Multiparametric MRI - local staging of prostate cancer and beyond

The Prostate Imaging-Reporting and Data System (PI-RADS) guidelines19 are aimed at standardizing MRI acquisition and interpretation and recommend MRI to be performed at 3T in order to increase signal-to-noise ratio (SNR) and spatial resolution, and decrease acquisition times.20, 21, If acquisition protocols are optimized and contemporary technology is employed, then 1.5T scanners are also able to provide sufficient diagnostic performance. 1.5T scanning may also be preferential when a patient has an implant non-compatible at higher field strengths, or with bilateral hip replacements in order to minimise artefact.22, 23, The routine use of an endorectal coil (ERC) is no longer recommended.19 3T scanners or contemporary 1.5T scanners can provide sufficient imaging quality and although ERC increases SNR bring disadvantages, including deformation of the gland contour, near field coil flare, increased cost and time of examination as well as higher patient discomfort.19, 21,

Standard prostate MRI protocols should incorporate anatomical T1W and T2W imaging in combination with the two functional sequences of diffusion weighted imaging (DWI) and dynamic contrast enhanced imaging (DCE).24 A set of minimal technical parameters for each of these sequences is outlined in Table 225, although institutions are encouraged to optimize imaging protocols based on their own equipment, capacity and expertise. It is mandatory for axial T2W, DWI and DCE to be acquired in the same location, angle, slice thickness and gap to allow for synchronous scrolling through the images and direct evaluation of suspicious findings between the sequences. Axial T1WI is essential to assess post-biopsy haemorrhage, and is typically employed as the sequence to cover the pelvis to the aortic bifurcation to enable bone and nodal assessment.

T2W imaging is the key sequence for local T-staging of the prostate. The high in-plane spatial resolution allows for accurate evaluation of extra-capsular

extension, neurovascular bundle assessment and seminal vesicle invasion. Fast-spin-echo (FSE) or turbo-spin-echo (TSE) imaging should be obtained in the axial plane and in least one additional orthogonal plane (sagittal or coronal) with the highest quality possible and thin slices at 3 mm with no gap.25 3D T2 weighted imaging with isotropic voxels and slice thickness at 1 mm may be obtained, with evidence suggesting utility for assessment of extraprostatic extension26 and for nodal and bone staging when combined with DWI of the entire pelvis (b-values 0–1000 s/mm2).27

Motion artefact Bowel peristalsis is a well-known cause of motion artefact in abdominal imaging,

Figure 1

but the relatively low position of the prostate, remote from small bowel combined with limited evidence prior to PIRADS version 2 means that anti–spasmodic agents are not recommended in current guidelines. However, recent studies have shown use of anti-peristaltic agents significantly improves image quality of T2W imaging28, 29 with better depiction of anatomical details (e.g. prostatic capsule and neurovascular bundles) as well as reducing non-diagnostic MRI to < 1%.29 Routine use of antiperistaltic agents (recommended dose 20 mg HBB i.v or 1 mg glucagon) prior to prostate mp-MRI may therefore be beneficial for optimisation of T2W image quality, a key sequence of mpMRI for local staging.30 The risk of side-effects with these agents is low and these are usually minor and self-limiting.31

Extension of the tumour into the periprostatic fat is defined as T3a disease, termed extracapsular extension (ECE). Of note, however, in a strict sense the prostate lacks a true capsule as an anatomic structure that encloses the gland but has rather an outer fibromuscular layer which is inseparable component of the prostatic stroma.32, 33, T3a disease also incorporates invasion into the neurovascular bundle, internal sphincter and bladder neck.1 Histopathologically, extracapsular extension (ECE) is sub-classified into focal and established with the latter carrying a worse prognosis.34 However, there is currently no clear consensus on the exact definitions of these, which can vary from a few glands beyond the capsule to cancer extending up to 0.5 mm radially from the capsule.1 In addition, focal ECE cannot be detected by MRI due to inherent resolution limits.35

Extracapsular extension has traditionally been evaluated by clinical criteria and nomograms such as Partin tables, which are based on PSA, DRE and Gleason score at biopsy.36 However, nomograms represent a patient level risk score alone, have been shown to be inferior to MRI11, 37,, and unlike MRI offer no information on location and extent of ECE. A meta-analysis by de Rooij et al. in 2016 showed MRI to have a high specificity of 91% but only moderate sensitivity at 57% in diagnosing ECE. Of note, this analysis included studies with both uni- and multi-parametric protocols at both 1.5T and 3T, and sub-analysis of 3T studies improved overall performance with specificity 86% and sensitivity 68%.38 The main reasons for improvement being higher spatial resolution at 3T and improved lesion identification with a multiparametric approach, allowing interrogation of the capsule and neurovascular bundle in the adjacent region (Figure 2).39

Figure 2

Several approaches have been proposed and explored in order to increase diagnostic accuracy for the evaluation of ECE. Prostate imaging-reporting and data system (PI-RADS) guidelines recommends various morphologic criteria (Table 3).19 These have been evaluated and demonstrated sensitivity and specificity between 60%–81% and 75%–78%, respectively, and showed moderate inter-reader agreement (K = 0.45) for the prediction of T3a disease.40, 41,

In addition, the length of tumour contact with the capsule at MRI (Figure 3) has also been shown to be a strong predictor of ECE35,42,43 with good to excellent inter-reader agreement26, 35, (Figure 3).26,35 However, a reliable threshold is yet to be established, with reported rates varying from 6–20 mm, the PI-RADS v2 guidelines recommend an arbitrary threshold of 10 mm19, which pre-dates many of these studies. The reason for variability is likely multifactorial with different methodology employed and variations in scanner strength, vendor and protocols. For instance, Rosenkrantz et al. measured the length in a linear rather than curvilinear fashion which likely explains their lower reported threshold of 6 mm.35 In addition, a more recent study suggests that thresholds differ between low-(Grade Group 1–2) and high-grade (Grade Group 3–5) cancers, with the former having a positive predictive values (PPV) of 90.4% for ECE at 12.5 mm and the optimal cut-off for the latter being 5 mm.26 This finding was further confirmed by Matsuoka et al. who reported significantly increased upstaging in low- versus high-grade cancers when the same threshold (10 mm) was applied.44 Given that lower apparent diffusion coefficient (ADC) values in prostate cancer correlate with higher Gleason score45,46, this could potentially be exploited as an adjunct for more accurate diagnosis of T3a disease prior to biopsy results. To date however, there have been mixed results when applying ADC values for stage assessment, which may relate to difficulties in applying uniform quantitative values.39,47,48,49,50

Figure 3

Another potential approach to improve sensitivity is utilisation of an isovolumetric 3D T2 imaging sequence to acquire thinner slices with less partial voluming and reformatting of isotropic images in multiple planes. Studies using 3D-T2 sequence have reported encouraging results with sensitivity and specificity ranging from 58.3%–84% and 73.1%–89%, respectively.44,51,5253 In addition, Caglic et al. proposed a new criterion of 3D Contact which significantly improved detection of ECE (sensitivity, specificity: 73.7% and 87.8%) when compared to the length of capsular contact measured on conventional T2 imaging in axial plane (sensitivity, specificity: 59.6%, 87.8%).26 This approach exploited the reduced partial voluming due to thinner slices (Figure 4) and reconstruction of images in multiple planes in order to measure the more representative a truer length of capsular contact. Although not supported by work of Jäderling et

Figure 4

al. using 3D-T2 reconstructions, it should be noted that their analysis was based on morphological criteria and not on quantifying capsular contact.54

Although diagnostic accuracy for early ECE is improving, sensitivity remains relatively poor, and it should be noted that these results come from experienced centres, utilizing modern equipment and optimised protocols. As a result, equivocal MRI findings should not change the planned treatment course, but rather ensure discussion between radiologists and urologists at multidisciplinary meetings on a case-by-case basis. Practical advice would be to flag indeterminate features of ECE, to allow wider surgical margins to be taken in the corresponding region.10 Furthermore, reporting the exact location of T3a disease is important, as clear margins are harder to obtain at the apex whereas tumours remote from the neurovascular bundle (NVB) such as in the anterior location will allow nerve sparing surgery and reduce resultant morbidity from urinary incontinence or erectile dysfunction.

T3b disease is defined as involvement of one or both seminal vesicles (SV) by prostate cancer, with the prevalence of SV invasion in surgical series being reported at 4–23%.55, 56, Patients with T3b disease carry an increased risk of lymph node involvement, local recurrence and distant metastases57, making preoperative identification of SV involvement an important factor for prognosis and treatment planning. Patients with T3b disease are typically not offered radical prostatectomy or brachytherapy unless as part of a multimodal approach, and are usually offered external beam radiotherapy (EBRT) and androgen deprivation therapy (ADT).

MRI has been shown to outperform clinical risk assessment tools such as Kattan nomogram and Partin tables in predicting SV involvement13,58 with meta-analyses showing moderate sensitivity of 73% and specificity of 95% for multiparametric MRI studies at 3T.38 Recent work by Grivas et al. including 527 patients at 3T mpMRI achieved similar results with sensitivity, specificity, PPV and negative predictive values (NPV) at 75.9%, 94.7%, 62% and 97%, respectively.59

Histopathologically, SV invasion is defined as prostate cancer penetrating the SV muscular wall, with tumour involving the extraprostatic portion of the vesicles rather than the intraprostatic ejaculatory ducts.57 Three routes of invasion have been described, Type I: direct spread via the ejaculatory duct complex (Figure 5), Type II: extracapsular spread of disease with invasion via the outer seminal vesicle wall and type III: metastatic involvement from a remote intraprostatic lesion (Figure 6). The first two types individually or in combination account for more than 95% of cases, with type III spread being extremely rare.60, 61

Figure 5

Figure 6

Seminal vesicles are best evaluated on T2W imaging in combination with functional imaging. Coronal or sagittal reformats are especially useful in demonstrating the type of spread. In Type I invasion, invasion via the erectile dysfunction (ED) causes SV expansion with a low signal intraluminal mass and may cause diffuse or focal wall thickening. In Type II involvement, there is obliteration of the angle between the base of the prostate and SV.60 In addition, in 2009 Jung et al. proposed a novel six-tier classification system for SV invasion based on morphological appearance of SV on T2W imaging (Class 0 = normal SV appearance, Class 5 = apparent mass lesion with destructive architecture) showing sensitivity and specificity of 71.4 and 96.6%, respectively.62 More recent work incorporating functional sequences has further increased accuracy, with DWI proving to be of more incremental value than DCE.56, 63

There are known pitfalls to be aware of when assessing for SV involvement, such as diffuse wall thickening due to SV atrophy or asymmetry. In addition, there can be large variation with a mean right - left asymmetry of 20% and surgical series suggesting SV length between sides can vary up to 9-fold.64 Post-biopsy haemorrhage can mimic the low T2 signal of prostate cancer and review

of T1W imaging for high signal within the SV is therefore essential. Another important mimic of prostate cancer is amyloidosis which exhibits low T2 signal but does not show diffusion restriction (Figure 7). In cases with equivocal SV findings on prebiopsy MRI, patients can undergo subsequent targeted biopsy in order to correctly stage the disease (Figure 7).65

Figure 7

Although PI-RADS v2 does not recommend abstinence from ejaculation prior to prostate mpM-RI19, some centres require patients to refrain from ejaculation prior to imaging in order to achieve maximal distension. Recent studies suggest 72 hours abstinence as the recommended interval to achieve maximal seminal vesicle distension.66, 67, 68, 69, This may be beneficial in evaluation of seminal vesicle invasion but further prospective studies including patients with prostate cancer are required to determine the effect on local staging accuracy.