slice thickness = 3 mm, interslice gap = 0 mm, FOV = 14 cm, matrix , frequency direction anteroposterior (to prevent obscuration of the prostate by endorectal coil motion artifact), and three excitations. Fat saturation does not seem to have either a positive or negative effect on staging accuracy [27], and its use is therefore a matter of preference. Gadolinium-enhanced MR imaging The central gland and peripheral zone enhance homogeneously in the normal prostate, with the central gland enhancing more than the peripheral zone [28 30]. Benign prostatic hyperplasia results in marked inhomogeneity of central gland enhancement [28,29]. Some reports suggest that prostate cancer enhances more rapidly than adjacent peripheral zone tissue, and can be demon- Fig. 8. (A) Axial T1-weighted image of the prostate in a 78-year-old man shows a low T1 signal intensity focus (arrow) in the left superior pubic ramus, suspicious for metastasis. (B) Bone scintigraphy shows increased uptake (arrow) in the left superior pubic ramus, confirming the presence of a metastasis. The bone scan was interpreted initially as showing increased uptake as a result of urinary contamination, and the correct interpretation was rendered only after the MR imaging was performed and both studies were correlated directly.

9 A. Rajesh, F.V. Coakley / Magn Reson Imaging Clin N Am 12 (2004) Fig. 9. Axial T2-weighted image of the prostate, showing a tumor in the left midgland as a focus of reduced signal intensity (arrow). strated on early dynamic contrast-enhanced images [31]. Other studies have shown no significant difference between the enhancement pattern of prostate cancer and noncancerous peripheral zone tissue [32]. In clinical practice, gadolinium-enhanced images have not improved tumor localization or staging [28,29], with the possible exception of better visualization of seminal-vesicle invasion in equivocal cases [28,30]. In general, gadolinium-enhancement is not considered helpful in routine prostate MR imaging. Macromolecular contrast media are in development and may prove more useful because tumor microvessels are permeable to macromolecular contrast molecules and normal microvessels are not (standard small molecular contrast media pass through the endothelium of normal and neoplastic microvessels) [33]. Technique of MR spectroscopic imaging Technical aspects of MR spectroscopic imaging MR imaging uses strong magnetic fields to induce coherent spinning of hydrogen protons, and then applies radiofrequency pulses to generate a map of proton signal intensity by spatial location. The signal intensity of all hydrogen protons is combined, although the signals from hydrogen protons in different molecules have slightly different frequencies (a property known as chemical shift). MR spectroscopic imaging exploits the chemical-shift property to produce a map of signal intensity versus frequency (ie, a spectrum) and spatial location. The x and y axes of the spectral trace from an individual voxel represent frequency and intensity, respectively. By convention, the x axis is plotted as the downward frequency shift relative to water expressed in ppm (this denominator adjusts for magnetic field strength, so the x-axis units are fixed irrespective of the type of MR-imaging scanner used). Chemicals with greater degrees of shift are plotted further to the left, and vice versa. The metabolic peaks relevant to prostate MR spectroscopic imaging are choline, creatine, and citrate, occurring at shifts of approximately 3.2, 3.0, and 2.6 ppm, respectively. MR spectroscopic imaging of prostate cancer is characterized by raised choline (a normal cell-membrane constituent that is elevated in many tumors), reduced citrate (a constituent of normal prostatic tissue), or both (Fig. 10) [34]. The ratio of choline and creatine to citrate in sextants with normal prostatic tissue has been established as (John Kurhanewicz, personal communication, 2003). The spectral peaks of creatine and choline often overlap, and may be inseparable. Inclusion of creatine in this ratio is not considered a potential source of error because creatine seems to remain at a relatively constant level in healthy and cancerous prostatic tissue. The y axis lacks absolute units. Each spectrum is derived from a small volume of tissue known as a voxel. Several mechanisms have been developed for overlaying the spectral information from the voxels on the corresponding anatomic images, including overlaid coded grids and color overlays. The ratio of choline and creatine to citrate is related to the Gleason score of the tumor (Fig. 11) [35]. No other imaging modality can provide such a direct evaluation of tumor aggressiveness. The use of MR spectroscopic imaging to detect tumor metabolism is conceptually appealing, and in neuroimaging virtual biopsy refers to the ability of MR spectroscopic imaging to provide noninvasive tissue characterization of brain tumors [36]. Such optimistic terminology would not be appropriate for prostatic MR spectroscopic imaging, which is constrained by substantial technical hurdles and a relatively limited number of validation studies. Some of the technical challenges in obtaining high-quality MR spectra of the prostate can be appreciated by considering the resolution required in the x and y axes. Citrate protons spin with a frequency that is 2.6 Hz per T less than water protons, which spin with a frequency of 42.6 MHz (ie, 42,600,000 Hz) per T. The concentration of metabolites detected at MR spectroscopic imaging is 1 to 10 mmol/l, which is approximately

10 566 A. Rajesh, F.V. Coakley / Magn Reson Imaging Clin N Am 12 (2004) Fig. 10. (A) MR spectrum from a voxel of normal peripheral zone tissue. The identifiable peaks are choline, creatine, and citrate. The polyamine peak cannot be distinguished at 1.5 T, but accounts for the blurring or filling in of the dip between choline and creatine. (B) MR spectrum from a voxel of prostate cancer in the peripheral zone. When compared with the normal spectrum in (A), the choline peak is elevated, the citrate peak is absent, and the polyamine peak is reduced (seen as a deepening of the dip between choline and creatine). 10,000 to 100,000 times less than the molar concentration of water protons. For these and other reasons, the possible sampling volume that can be interrogated with current MR spectroscopic imaging technology is small, and the voxels are relatively large. For example, the authors standard endorectal MR spectroscopic imaging protocol has a voxel size of 0.34 cm 3. A spherical tumor must be at least 0.66 cm 3 in size to fill complexly a cm 3 voxel, assuming the ideal scenario of tumor and voxel having the same geometric center. Otherwise, incomplete filling of a voxel by tumor may result in partial volume artifact. Combined MR imaging and MR spectroscopic imaging of the prostate can be performed in less than 1 hour using a standard clinical 1.5T MR imaging scanner and commercially available endorectal coils [37 39]. An endorectal coil is essential for performance of spectroscopy, and improves the MR imaging component, resulting in higher staging accuracy [25]. The total examination time includes coil placement, patient positioning, and MR imaging and MR spectroscopic imaging data acquisition. Several vendors are offering or are close to releasing product versions of this combined MR imaging and MR spectroscopic imaging examination. Spectral acquisition and processing After review of the axial T2-weighted images, a spectroscopic imaging volume is selected to maximize coverage of the prostate and minimize the inclusion of periprostatic fat and rectal air. Three-dimensional (3D) MR spectroscopic imaging data are acquired using a water- and lipidsuppressed double-spin echo point-resolved spectroscopy sequence optimized for the quantitative detection of choline and citrate. Water and lipid suppression is achieved using the band-selective inversion with gradient dephasing technique [40]. Susceptibility artifacts from periprostatic fat and rectal air are eliminated using outer voxel saturation pulses [41]. Data sets are acquired as phase-encoded spectral arrays (1024 voxels) with a nominal spectral resolution of 0.24 cm 3 to 0.34 cm 3, TR/TE = 1000/130 milliseconds, and 17-minute acquisition time. The 3D-MR spectroscopic imaging data are processed on an UltraSparc workstation (Sun Microsystems, Moutainview, California) using research software previously developed specifically for 3D-MR spectroscopic imaging studies. All spectral data are apodized with a 1-Hz Gaussian function. Data are Fourier-transformed in the time domain and three spatial domains. Corrections to phase,

11 A. Rajesh, F.V. Coakley / Magn Reson Imaging Clin N Am 12 (2004) Fig. 11. (A) Coronal T2-weighted image of the prostate, showing the spectral grid from which the spectra in (B) were acquired. Asymmetric low T2 signal intensity, suggestive of tumor, is visible throughout the right side of the prostate. (B) Spectra from the grid shown in (A) show marked abnormality (elevated choline and reduced or absent citrate) throughout the left side of the prostate (outlined area). Such marked metabolic abnormalities are associated with more aggressive cancers, and biopsy confirmed the presence of Gleason score 9 adenocarcinoma. baseline, and frequency are performed using research software. Spectral evaluation Spectra are interpreted and scored based on prior research and current understanding of prostate cancer metabolism [39]. The level of citrate is characteristically high in normal prostatic tissue [42,43]. Citrate levels fall in prostate cancer, but also can be reduced by prostatitis or postbiopsy hemorrhage [44]. The level of choline, a cell-membrane constituent, is increased in prostate cancer as a result of increased membrane turnover [39,45]. The level of polyamines recently has been recognized to decrease in prostate cancer [46,47]. The polyamine peak occurs between the creatine and choline peaks, and currently cannot be resolved entirely from these peaks. Decreased polyamine levels can be recognized subjectively as a sharper separation of the creatine and choline peaks, however [39]. Based on these considerations, the levels of creatine, choline, and citrate are integrated to determine their relative concentrations. Peak area ratios of choline plus creatine to citrate, citrate to normal citrate, and choline to creatine are calculated. The estimation of choline-to-creatine peak area ratio is only possible in regions of cancer because in healthy voxels there is poor in vivo resolution of these species as a result of the presence of a large polyamine resonance. Spectroscopic voxels are scored on a standardized fivepoint scale, using the following criteria: 1. A primary score of 1 to 5 is assigned based on mean healthy value ratios of choline plus creatine to citrate. A score of 1 is assigned to voxels with a choline plus creatine to-citrate ratio within one standard deviation of the mean healthy value. A score of 2 is assigned to voxels with a choline plus creatine to-citrate ratio between one and two standard deviations above the mean healthy value. A score of 3 is assigned to voxels with a choline plus

12 568 A. Rajesh, F.V. Coakley / Magn Reson Imaging Clin N Am 12 (2004) creatine to-citrate ratio between two and three standard deviations above the mean healthy value. A score of 4 is assigned to voxels with a choline plus creatine to-citrate ratio between three and four standard deviations above the mean healthy value. A score of 5 is assigned to voxels with a choline plus creatine to-citrate ratio more than four standard deviations above the mean healthy value. 2. An initial adjustment is made to the primary score to incorporate elevation of choline relative to creatine and reduction in polyamines. When the choline-to-creatine ratio is greater than or equal to 2 with a primary score of 2 or 3, the overall score is increased to a 4. When the choline-to-creatine ratio is less then 2 or there was no reduction in polyamines with a primary score of 4 or 5, the overall score is decreased by 1 (ie, 3 and 4). 3. A final adjustment is made to the score to incorporate poor voxel signal-to-noise ratio. Poor signal-to-noise ratio is defined as a ratio of less than 8 for voxels with a score of 3 to 5, and less than 5 for voxels with a score of 1 to 2. In the presence of poor signal-to-noise ratio, a score of 1 becomes 3, scores of 2 or 4 become 3, and scores of 5 become 4. Scores of 3 are not changed by low signal-to-noise criteria. This standardized scoring system results in a final score from 1 to 5, and is designed so that the following interpretative scale can be applied: 1 is considered probably benign, 2 is possibly benign, 3 is equivocal, 4 is possibly malignant, and 5 is probably malignant. In addition to the 5- point scoring system, readers are allowed to designate spectra as unusable if they showed significant lipid contamination or misalignment of metabolite resonance peaks. This five-point scale distinguishes benign and malignant tissue with reasonable accuracy [48]. Applications of MR imaging and MR spectroscopic imaging Role of MR imaging and MR spectroscopic imaging in prostate-cancer diagnosis MR imaging findings in prostate cancer first were described in the early 1980s [49]. These early reports were uncertain about the exact MR imaging appearance of prostate cancer because of the limitations in contrast and spatial resolution of the MR imaging technology available at that time. Further research established that prostate cancer is characterized by low T2 signal intensity in the normally high T2 signal intensity peripheral zone [50]. The presence of reduced T2 signal intensity in the peripheral zone is of limited sensitivity, however, presumably because some tumors are isointense (Fig. 12) [51,52]. MR spectroscopic imaging also may show no convincing metabolic abnormality in some patients (see Fig. 12). The finding is also of limited specificity because there are other causes of low T2 signal intensity in the peripheral zone, such as hemorrhage, prostatitis, scarring, radiotherapy, cryosurgery, and hormonal therapy (Figs. 13 and 14). Only a few studies have investigated the accuracy of MR imaging in the diagnosis of prostate cancer because MR imaging generally is reserved as a staging study in patients with biopsy-proven prostate cancer. In a study of 33 patients with an elevated PSA and at least one negative sextant biopsy before MR imaging [53], two readers rated the likelihood of malignancy as low, intermediate, or high. Repeat biopsy was positive in 1 of 18 patients considered low likelihood, one of eight patients considered intermediate likelihood, and five of seven patients considered high likelihood. A more recent similar study of 38 patients with prior negative biopsies, 12 of whom had a positive post MR imaging biopsy, demonstrated a sensitivity of 83% and a positive predictive value of 50% for the MRimaging diagnosis of prostate cancer [54]. These results suggest MR imaging can be used in this setting to stratify patients with a high and low probability of a subsequent positive biopsy. The potential incremental benefit of MR spectroscopic imaging in this setting has not been reported, but based on the results for tumor localization, one would expect a positive impact. The use of MR imaging as screening tool in the general population would result in an unacceptably large number of false positive and negative results. This would also be impractical for logistical and financial reasons. Role of MR imaging and MR spectroscopic imaging in prostate cancer localization Sextant biopsy traditionally has been regarded as the standard of reference for nonsurgical tumor localization [55]. The limitations of sextant biopsy are being recognized increasingly, however [56,57]. In a study with two readers using

13 A. Rajesh, F.V. Coakley / Magn Reson Imaging Clin N Am 12 (2004) Fig. 12. (A) Axial T2-weighted image of the prostate, showing the spectral grid from which the spectra in (B) were acquired. No convincing mass-like foci of low T2 signal intensity are visible. (B) Spectra from the grid shown in (A) demonstrate no convincing metabolic abnormality in the peripheral zone. (C) Corresponding whole mount step section pathologic preparation with areas of cancer outlined. A large tumor is present in the right side of the prostate, despite the lack of findings at MR and MR spectroscopic imaging.

14 570 A. Rajesh, F.V. Coakley / Magn Reson Imaging Clin N Am 12 (2004) Fig. 13. Axial T2-weighted image of the prostate in a patient who has received hormonal therapy. The prostate is small, of uniformly low T2 signal intensity, and relatively featureless (compare to Fig. 5). step-section histopathology as the standard of reference in 53 patients [58], MR imaging alone had a sensitivity of 77% to 81% and a specificity of 46% to 61% for the sextant localization of prostate cancer. The addition of MR spectroscopic imaging reduced the sensitivity slightly to 68% to 73%, but specificity increased substantially to 70% to 80%. In another study of 47 patients using step-section histopathologic analysis of radical prostatectomy specimens as the standard of reference [59], the separate and combined accuracy of preoperative sextant biopsy, MR imaging, and MR spectroscopic imaging were compared. When all three tests were positive for cancer in a sextant, sensitivity was 33% and specificity was 98%. Conversely, a single positive test had a sensitivity of 94% and a specificity of 38%. This suggests that when correct tumor localization is crucial, only sextants that are positive at biopsy, MR imaging, and MR spectroscopic imaging should be considered as definitely cancer-containing. The results described above refer to the sextant localization of prostate cancer, which is not synonymous with volumetric localization. In a recent study [60], MR imaging and MR spectroscopic imaging were performed in 37 patients before radical prostatectomy. Two independent readers recorded peripheral-zone tumor-nodule location and volume and results were analyzed using step-section histopathologic tumor volumetry as the standard of reference. The mean volume of all peripheral-zone tumor nodules (n = 51) was 0.79 cm 3 (range cm 3 ). Readers detected 20 (65%) and 23 (74%) of the 31 peripheral-zone tumor nodules greater than 0.5 cm 3. For these nodules, tumor volume measurements by MR imaging alone and combined MR imaging and MR spectroscopic imaging all were correlated positively with histopathologic volume (Pearson s correlation coefficients of 0.49 and 0.55, respectively), but only measurements by combined MR imaging and MR spectroscopic imaging reached statistical significance (P <.05). When all nodules were analyzed, tumor volume measurements were correlated poorly and not statistically significantly with histopathologic tumor volume (Table 4). These results for prostate cancer tumor volume measurement may seem disappointing, particularly in the context of other studies indicating high accuracy for sextant localization. Two factors probably account for this discrepancy. First, because per-sextant rather than per-nodule analysis does not require size concordance between imaging and pathology, a small imaging abnormality counts as a true positive even if the tumor is pathologically much larger, and vice versa. Second, there has been a general downward stage migration of prostate cancer because of widespread PSA testing. For example, the rate of organ-confined disease in patients undergoing prostatectomy and MR imaging at the University of California, San Francisco between 1992 and 1995 was 56% (43 of 77) [61], compared with 89% (33 of 37) in 1999 [60]. High accuracy is difficult to achieve when imaging smaller tumors. Role of MR imaging and MR spectroscopic imaging in prostate cancer staging From the inception of prostate MR imaging, the hope has been that the modality would be more accurate in local staging and detection of extraprostatic disease. Disappointingly, an early multi-institutional study examining the detection of extracapsular extension showed no difference in the area under the receiver operating characteristic curve for MR imaging (0.67) compared with TRUS (0.62) [4]. Single-institution studies since then have shown higher overall staging accuracies of 86% to 88% [62,63], which probably reflects improved technology and better diagnostic criteria. Multivariate feature analysis has shown the MR imaging findings that are most predictive of extracapsular extension are a focal irregular capsular bulge, asymmetry or invasion of the neurovascular bundles, and obliteration of the rectoprostatic angle (Fig. 15) [61]. The addition of MR spectroscopic imaging to MR imaging

15 A. Rajesh, F.V. Coakley / Magn Reson Imaging Clin N Am 12 (2004) Fig. 14. (A) Axial T2-weighted image of the prostate shows foci of reduced T2 signal intensity (arrows) in the peripheral zone bilaterally. (B) Axial T2-weighted image of the prostate, showing the spectral grid from which the spectra in (C) were acquired. (C) Spectra from the grid shown in (B) show several voxels with elevated choline in the left gland (outlined area), but no abnormal metabolism elsewhere. Biopsy confirmed Gleason score 7 adenocarcinoma in the left midgland and apex. increases staging accuracy for less-experienced readers and reduces interobserver variability [63]. The value of MR imaging staging has been demonstrated in a 5-year follow-up study of 1025 men who were staged radiologically before prostatectomy [64]. The MR-imaging detection of extracapsular extension conferred a significantly worse prognosis in the 39% of patients with moderate or high-risk tumors. Similarly, a decision analysis model suggested that preoperative MR imaging was cost-effective for men with moderate or high probability of extracapsular disease [65]. Reader variation in the interpretation of prostate MR imaging remains a barrier to greater acceptance of this technique for preoperative staging [66,67]. Radiologists interpreting prostate MR imaging should be aware that high specificity, even if accompanied by low sensitivity, is a more

16 572 A. Rajesh, F.V. Coakley / Magn Reson Imaging Clin N Am 12 (2004) Table 4 Correlation between MR imaging and MR spectroscopic imaging measurement of peripheral zone tumor volume and histopathologic tumor volume in a study of 37 men with 51 proven peripheral zone tumor nodules at radical prostatectomy [60] All nodules Nodules > 0.5 cm 3 Method of tumor volume measurement Correlation coefficient (95% confidence interval) P value Correlation coefficient (95% confidence interval) P value MR imaging alone 0.21 (ÿ0.22, 0.54) (ÿ0.06, 0.80) 0.07 MR imaging and MRS imaging 0.32 (ÿ0.22, 0.65) (ÿ0.04, 0.82) 0.04 Only MR and MRS imaging volume measurement of tumors >0.5 cm 3 showed a statistically significant correlation with histopathologic volume. cost-effective approach in patients being considered for surgery [68]. Extracapsular extension generally is reported as a simple binary observation, either present or absent. This is an oversimplification because extracapsular extension is a continuum that varies in degree. It is instructive to consider how this affects prognosis and imaging accuracy. The prognostic importance of the degree of extracapsular extension was evaluated in a 10-year followup study of 617 men with T1 to T3A node-negative prostate cancer after radical prostatectomy [69]. The 10-year risk for recurrence (clinical, radiologic, or biochemical) was 32% in patients with microscopic extracapsular extension (the presence of no more than a few malignant cells immediately outside the capsule on no more than two sections) compared with 42% in patients with more established extracapsular extension, suggesting microscopic extracapsular extension is nearly as important as macroscopic extracapsular extension with respect to prognosis. This is an important finding because microscopic extracapsular extension is unlikely to be visualized directly by any currently available radiologic test. The degree of extracapsular extension is an important variable that almost certainly affects MR imaging staging accuracy. This was shown directly in one study of 34 patients undergoing endorectal MR imaging before radical prostatectomy [70]. The sensitivity of MR imaging for extracapsular extension of less than 1 mm was only 14% (one of seven sites detected), compared with 71% (five of seven sites detected) for extracapsular extension greater than 1 mm. Role of MR imaging and MR spectroscopic imaging in treatment planning Anecdotally, MR imaging and MR spectroscopic imaging results may be of value to patients with prostate cancer deciding whether to undergo Fig. 15. (A) Axial T2-weighted image showing an irregular capsular bulge at the left base associated with an underlying region of low T2 signal intensity (arrow). The findings are consistent with extracapsular extension of tumor (T3A disease). (B) Axial T2-weighted image showing asymmetric low T2 signal intensity in the left seminal vesicle (arrow), consistent with seminal vesicle invasion (T3B disease).

17 A. Rajesh, F.V. Coakley / Magn Reson Imaging Clin N Am 12 (2004) Fig. 16. Coronal T2-weighted image in a 58-year-old man with a PSA of 9.2 and biopsy-proven Gleason 7 cancer. The tumor was considered organ-confined on a DRE. MR imaging shows asymmetric low T2 signal intensity extending into the left seminal vesicle (arrow), consistent with seminal vesicle invasion. In view of this finding, the patient chose radiation treatment rather than radical prostatectomy. surgery or radiation therapy (Fig. 16), although the influence of MR imaging and MR spectroscopic imaging on such decision-making is difficult to quantify in systematic studies. The display of the tumor relationship to the neurovascular bundles on MR imaging is helpful to the urologist in planning the surgical approach, although this has not been validated scientifically. The relationship between periprostatic anatomy and operative outcomes has been investigated. In a study of 211 consecutive patients with newly diagnosed prostate cancer undergoing radical prostatectomy performed by a single surgeon [71], membranous urethral length was measured on preoperative endorectal MR imaging (Fig. 17) and correlated with postoperative urinary continence. After controlling for age and surgical technique, multivariate analysis showed that membranous urethral length was related to the time to stable postoperative continence (P =.02), such that a longer membranous urethra was associated with a shorter time to stable continence. In another study, MR imaging was performed in 143 patients with newly diagnosed prostate cancer before radical prostatectomy [72]. Two independent readers rated the prominence of the periprostatic veins (based on number and size) at four anatomic sites on a 3-point scale. Prominence of the anterior and posterior apical periprostatic veins was associated positively with Fig. 17. Coronal T2-weighted image of the prostate, demonstrating measurements of the membranous urethral length. blood loss (correlation coefficients of 0.22 and 0.17, P <.01 and.05, respectively). Several possible roles for MR imaging and MR spectroscopic imaging have been suggested in radiation treatment planning. Using fiducial markers and image fusion software, it has been shown that CT overestimates the clinical target volume by 34% when compared with MR imaging [73], and that dose-volume planning with MR imaging would decrease radiation dose to the bladder, rectum, and femoral heads. This result is hardly unexpected because the radiation-planning CT scans are performed without intravenous contrast, which limits soft-tissue contrast of CT in the prostate and perineum (Fig. 18). Several groups have reported using MR spectroscopic imaging to increase the brachytherapy radiation dose in prostatic locations considered suspicious for cancer [74,75]. Such studies, which suggest technically successful dose escalation in spectroscopically suspicious locations implies improved clinical outcome, must be viewed with caution, given the limited ability of MR imaging and MR spectroscopic imaging to assess tumor volume. More convincing evidence of benefit was shown in a study of 390 patients with prostate cancer treated by brachytherapy, either alone (46%) or in combination with external beam radiotherapy (54%). MR imaging was used to evaluate stage and guide therapy in 327 patients [76]. MR imaging findings changed the overall treatment recommendation in 60 patients with most of these

18 574 A. Rajesh, F.V. Coakley / Magn Reson Imaging Clin N Am 12 (2004) Fig. 18. (A) Axial CT image in a 72-year-old man with recurrent tumor after radical prostatectomy. No obvious abnormality is appreciable because the soft-tissue contrast of CT is limited, particularly in the distinction of perineal structures. (B) Axial T2-weighted MR image in the same patient shows a large mass (arrows) in the prostatectomy bed. patients receiving combined therapy instead of monotherapy after MR imaging documented more extensive disease. Seed distribution was modified in 183 patients, mostly related to coverage of bulky or extracapsular disease seen on MR imaging. Freedom from PSA progression at a mean followup of 38 months was used as the endpoint. Cox regression analysis showed that only the percentage of positive cores (P =.001) and failure to have MR imaging staging (P =.0008) predicted for failure. This suggests that MR imaging improves treatment planning in terms of technical success and with respect to clinical outcome. Role of MR imaging and MR spectroscopic imaging in posttreatment follow-up The role of MR imaging and MR spectroscopic imaging in posttreatment follow-up of patients with prostate cancer is not well established, partially because it is frequently clinically unclear whether patients with a rising PSA after treatment have local or distant recurrence. In the authors experience, MR imaging and MR spectroscopic imaging are of limited utility after prostatectomy, and only the occasional patient demonstrates an unequivocal locally recurrent tumor (Fig.19). The ability of MR imaging to detect local recurrence is limited after radiation or hormonal therapy because the prostate becomes shrunken with diffusely low T2 signal intensity and indistinct zonal anatomy [77,78]. MR spectroscopic imaging may be of value in the detection or exclusion of locally recurrent prostate cancer after radiation therapy. In a preliminary study of 21 patients with biochemical failure after external beam radiation therapy for prostate cancer in whom subsequent biopsy confirmed locally recurrent prostate cancer in nine hemiprostates of six patients, Teh and colleagues [79] found that the presence of three or more MR spectroscopic imaging voxels with isolated (ie, absent citrate) elevation of choline showed a sensitivity and specificity of 87% and 72%, respectively. Conversely, the presence of complete metabolic atrophy demonstrated a negative predictive value of 100% for the exclusion of local recurrence (Fig. 20); that is, a negative spectroscopic study in a patient with a rising PSA after radiotherapy may be a presumptive marker of distant failure, and indicate that systemic rather than local therapy is required. Fig. 19. Sagittal T2-weighted image in an 83-year-old man with a rising PSA 8 years after a radical prostatectomy. The surgical specimen demonstrated T3B cancer. A large mass of recurrent tumor is visible in the prostatectomy bed.

19 A. Rajesh, F.V. Coakley / Magn Reson Imaging Clin N Am 12 (2004) Fig. 20. (A) Axial T2-weighted image of the prostate, showing the spectral grid from which the spectra in (B) were acquired. The patient had a rising PSA after receiving external beam radiation for prostate cancer 2 years previously. (B) Spectra from the grid in (A) show random noise, with no detectable metabolites in the prostate. The finding of complete metabolic atrophy in an irradiated gland is consistent with satisfactory local tumor control, and therefore suggests that the rising PSA is a result of distant failure. Summary The primary indication for prostate MR imaging and MR spectroscopic imaging is the evaluation of men with newly diagnosed prostate cancer with a moderate or high risk for extracapsular extension who are uncertain whether to go undergo surgery or radiotherapy. Other applications of MR imaging and MR spectroscopic imaging in prostate cancer are under investigation and are yet to be defined fully. Areas of active research interest include volumetric localization of prostate cancer, in vivo MR spectroscopic imaging findings at high field strength (3 T), in vitro MR spectroscopic imaging findings at very high field strength (7 11 T), novel spectroscopic markers of malignancy such as polyamines and spermine, and interventional MR guidance of biopsy and therapy [39,80,81]. MR spectroscopic