The PI-RADS (Prostate Imaging-Reporting and Data System) standardized image acquisition and reporting is designed to be used by medical professionals in the initial evaluation of patients to assess the risk of clinically significant prostate cancer leading to biopsy and treatment [9]. The longest axis tumor diameter was more strictly introduced in the PI-RADS version 2 [10], with the cutoff of 15 mm between scores 4 and 5 on T2-weighted images. Although this PI-RADS scoring system was not designed for tumor volume assessment, some minimal requirements on tumor volume estimation were introduced, reporting the largest dimension of a suspicious finding on an axial image. If the largest dimension of a suspicious finding is on sagittal and/or coronal images, this measurement and imaging plane should also be reported. If a lesion is not clearly delineated on an axial image, the measurement on the image which best depicts the finding should be reported. If preferred, lesion volume may be determined using appropriate software, or three dimensions of lesions may be measured so that lesion volume may be calculated (max. a-p diameter × max. l-r diameter × max. c-c diameter × 0.52).

Peripheral zone lesions should be measured on apparent diffusion coefficient (ADC) , reconstructed from diffusion-weighted MR images. Transition zone lesions should be measured on T2-weighted images. If lesion measurement is difficult or compromised on ADC (for peripheral zone) or on T2-weighted images (for transition zone), measurement should be made on the sequence that shows the lesion best.

Recently, the size threshold for the highest level of suspicion (score 5) in the PI-RADS version 2 was proposed to be reduced from ≥15 mm to ≥10 mm [11], corresponding the 0.5 cc for significant prostate cancer addressed by Wolters and colleagues. However, a clear size or volume threshold on MRI remains difficult to define to discriminate significant from insignificant prostate cancer in low-grade disease. Visibility on MRI (instead of actual tumor volume) could be of additional value in discriminating significant from insignificant prostate cancer.

Tumor volume estimation becomes more important if MRI is used in monitoring patients on active surveillance. In particular, stable disease on clinical parameters and MRI may defer or avoid systematic prostate cancer biopsies. In an international consensus meeting, recommendations have been developed to collect data in men having MRI on active surveillance (PRECISE criteria—Prostate Cancer Radiological Estimation of Change in Sequential Evaluation) [12]. Besides data on MRI suspicion score (Likert or PI-RADS v2) and the likelihood of progression being present, the absolute size of a lesion at baseline and follow-up should be reported, addressing the importance of MRI tumor volume assessment. However, to identify tumor progression on MRI, MRI tumor volume estimation should be robust. Today, no clear recommendations on MRI tumor volume assessment are present.

It is suggested that all metastatic sites in a single patient may derive from a single monoclonal precursor cell [13], indicating that a single tumor focus is responsible for tumor progression and death, despite the multifocality of prostate cancer. This focus has been labeled the index lesion and has been considered to be defined by the largest tumor focus [14–16].

Increasing data supports the validity of the index lesion as the driver of prognosis and any adverse oncologic outcome and therefore necessitates consistently localizing the index lesion in each patient. MR imaging may offer a more robust evaluation of the entire prostate that can facilitate index lesion localization.

Index lesion identification by MRI has been proven to be well associated with the largest tumor focus in histopathology analysis, with correct identification and localization of 94–98% by MRI [17, 18]. Clearly, these high percentages of correct index lesion identification may be dependent on tumor size and Gleason grade.

Intuitively, one might think that tumors grow expansively in all directions (rounded volume); however, in reality, prostate cancer may spread/diffuse through the tissue or along the prostatic border. This may hamper the volume estimation by maximum diameter measurements on imaging. Planimetric volume measurements (delineation of the tumor at each axial slide) would be most accurate; it will take significantly more time during reporting in radiological practice, and sufficient software should be available. Furthermore, for lesions best seen on DWI/ADC image sequences, a single diameter may be more reproducible than a volume because of the need to use larger voxel sizes in sequence acquisitions.

If volume on MRI could be estimated from an ellipsoid or oval tumor, the axial diameter of a 0.5 cc, 1.3 cc, and 2.5 cc tumor would be approximately ~10 mm, ~14 mm, and ~17 mm, respectively. Robustness of MRI measurements is crucial for MRI tumor volume estimation.

The clinical significance of prostate cancer tumor volume in radical prostatectomy specimens is controversial [19, 20]. In univariable analysis, prostate tumor volume is of prognostic significance, but most studies report that its prognostic significance is lost in multivariable analysis [21–25]. This may be attributed to the close correlation of tumor volume with Gleason score, T-stage, and surgical margin status of radical prostatectomy specimens. Due to the lack of independent prognostic significance, most pathologists choose not to report routinely an exact measure of tumor volume [26]. Despite this controversial issue and difficulties in accurately measuring tumor volume, pathologists agreed that some form of semiquantitative measurement of tumor volume should be reported, although no consensus was reached regarding method(s) [5].

Although there are many methods available to calculate prostate tumo r volume, obviously, computer-assisted image analysis systems (planimetrically) are considered the most accurate means of determining tumor volume [27]. Due to the high labor intensity, these planimetric methods are not commonly used in routine pathology practice. Alternative methods for the estimation of tumor dimensions of tumor volume (or size) rely on calculation of basic tumor dimensions [28–31]. In routine pathology laboratories, the ellipsoid formula (0.52 × length × width × height) (Fig. 10.2) may provide a close estimate of the tumor volume [31], as compared to planimetrically determined tumor volume by computer image analysis, with sensitivity and specificity of, respectively, 94% and 92% for tumor volume >0.5 cc. However, others showed the ellipsoidal method may overestimate tumor volume by approximately 30% [28, 29] and would use the formula 0.4 × length × width × height. This ellipsoid measurement is simple and reproducible and is considered appropriate for pathological tumor measurements. Nevertheless, it is not routinely applied, except in a research setting. Most clinicians do not use the measurement of cancer volume by this technique in clinical decision making, and it is not currently recommended.

Relationships between MRI signal intensities and the underlying architectural prostatic tissue are complex. Langer and colleagues showed the association between specific alterations in tissue composition and MR imaging measurements [32]. They have shown that the increased percentage area of nuclei and cytoplasm, in combination with the decreased percentage area of luminal space, within equal stroma, corresponds to a decrease in T2-weighted and ADC signal intensities (Figs. 10.3 and 10.4). Furthermore, these morphological changes also correspond to the increase of the DCE-MRI volume transfer constant [K^trans] and decrease to the DCE-MRI extravascular extracellular volume fraction [v _e] (Figs. 10.3 and 10.4).

They also determined that the number of cellular components is significantly different between malignant and benign peripheral zone tissue, and thus these mechanisms influence prostate cancer detection with MR imaging. T2-weighted imaging is sensitive to extracellular water; ADC is sensitive to diffusion in the lumen; thus increased cellular texture corresponds to decrease in T2-weighted and ADC signal intensities. Because DCE-MRI-derived parameters are expected to be related to properties of the vasculature (rather than cellular composition), changes in cellular components occur in abnormalities associated with vascular density. Figure 10.4 shows graphically the signal intensity decays and increases of T2w, ADC, K^trans, and v_e, based on their published mean slopes for MRI measurements versus proportion of cellular components. Other groups have also shown this correlation between ADC and cellular density [34, 35] and between contrast MR imaging parameters and microvessel density [36] on radical prostatectomy specimen.

MRI-derived parameters are reflective of pathologically determined characteristics of prostate cancer; however, there is great overlap with benign conditions, such as benign prostate hyperplasia, inflammation, or fibrosis. For example, nucleomegaly and increased cellular density, in addition to indicating prostate cancer, can also be indicative of inflammation and prostatic intraepithelial neoplasia. Luminal size can be altered in cystically dilated glands, regions of atrophy, or fibrosis. These benign abnormalities have been implicated as sources of false-positive MR imaging findings [37–40] or poor radiologic-pathologic volumetric correspondence [41].

In all such studies, MRI measurements have been determined unblinded to pathology results of prostatectomy specimen. Furthermore, in these studies, the impact on tumor detection was not investigated. In clinical practice, this detailed knowledge of tumor distribution obtained from prostatectomy specimen is of course not at hand at the time of prospective patient management. Therefore, it is the relationship between tumor detection and histopathological features that is of clinical relevance.

Definitely, tumor size and tumor aggressiveness may have serious impact on tumor detection on MRI. Vargas and colleagues found that the integrated PI-RADS v2 scores resulted in the correct classification of 94% (118/125) peripheral zone tumors and 95% (42/44) transition zone tumors with ≥0.5 cc on pathology with any Gleason grading [42]. This correct classification was limited for the assessment of small tumor volumes of ≤0.5 cc with Gleason ≥4 + 3. The majority of GS ≥4 + 3 tumors with volumes <0.5 cc on pathology were not detectable on MRI; MRI was only able to identify 26% (7/27) peripheral zone tumors and 20% (2/10) transition zone tumors (suspicion scores 4 and 5, PI-RADS v2). This is not surprising: previously this group showed that lesion detectability on MRI was volume dependent and Gleason grade dependent [43]. Nonetheless, this is important to consider the limited ability of MRI to detect small lesions. This was a retrospective analysis biased by the inherent limitations of such a study design. Most likely, without the knowledge of the presence of prostate cancer, prospective studies may show even worse outcomes for MRI lesion detection of <0.5 cc. These findings have implications for the use of MR imaging in the management of patients with clinically low-risk prostate cancer, who now constitute about one-half of all patients diagnosed with prostate cancer in the Western world [44].

Bratan and colleagues found similar results [45]. They explored the tumor detection rates of peripheral and transition zone tumors on MRI as a function of histological parameters. MRI detection rates were 57–63% and 36–41% for all peripheral zone tumors and transition zone tumors, respectively, with small interobserver variation between two readers. They found an increased tumor detection rate on MRI, related to increased histology tumor volume: MRI tumor detection rates of 28%, 38%, 76%, and 96% corresponded to <0.05 cc, 0.05–0.5 cc, 0.5–2.0 cc, and >2.0 cc on planimetry histology analysis. Although the ellipsoid tumor volumes 0.05 cc, 0.5 cc, and 2.0 cc may correspond to an axial diameter on MRI of approximately 5 mm, 10 mm, and 16 mm, most likely these diameters would not always be measured on MRI, as underestimation is known when the tumor phenotype is not represented by dense tissue. Interestingly, this group also investigated the MRI tumor detection rate in relation to histological architecture. Increased MRI tumor detection rates of 35%, 40%, 62%, and 66% corresponded to lobulated, infiltrative, mixed, and dense tissue on histology analysis. Furthermore, increased MRI tumor detection rates of 38%, 76%, and 96% corresponded to increased Gleason scores 6, 7, and 8–9. From these data, we may conclude that prostate cancer location, histological volume, Gleason score, and histological architecture were independent significant predictors of tumor detection on multiparametric MRI.

In some patients, MRI is not able to visualize an index tumor. In some studies, these tumors proved to be microscopic at histopathology with volumes almost all below the 0.2 cc [17, 46]; however, in other studies, these tumors proved to be larger than 0.2 cc [42, 45, 47]. Although patients with small tumors may be better candidates for active surveillance, irrespective of low-risk or intermediate-risk disease, non-visible tumors are not always small tumors.

Langer and colleagues showed that no significant differences in ADC or quantitative T2-weighted values were present between the surrounding normal peripheral zone tissue and the “sparse” prostate tumors, which contain a high percentage of normal peripheral zone tissue, intermixed with prostate cancer [33]. The presence of regions within prostate tumors that are intrinsically invisible by using T2-weighted and ADC-based tissue contrast may limit accurate determination of tumor volume and target definition for active surveillance but also for MRI-driven targeted biopsies or focal therapy.

Some histological characteristics are predisposed to be more visible by MRI. Rosenkrantz and colleagues described the presence of “solid tumor growth” as a key contributor to tumor detection on MRI that was present in only 58% of their investigated tumors [48]. Although there was a significant difference between the detected and missed tumors on MRI for all assessed histological features (i.e., loose stroma, desmoplastic stroma, solid tumor growth), only size, Gleason score , and the presence of solid tumor growth were independent predictors for MRI visibility on multivariate analysis. Solid tumor growth had a substantially high odds ratio (17.83) after accounting for the effects of other features, supporting the particular importance of the formation of a discrete nodule of continuous tumor growth to facilitate MRI detection. They hypothesized that the inability to visualize a known tumor on MRI may indicate the absence of a distinct nodule of packed malignant glands encompassed by the tumor. In addition, desmoplastic tumor-associated stroma was present in all detected tumors, supporting the important role of stromal as well as epithelial components in tumor detection by MRI [48].

Although MRI has the big advantage of in vivo tumor volume estimation and could therefore be valuable in diagnosis, in monitoring, and in treatment decision-making, critical analysis of the literature shows still some big hurdles to overcome.