IN THE DIAGNOSIS of prostate cancer (PCa) parameters such as location, TNM stage, aggressiveness (represented by the Gleason score), and volume estimation are essential to both treatment choice as well as its success. Urologists base their everyday decision-making on nomograms that have incorporated these parameters (1).
However, the generally used diagnostic methods, like digital rectal examination (DRE), prostate-specific antigen (PSA) level, and gray scale transrectal ultrasound (TRUS)-guided biopsy, which are the input for the nomograms, are imperfect. DRE misses a substantial proportion of cancers and identifies predominantly tumors when they are large and in a more advanced pathologic stage (2, 3). PSA is considered the most useful tumor marker for diagnosis, staging, and monitoring of PCa. It correlates well with advanced clinical and pathological stages in most cases, but due to low specificity it cannot accurately stage an individual patient, as there is large overlap between different tumor stages (4). Shortcomings of gray-scale TRUS-guided prostate biopsy are an inherently low sensitivity for detection of PCa (5), a high false-negative biopsy rate, and over- or underestimating of true Gleason score (1, 6). More rigorous biopsy schemes such as 3D template-guided transperineal saturation biopsies, up to 80 cores, have been shown to augment cancer detection rates (7–9). Nevertheless, such drastic biopsy methods have many drawbacks, like patient discomfort, the need for anesthetics, and intensive pathologic processing.
In summary, PSA, DRE, and TRUS-guided biopsy do not have the ability to correctly localize, stage, determine volume, and aggressiveness of PCa. In the new era of more precise therapy forms such as intensity-modulated radiation therapy (IMRT) and focal ablative therapies (eg, cryosurgery, high-intensity-focused ultrasound, thermal therapy, and radiofrequency ablation), this information is of utmost importance.
The ideal PCa assessment tool should be able to gather this information in a noninvasive way. This information can then serve therapy choice, guidance of interventions, and treatments. Multiparametric magnetic resonance imaging (MRI) has the potential of being such a tool.
The purpose of this review is to discuss the potential role of multiparametric MRI in focal therapy with respect to patient selection and directing (robot-guided) biopsies and IMRT.
Screening, Detection, and Localization
MRI has been applied in the assessment of PCa for multiple purposes. It has been used for screening, detection in patients with elevated PSA level, and multiple negative TRUS-guided biopsy sessions, recurrent disease, localization, and staging local as well as distant.
The reported diagnostic performance of T2-weighted (T2-w) imaging in the detection and localization of PCa has a wide range, with sensitivities and specificities between 47.8%–88.2% and 44.3%–81%, respectively (10–14). Therefore, the incremental value of multiparametric MRI such as dynamic contrast-enhanced (DCE-MRI), diffusion-weighted imaging (DWI), and MR spectroscopic imaging (MRSI) have been investigated. Studies reporting on the combination of these techniques describe the additional value in diagnostic performance (11, 12, 14–20). Futterer et al (14) reported an area under the curve (AUC) of 0.90 when T2-w, DCE-MRI, and MRSI were combined in localizing PCa. It is recommended to perform multiparametric MRI consisting of at least T2-w, DCE-MRI, and DWI. However, the lack of consensus in imaging protocols (eg, with/without endorectal coil, field strengths, b-values, postprocessing methods) makes defining definite guidelines troublesome. Hence, the next focus point in future studies should be optimizing these protocols and reaching consensus.
T2-w, endorectal coil imaging, according to a meta-analysis by Engelbrecht et al (21), reported a maximum combined sensitivity and specificity of 71% in distinguishing clinical stages T2 and T3 at a 1.5T MR system (18). More recent studies report maximum sensitivities and specificities of 80%–88% and 96%–100%, respectively, at a 3T MR system (22, 23). Adding metabolic information obtained with MRSI to T2-w imaging improves staging accuracies (24). A combination of this information with nomograms, used in daily clinical practice by urologists, in evaluating extracapsular extension and seminal vesicle invasion in staging PCa yields higher accuracies (25–27). In less experienced readers of MRI in staging PCa, DCE-MRI also appears to be of added value (28). Presumably, the extra information collected with DCE-MRI, DWI, and MRSI draws the attention of the reader more than T2-w imaging alone to areas where extracapsular extension could be the case (Fig. 1).
Assessment of skeletal metastases is mostly done by performing a technetium-99m-diphosphonate bone scintigraphy. Although bone scintigraphy is considered the gold standard for skeletal imaging, its low positive yield makes its use controversial (29). Several studies have evaluated the role of this technique compared to MRI in detecting skeletal metastases (30–33), but only one study conducted this in a patient group with only PCa as primary cancers (33). According to this study, MRI when used alone is superior in detecting bone metastases with a sensitivity and specificity of 100% and 88%, respectively, compared to bone scintigraphy combined with targeted x-rays with a sensitivity and specificity of 63% and 64%. However, the highest specificity of 100% (at the cost of a lower sensitivity of 88%) was reported when bone scanning together with targeted x-rays and MRI was performed. Larger trials can possibly clarify what the best approach in clinical practice should be. Noteworthy, limiting MRI to the axial skeleton will suffice, as the probability of finding solitary metastases outside this area is negligible (34, 35). New insights for DWI concerning skeletal metastases could possibly lead to even higher detection rates (Fig. 2) (36–38).
Computed tomography (CT) and MRI demonstrate an equally poor performance in the detection of lymph node metastases from PCa, especially concerning sensitivity (39). For this reason two other MR techniques have been developed: MR lymphography (MRL) (which uses a lymph node-specific contrast agent, based on ultrasmall superparamagnetic particles of iron oxide [USPIO]) and DWI-MRI (Fig. 3).
In 1998, Bellin et al (40) reported on the initial clinical experience with MRL and found a perfect sensitivity of 100% at 80% specificity. In another prospective study with 334 lymph nodes in 80 patients, sensitivity and specificity were 90.5% and 97.8%, respectively (41). More recently, it has been shown that MRL is significantly more accurate than multidetector-row CT (42), and that in 41% of PCa patients MRL can detect lymph node metastases outside the surgical area of routine pelvic lymph node dissection (43). Although these results are very promising, MRL has not yet become available for widespread clinical use due to the lack of an U.S. Food and Drug Administration (FDA)-approved lymph node-specific contrast agent.
The added value of DWI compared to USPIO-MRL does not improve accuracy, but rather reduces reading time (44). However, one study reported a good accuracy based on apparent diffusion coefficient (ADC) value alone, with a sensitivity of 86.0% and a specificity of 85.3% (45).
For focal therapy, merely localizing PCa is not enough. Determining cancer foci volume is of paramount importance for targeting this form of therapy. Studies regarding this matter are few in number and not consistent (46–49). In 2002 the value of MRSI in measuring tumor volume in nodules greater than 0.5 cm3 was assessed and researchers found that it was positively correlated with histopathologic tumor volume, with a Pearson correlation coefficient of 0.59 (50). More recently, DCE-MRI and DWI have as well been investigated and some promising results have been presented (51, 52).
The Gleason grading system remains one of the most effective prognostic factors in PCa (53). Gleason score, PSA level, and clinical stage have a major role in therapy decision-making. They have been associated with biochemical failure, local recurrences, and distant metastases such as skeletal and lymph node metastasis after prostatectomy or radiation therapy (54–58). Since Gleason scores of 3+4, or lower, are associated with lower disease progression rates, and Gleason scores of 4+3, or higher, are associated with higher disease progression rates (59), a test differentiating between both is meaningful.
Although studies reporting an association of Gleason grade with MRI are scarce (60), with DWI a significant negative correlation between Gleason grade and ADC values has been found (61–63). Furthermore, choline plus creatine-to-citrate ratios determined by using MRSI have also been correlated with Gleason grade (64, 65). One study even reported on the correlation of signal intensities on T2-w imaging with Gleason grade (66). More research needs to be done with regard to what role multiparametric MRI can play in the assessment of PCa aggressiveness.
Recurrence of Local PCa
The majority of patients with PCa are either treated with radical prostatectomy (RP) or radiotherapy (RT). Approximately 30% will experience recurrent disease (67). Currently, PSA elevation after RP or RT is the best indicator of biologically active tumor (68, 69). Whenever such an elevation of PSA has taken place, the main objective is to find out whether the increase in PSA is due to local or distant recurrent disease. Imaging is required as further detailed information about the site of recurrence is needed.
After RT the prostate will be characterized by a diffuse low signal (70, 71). For this reason, T2-w imaging is less helpful in localizing recurrence of PCa. Gadolinium-based contrast agents are of added value. Studies evaluating DCE-MRI after RT report sensitivities between 70%–74% and a specificities between 73%–85% (72, 73). Subsequently, a targeted MR-guided biopsy can be performed (Fig. 4) (74). The incremental value of MRSI in localizing PCa after RT has also shown to be useful (75, 76). Although DWI after radiotherapy can be supportive (77, 78), after RP it is of limited value because of surgical clips (left behind by the urologist during surgery) causing magnetic field inhomogeneities. After RP, DCE-MRI and MRSI provide promising results (Fig. 5) (79–81).
Computer-assisted technologies have been developed to help interpret the often multidimensional and multiparametric data. In mammographic detection of breast cancer such aids have been successfully developed and have become part of routine clinical work (82). Applications to the prostate are emerging.
Two types of computer-assisted technologies exist: computer-aided detection (CAD) and computer-aided diagnosis (CADx). CAD provides fully automatic feedback on zero or more locations containing malignant cancers. CAD has been used in mammographic screening, whereas CADx, often confused with CAD, only provides a probability of malignancy of a user-identified lesion. CAD and CADx both contain a supervised pattern recognition module trained on annotated cases. Although companies provide prostate MR-specific tools with the name CAD in it, neither CAD nor CADx is currently commercially available for prostate MR.
CADx research for prostate is still limited. The main focus in CADx is to have objective features or thresholds for pattern recognition system of PCa. Studies using simple raw image derived methods report lower AUCs of 0.72 (83–85) than studies using quantitative features (AUCs of 0.81) (86, 87). Combining quantitative features from all available sequences (multiparametric MRI) is currently lacking, but first attempts are promising (AUC of 0.89) (88, 89). An important obstruction for prostate MR CAD and CADx to become widely available is the lack of standardized sequences and objective quantitative features of PCa.
Furthermore, without evidence of effectiveness of CAD or CADx in clinical situations, clinical use should be avoided.
INTRAPROSTATIC HIGH DOSE BOOSTING WITH INTENSITY-MODULATED RADIOTHERAPY
In the treatment of PCa, RT has evolved from the use of rectangular beams to 3D conformal beams, and eventually to segmentation and dose intensity-modulation of these beams known as IMRT. IMRT allows the radiation oncologist to plan a precise dose distribution in the targeted organ. Potentially, this results in a high-dose escalation in the targeted cancer foci (dominant intraprostatic lesion [DIL]), while lowering treatment-induced toxicity to noncancerous prostate tissue and organs that abut the prostate, and thus expectantly improving the therapeutic ratio (90–95). This is achieved by using an inverse treatment planning method consisting of defining the desired dose distribution to the target volume and dose restriction to uninvolved tissue or organs (96).
For such a treatment planning strategy, a reliable imaging technique is essential. CT is probably the most frequently used treatment planning technique. CT has some benefits over MRI as it is helpful in RT dose calculations, because of the accurate electron densities provided by it, and its low geometric distortion. Some promising efforts have been made to overcome both matters concerning MRI (97, 98). Still, there is more work to be done in the consideration of performing focal high-dose DIL-IMRT planning with MRI alone. The fault of CT compared to MRI is that it lacks sufficient soft-tissue contrast to precisely delineate the prostate gland, especially in the apex of the prostate, consequently overestimating prostate volume (99–102). Furthermore, CT struggles with significantly higher interobserver variability than MRI (101). In addition to all the foregoing, MRI as opposed to CT has the ability to localize the cancer foci in the prostate (see Localization of PCa, above). Eventually, with a precise dose delivery technique such as DIL-IMRT, one has to bear in mind that its success rate is highly dependent on the accurate delineation of the target volume.
Future developments will probably make it possible to use MRI directly in the treatment planning of RT. Until then we need to combine both techniques and use all assets provided by both involving the method of fusion. By using a gold marker-based 3D fusion of CT and MRI it is feasible to plan a safely delivered intraprostatic high-radiation-dose DIL-IMRT, where a DIL volume, for example, can be defined by using DCE-MRI and MRSI (Figs. 6, 7) (92, 103). With the help of special software programs the merging of MR and CT images can be realized with fusion inaccuracies with a mean of 1.1 mm at the border of the prostate (104). Other efforts in finding new fiducial markers for this fusion process have been made (105).
MR-GUIDED BIOPSY AND ROBOTICS
Cancer suspicious regions (CSRs) seen on multiparametric MRI can be targeted for biopsy. This can be done by either performing a TRUS-guided biopsy, a TRUS-MR fusion-guided biopsy, or an MR-guided biopsy. No studies have been performed comparing these different biopsy methods.
MR-guided prostate biopsies have been shown to improve cancer detection rates in subjects with an elevated PSA and repetitive negative TRUS-guided biopsies (106–108).
So far, the only commercially available MR-guided prostate biopsy device is a manually adjustable standard for needle guide positioning (106, 109). The radiologist has to adjust the needle guide into the direction of the region of interest by hand, based on MR images (109). Consequently, the patient has to be withdrawn from the scanner bore each time to adjust the direction of the needle guide. It is conceivable that during these time-consuming maneuvers there is an abundance of time for movement of the patient as well as movement of the prostate, possibly leading to less accuracy. Furthermore, one can imagine that this procedure is unpleasant for the patient.
For these reasons MR-compatible robots for transrectal prostate biopsy are being developed. Preliminary results found in phantom and patient feasibility studies are promising (110–114). In future studies robotics, for instance, can also play an important role in guiding focal treatment of PCa. But before robot-assisted MRI-guided focal therapy can be realized further extensive research needs to be done.
In conclusion, multiparametric MRI is a versatile and promising technique. It appears to be the best available imaging technique at the moment in localizing, staging (primary as well as recurrent disease, and local as well as distant disease), determining aggressiveness, and volume of PCa. With this information guidance of focal therapy, IMRT and biopsy can be realized.
However, larger study populations in multicenter settings have to confirm these promising results. But before such studies can be performed, more research is needed in order to achieve standardized imaging protocols regarding issues in obtaining data (eg, b-values in DWI, T1-w or T2*-w in DCE-MRI, 1.5T vs. 3T, the use of an endorectal coil) and quantifying and determining objective parameters or thresholds (eg, ADC values in DWI, calculating pharmacokinetic parameters in DCE-MRI) for PCa. As a consequence, supportive techniques such as computer-aided diagnosis can be developed further.