Magnetic resonance imaging and endorectal ultrasound for diagnosis of rectal lesions

Magnetic resonance imaging (MRI), with both endorectal and pelvic phased array coils, is the gold standard, especially in oncological rectal examination. An endorectal coil is a surface coil and provides very good images of the wall of the organ, but offers limited information on surrounding structures. Moreover, the best images are taken only at the level of the coil. Its usefulness is also limited because of the patient discomfort and in case of anal stenosis. Rectal wall motion is also responsible for the coil migration and misinterpretation of the observed lesions. For these reasons phased array coils are recommended in routine cancer staging. They obtain higher signal, but with wider coverage and improved homogeneity [24]. In rectal cancer, the pelvic coil showed high accuracy in tumor (59-95%) and nodal metastasis evaluation (39-95%). Similar data for the endorectal coil reached 66–91 and 72-79%, respectively [25,26], and for T staging was similar to ERUS [27,28]. However, it has to be stressed that unlike for ultrasound, there are a number of absolute and relative contraindications for MRI (Table 2). The examination can be performed in pregnant women, but without contrast media administration and using a limited number of sequences [29,30].

The examination does not require any previous rectal preparation but in some medical centers a prior enema is recommended. Furthermore, the usage of intrarectal water, ultrasound gel, gadolinium or air insufflations as well as pretreatment with spasmolytic agents is debatable, but they may strongly improve image quality [2]. Most skeptics stress that any kind of artificial substances inside the rectum may be uncomfortable to the patient and, more importantly, may compress the mesorectal fat and influence evaluation of the circumferential resection margin. The main exception to the contrast enema is a dynamic rectal examination, also known as MRI defecography or proctography, performed for various pelvic floor dysfunctions [31]. However, similar data, such as for Park’s angle and various mobile rectal diameters, could be obtained during ERUS and ultrasound examination of the perineum with a linear probe [7,13].

The rectal MRI examination is performed on 1.5-T or higher systems, but the reported sensitivity and specificity for 3 T are very similar to those obtained for 1.5 T. However, due to the high signal, 3-T equipment may obtain thinner T2-weighted images, which are the most suitable for rectal evaluation [32].

During the examination, the patient has to be comfortably positioned in a supine position, since image acquisition usually takes about 40–60 min. Like in most pelvic examinations, breath holding is not recommended. After initial coronal and sagittal localization images, usually a sagittal T2-weighted, fast (turbo) spin-echo sequence is performed. It is followed by large-field-of-view axial sections of the whole pelvis and T2-weighted thin-section axial sections perpendicular to the long axis of the rectum (Tables 3 and 4; Figures 3 and 4). In case of low rectal cancers, high-spatial-resolution coronal imaging is recommended just to see their position in relation to the levator ani and sphincter complex [33]. It is especially important in planning sphincter-sparing surgery [34]. In perianal fistulas, small-field-of-view axial T2-weighted sections seem to be more reasonable. Moreover, radial water- and fat-saturation sequences are useful to obtain spatial pseudo 3D reconstruction. Such MRI hydrography has been previously used routinely to visualize the biliary and pancreatic ducts (MRCP) and urinary pathways through depiction of static fluid. Currently, in most medical centers the examination is performed according to the suggestions of The Magnetic Resonance Imaging and Rectal Cancer European Equivalence Study (MERCURY) [35]. Three-dimensional (3D) T2-weighted sequences permit application of 1–2-mm-thin sections with no intersection gap. They should be able to compensate for difficulties with angulation of the tumor such as tortuosity and redundancy of the rectum [18]. Usefulness of the gadolinium contrast enhancement is debatable, since it does not improve evaluation of local staging [36]. Moreover, since contrast enhancement requires fat suppression, it results in reduction of the signal-to-noise ratio and potential overstaging of the tumor due to enhancement of adjacent nonmalignant structures such as the vessels, desmoplastic stromal reaction and normal nodes [24]. Unlike other authors, Zhang et al. [34] indicated that 3D fat-suppressed dynamic contrast enhancement is the best technique to delineate the tumor margins. On the other hand, contrast injection may be helpful in examination of the internal morphology of the tumor as well as various perfusion values, which are important in the evaluation of an early treatment response and in tumor recurrence. For the same reason, diffusion-weighted images (DWIs) should be routinely performed, using increasing b values (in our hospital usually 0, 50, 500 and 1,000 s/mm²) (Figure 5). Even though the DWI images together with ADC maps are very helpful in detection of the lesions, they cannot be applied to confirm malignancy without examination of classical T1- and T2-weighted sequences [37]. In both perfusion and ADC measurements, the region of interest (ROI) should only fit the pathological lesion, without the neighboring tissues. Even ADC measurement might be subjective; Attenberger et al. [38] indicated that by using strict criteria they could be analyzed with good interobserver agreement in patients with rectal cancer.

From the clinical point of view, T2-weighted images without fat saturation are the most helpful in distinguishing normal morphology of the rectal wall from organ abnormalities. The most internal hyperintense layer is formed by the mucosa and submucosa, which cannot be differentiated by either MRI or ERUS. The next hypo- and hyperintense layers represent the muscularis propria and perirectal fat, respectively. Usually, the mesorectal fascia should be also detected as a low-signal line visible on the margin of the mesorectum [2]. However, there are studies in which five layers were clearly described. Stollfuss et al. [39] were able to distinguish hypointense mucosa and hyperintense submucosa in 19 of 24 post-resection specimens using T2-weighted images obtained on a 1.5-T system. The internally located circular-directed muscular fibers had higher signal than the outer longitudinal-directed fibers.

It is also important that MRI may differentiate mucinous from nonmucinous adenocarcinomas [40,41] and may be helpful in the identification of early recurrence from postoperative changes (perianal fistulas and sinuses) [11,12]. Moreover, unlike CT, MRI does not use ionizing radiation and highly nephrotoxic contrasts.

In the final radiological report, any abnormalities should be identified and clearly described. In both malignant and nonmalignant rectal lesions various calcifications have been found. Anatomically or for general description, rectal lesions, like the organ itself – as described in detail in the introduction – may be divided into lower, middle and upper ones. However, in clinical practice, the Parks [8,12] and WHO TNM classifications are commonly used (Table 5) [42]. The former describes four primary types of perianal fistulas: intersphincteric, transsphincteric, suprasphincteric and extrasphincteric. The latter is used for tumor staging and helps to divide patients for surgery (stage T1 and T2) or for preoperative therapy (>T2). Nowadays, patients with T1 stage undergo transanal endoscopic microsurgery, while those with stages T2 and T3 are usually treated with a total mesorectal excision, without or after preoperative neoadjuvant therapy (radio- or radiochemotherapy), respectively. Such a therapeutic schema highly increases the 5-year survival rate when compared with conventional surgery. Moreover, patients with preoperative neoadjuvant therapy had a substantially lower rate of local recurrence compared with those who received similar postoperative treatment [43]. On the other hand, low rectal tumors form a distinct entity among all neoplasms of the organ as they have a high risk of local recurrence and poor outcome compared to lesions located in the middle and upper rectum [44]. It is especially important in the abdominoperineal excision, which is characterized by a relatively high risk of local recurrence (>30%) [35].

In nonmalignant lesions ERUS offers similar accuracy to MRI and could be easily applied in all pathologies located close to the rectal wall. Due to the higher spatial resolution, it is even better than MRI in differentiating T1 (T1sm1 and T1sm2) and T2 tumors, but is more subjective and depends highly on the sonographer’s experience [45]. In contrast, MRI is more reproducible and allows accurate evaluation of neoplastic and distal spread, including measurement of mesorectal involvement and establishing the potential surgical circumferential resection margin. According to Wieder et al. [46], MRI led to accurate prediction of the circumferential resection margin with 100% sensitivity and 88% specificity, which depend on the minimum distance of the tumor to the mesorectal fascia seen in histological examinations. It is also a method of choice to exclude infiltration of structures located nearby, but in such an examination the gadolinium contrast enhancement strongly increases the accuracy [18,46,47]. The main problem with MRI is an overstaging usually caused by a desmoplastic reaction in nonneoplastic structures located close to the tumor margin [47]. It usually happens in case of stage T2 and T3. However, in stage T3 the muscularis propria signal intensity is unclear, while its external margin loses smoothness and becomes more nodular or spiculated. On the other hand, hyperplasia of the muscularis propria is associated with radiation and may also increase overinterpretation [39]. It is worth mentioning that computed tomography (CT) is not indicated in anal tumor diagnosis because of low sensitivity (66%) [48,49].

Irrespective of the applied method for each tumor, clinicians require the following information: tumor spread, including detailed evaluation of the mesorectal fascia (T feature) and extramural venous invasion, lymph node involvement (N feature) and presence of distal metastases (M feature). However, in case of rectal abnormalities the distance from the sphincter anal complex as well as the sphincters’ morphology and their function are also important because a low rectal tumor without any sphincter invasion and a distance between its inferior pole and the upper margin of the internal sphincter may be treated with low anterior resection consisting of en-bloc resection of the rectum with total mesorectal excision [18]. If possible, the distance to the mesorectal fascia or levator ani muscles for low rectal tumors should also be established. From a clinical point of view, it is more important to measure the depth of extramural spread in the mesorectal fat than to ascertain the T stage, since a T2 tumor has the same prognosis as a T3 tumor with less than 1 mm spread. Moreover, T3 tumors with less and more than 5 mm mesorectal invasion have different 5-year survival rates, i.e., 85 and 54%, respectively [50,51]. According to Beets-Tan et al. [52], perirectal spread terminating at least 5 mm from the fascia predicts an uninvolved circumferential margin of 1 mm at histological analysis with 97% confidence. However, such measurement is difficult and imprecise in tumors located in a lower third of the anal canal, on its anterior wall or in patients with a small amount of perirectal fat [2]. All these features are important prognostic factors and crucial for therapeutic management.

Extramural venous invasion – beyond the muscularis propria in an endothelium-lined vessel – is important since neoplastic cell embolism in the portal circulation may initiate distant metastases through hematogenous spread [53]. It is associated with a higher incidence of local and distant metastases as well as poorer overall survival rates [54,55]. Such complications were observed histopathologically in 10-54% cases of rectal cancer [56], but could be also evaluated by MRI but not CT [57]. A recent report [58] states that patients with MRI-detected venous invasion had a 3.7 times increased relative risk of metachronous metastatic disease. The most common system (Table 6) applied for such evaluations was introduced by Smith et al. [59] and divides the tumors into lesions without (score 0–2) and with venous invasion (score 3–4). Additionally, each vessel should be described as a small (a perforating vein that runs perpendicular to the rectal lumen), medium (an unnamed vein that runs parallel to the rectal lumen) or large named vein [58]. It is worth mentioning that in extramural venous invasion an abdominal MRI follow-up is recommended, since according to Scharitzer at al. [60], a gadoxetic-acid-enhanced 3-T system is more sensitive than 64-row multidetector CT in the detection of small (≤10 mm) hepatic metastases. On the other hand, MRI’s usefulness, especially in minute lesions, is limited by various artifacts related to the patient’s condition (e.g., respiratory and cardiac motion, vascular pulsation, nearby located small cysts, liver steatosis and iron accumulation, etc.) as well as artifacts and pitfalls related to the scanner and/or magnetic elements inside of the examined patient (i.e., aliasing, black boundary, chemical shift, entry slice phenomenon, Gibbs energy, magnetic susceptibility, moiré fringes, RF overflow, shading, slice-overlap, susceptibility, zebra stripes, zippers) [30,61,62]. This examination technique is also limited because of its cost and insufficient availability in many institutions. However, according to the algorithm of the European Registration of Cancer Care (EURECCA) [6], the M staging includes MRI or CT examination of the abdomen (Table 1). Furthermore, for detailed lung evaluation only CT is recommended [63]. The usefulness of PET-CT is limited only to multivisceral metastases and differentiating between fibrosis and tumor, particularly in locally recurrent rectal cancer. However, in such cases the DWI significantly improves the diagnosis [64].

Nowadays, the main problem in rectal cancer staging is evaluation of lymphatic spread. The difficulty lies in the lack of proper radiological criteria for nodal metastatic changes in the pelvis and various lymphatic pathways that carry the lymph from the rectum [1]. The lymph from the upper part of the ampulla is usually drained via pararectal lymph nodes located on the muscularis propria, either through sacral lymph nodes or directly into inferior mesenteric lymph nodes along the inferior rectal vessels. The remaining part of the rectum over the pectinate line drains the lymph directly to the sacral lymph nodes or via vessels surrounding the middle rectal artery into the internal iliac lymph nodes. All those nodes may not be visible in physiological conditions in ERUS and MR. The inferior third of the rectum, below the pectinate line, sends the lymph into the horizontal part of the superficial lymph nodes that are clearly visible in most radiological procedures.

For the evaluation of lymphatic spread, a pelvic phased array coil is recommended. Like CT, it gives an opportunity to examine most regional lymph nodes [6,35]. Discrimination between normal and metastatic lymph nodes by MRI remains problematic. For a very long time, the CT criteria that stress the lack of an oval shape and fatty sinus, round or irregular margin, as well as short transversal diameter over 10 mm, were applied to MR. However, while using such criteria, the reported accuracy was relatively low and reached only 43-85% [65]. Some authors described even lower thresholds for pararectal lymph nodes, but the sensitivity, specificity and accuracy in lymph node metastases with diameters greater than 6 mm were only 57, 88 and 76%, respectively [66]. Using a 5-mm short axis as a threshold resulted in lower sensitivity (66%) and specificity (76%) [67]. Currently, there is a tendency to report any pelvic lymph nodes since their transversal diameters have not been established in any large and multicenter studies. The irregular borders and signal intensity are principle features in node metastasis and allow much higher sensitivity (85-95%) and specificity (95-97%) [68]. Nowadays, according to the EURECCA principles, nodes >3 mm can be characterized as malignant or benign by signal and border features [6]. On high-resolution MR, identification of nodes <3 mm in diameter containing metastatic foci remains a challenge. Furthermore, most publications stress that the presence of ≥4 lymph nodes is associated with a higher risk of local recurrence [69]. However, based on new clinical observations [70,71], a well-preformed total mesorectal excision limiting nodal involvement is no longer a risk factor for a local recurrence. On the other hand, identification of involved lymph nodes located outside of the mesorectal fascia is important, as they will not be removed during a standard anterior rectal resection with total mesorectal excision [18]. Such nodes may require additional treatment since they are responsible for local recurrence.

In a problematic situation, high signal in DWI and low on ADC maps, as well as low ADC values, could also be helpful, since in metastatic infiltration similar features are seen for the tumor and nodes. However, the main limitation of the method is the relatively large size of the ROI (≥1 cm²) and at least three or four different b values used during the acquisition of the DWI sequence [30]. Attenberger et al. [38] found that ADC measurements and functional parameters were useful in differentiating N stages. However, Mizukami et al. [72] reported a high negative predictive value of DWI; even positive nodes had at least 1-cm diameter in the short axis. Promising data came from studies with new contrast agents (i.e., ferumoxtran-10, USPIO), but they are not routinely applied in clinical practice, and most of them are not registered for daily radiological practice [73,74]. Moreover, Lambregts et al. [75] introduced a novel method utilizing gadofosveset enhancement to assess chemical-shift artifacts associated with lymph node involvement, but data have not been confirmed in a multicenter study.

On the other hand, it has to be pointed out that Bipat et al. [76] indicated that in lymph node involvement and organ invasion, the estimated sensitivity and specificity of ERUS, MRI and CT were similar. However, in muscularis propria invasion, ERUS and MRI imaging had similar sensitivities, but the specificity of sonography was significantly higher (86 vs. 69%). Similar data were established for mesorectal invasion. The sensitivity of ERUS (90%) was higher than for CT (79%) and MRI (82%), while the specificities were comparable. In nodal metastatic spreading, the opposite data were recently presented by Puli et al. [77], who concluded that the lower accuracy of EURS in comparison to MRI and CT is due to lack of visualization of the entire mesorectum.