Focus of rectal cancer research has shifted to functional outcomes. Long-term morbidity after multimodal treatment for rectal cancer is suggested to be mainly made up by nerve-injury-related dysfunctions such as urinary, sexual, and anorectal dysfunctions [
1‐
3]. Recently, the performance of stereotactic navigation for minimally invasive transanal rectal surgery was reported [
4]. To assess the additional challenges related to stereotactic pelvic navigation compared with surgical navigation in other contexts such as neurosurgery and orthopedic surgery, a study with human male anatomical specimens was performed [
5]. It was concluded that accurate stereotactic pelvic surgical navigation should be feasible [
5]. Its application is suggested to improve safety and quality of the surgery as has been shown for stereotactic navigation in other contexts [
6]. Stereotactic navigation would be more effective when structures at risk of iatrogenic damage, such as the pelvic nerves, are highlighted.
Pelvic nerves consist of sympathetic, parasympathetic, and/or somatic fibers. The pudendal (PN), levator ani (LAN), and obturator nerve (ON) originate from sacral (SNs) and lumbar nerves. In addition, the SNs transfer fibers to the inferior hypogastric plexus (IHP) through the pelvic splanchnic nerves (PSNs), also known as the erigent pillar or nerves. The sympathetic trunk (ST) gives off branches to the SNs and fibers to the IHP through the sacral splanchnic nerves (SSNs). The superior hypogastric plexus (SHP) originates from the inferior mesenteric plexus and follows through in the hypogastric nerves (HNs), which follow through in the IHP. These nerves ultimately innervate the following anatomical structures: bladder, seminal vesicles, prostate, rectum, urethra, pelvic floor, dermis, anal sphincter, the vessels of the lower extremities, adductor muscles as well as corpora cavernosa in men, and uterus and vagina in women. Consequently, damage can result in several types of dysfunction.
Discussion
To our knowledge, this is the first study to report 3D MRI topography of all pelvic nerves at risk during pelvic visceral surgery in living humans. The likelihood of nerve delineation correctness was verified in three ways for validation purposes. Our study contributes to the earlier studies since these earlier studies analyzed only a part of the pelvic nerves and usually not in living humans. The LAN has neither been delineated on MRI in a living human nor on MRI in a human anatomic specimen.
Computed tomography (CT) imaging provides high resolution but with a poor contrast, and it is therefore not suitable to depict peripheral nerves. MRI is a noninvasive method which can provide high-resolution 3D images with a good contrast between anatomical structures. MR neurography is commonly performed in clinical routine to visualize nerve roots and large bundles of nerves, in order to identify points of compression or disruption. Most of the time, T2-weighted sequences with fat signal suppression are used [
51,
52]. However, these sequences provide dark images with a contrast which is not ideal for the visualization of small pelvic nerves [
53]. According to dedicated pelvic radiologists, high-resolution MRI is validated for the assessment of small pelvic nerves [
53].
Another MRI acquisition method for the visualization of pelvic nerves is tractography. Diffusion tensor MR imaging (DTI) seems suitable for tractography of the lumbosacral plexus and sacral nerves [
54]. In one study, tractography was successfully performed in the SNs in 10 healthy adults and in one 12-year-old patient [
54]. However, the PN could not be delineated in the majority of cases. The authors concluded that DTI had its limitation and should not replace anatomical plexus imaging but it could offer valuable complementary information. The main reason for the failure of DTI to depict smaller nerves was that tractography of voxels on the basis of water molecules could be hampered by a low signal/noise ratio (SNR) of peripheral nerves [
53], oversensitivity detecting all nerves fibers, also nonspecific [
55], and crossing fibers resulting in false-positive or false-negative tractography [
56]. In addition, tractography of the brain was processed via specific algorithms. Contrary to the brain, the pelvis is not a single organ but an inhomogeneous area made up of different tissues, complicating the use of these algorithms. Finally, DTI was also associated with a possible human error resulting in false-negative tractography since the most representative voxels necessitated a manual selection.
In 2014, Bertrand et al. performed pelvic 3-T MRI neurography and dissections in eight adult human anatomic specimen [
7]. They attempted to identify the HN, IHP, PSN, and cavernous nerves, and compared several key points of this innervation and three anatomical references on MRI to the same key points and references after dissection. They tested several settings including T1, T2, T1 fat saturation, T2 fast relaxation fast spin-echo, and different diffusion settings with 1–2.5-mm slice thickness. They reported a congruence between these key reference points on MRI compared with the dissections, and they concluded that MRI was suitable to detect autonomous pelvic innervations [
7]. However, they did not delineate the ST, PN, LAN, and ON.
In 2007, Mauroy et al. performed pelvic MRI and dissections of the IHP in 22 female human anatomic specimens to correlate anatomical data with MRI in order to determine the IHP path [
57]. MRI with turbo spin-echo T2 sequences was performed in five female patients with a 1.5-T MRI scan in three planes with slice thickness of 3 mm without a T1 sequence. They reported a good anatomical radiological correlation using MRI for the IHP with simple points of reference.
Brown et al. performed T2-weighted MRI using a 1.5-T system in human anatomic specimen sections and in patients before they underwent TME. They compared anatomical dissections of sagittally sectioned hemi-pelvises with MRIs obtained in vivo and concluded that they could depict the IHP [
51].
In our series, multiple sequences were scanned in different planes as mentioned in the protocol. The T1- and PD-weighted sequences provided images with a contrast which was better adapted to the visualization of small pelvic nerves compared with T2-weighted sequence. This coincided with the recent findings of pelvic radiologists performing MR neurography of the lumbosacral plexus and its pelvic neural branches [
53]. In our experience, the visualization of pelvic nerves was best achieved in the axial plane, and special care was taken to optimize the resolution in this plane while accepting longer scanning times (up to 18 min) in the axial plane. The PD scan took longer than a T1 scan (18 vs. 12 min) did. Consequently, this sequence was more prone to motion artifacts.
The SHP, HN, and IHP could be bilaterally identified in 14 (70%), 16 (80%), and 14 (70%) of the 20 volunteers, respectively. In two of four cases (50%) in which no HN could be identified, there were at least two or more factors hindering nerve delineation, namely motion artifact (patient motion, bladder contractions, bowel contractions), inappropriate emptying of the bladder or fast filling and a low pelvic fat amount. In the other two scans, there was a motion artifact. In these cases, the signal intensity between nervous tissue and neighboring other structures was too little to visualize the HN and for accurate delineation. This coincided with a low mean quality score of 2 to recognize the HN on these scans (assessed by a Likert scale 0–5).
A limitation in finding and delineating nerves remained the resolution of the images. They were acquired with a 3-T MRI system, which offered a substantially higher SNR compared with a traditional 1.5-T MRI system. However, even with this higher magnetic field, the likelihood of missing small nerve branches with a width of less than 1 mm was significant, given the resolution chosen in this study (0.80 × 0.80 × 1.0 mm3). Therefore, it was not possible to recognize smaller, more distal nerve branches. Acquiring images with an even higher resolution was possible, but at the expense of a prolonged acquisition time and decreased SNR. We tried to find the best compromise by choosing an image resolution higher than the common resolutions used in MRI, but still compatible with time constraints of clinical routine. This resolution proved to be adapted to find most of the nerves of interest.
Characteristically, the bigger nerves (lumbosacral plexus, SNs, ON, and ST) could be distinguished on MRI by their fascicles. For the smaller nerves (SHP, HNs, IHP, PN, and LAN), a distinction could be made on the basis of a higher signal intensity than the intensity associated with other structures like the ureter and vessels, although the assessment of differences in signal intensity is made by visual inspection and is therefore somewhat subjective. The intensity of these nerves compared with the bigger nerves did not seem to differ much. The explanation for this finding might be that even the smallest pelvic nerves like the LAN are myelinated [
45]. To our knowledge no data exist on inter-rater reliability for MRI pelvic nerve topography between radiologist, surgeon and anatomist.
Clinical routine pelvic MRI does not require the high level of detail for the depiction of peripheral nerves. At our hospital, these scans were acquired with a minimal slice thickness of 2 mm for the T2-weighted sequences and a 3.5 slice thickness for the T1-weighted sequences. For this reason, these scans were found to be unsuitable for topographic nerve mapping in a retrospective way. To summarize, T1-weighted high-resolution 3-T MRI seems to be the optimal sequence to depict normal anatomy of the peripheral pelvic nerves.
Although MRIs of anatomic specimens are not influenced by motion artifacts, they are hindered by cell death and temperature changes, rendering different contrasts compared with MRIs performed in vivo, and requiring adaptive setting [
7]. The current study was performed in volunteers. For this reason, no contrast agents were used and nonenhanced images were acquired. Consequently, a noncontrast angiographic sequence was used including a low velocity encoding (1 cm/s), allowing for the visualization of large and smaller blood vessels, where blood velocity was reduced. This sequence was helpful to distinguish the PN and pudendal artery which were very close to each other. All in all, angiographic images were useful for 16 of the 19 volunteers in which this sequence was acquired. The next step to optimize and verify the reproducibility of this method of 3D nerve topography would logically be a clinical study with enhanced angiographic MRI. Common contrast agents in MRI such as gadolinium chelates could help to differentiate between vessels and nerves and thereby speed up the process. Anticholinergic agents might also be helpful by lowering bowel wall activity, thereby decreasing motion artifacts.
The ability to preoperatively identify the pelvic nerves on a medical imaging modality would help surgeons to better inform patients about expected functional results and to plan the operative strategy. These topographic maps are expected to be applicable during preoperative planning and intraoperative pelvic stereotactic navigation which is currently being developed for pelvic surgery [
5]. Such an application during the laparoscopic stage of TME could facilitate dissection in the danger areas for nerve injury as previously described: the SHP at the level of the inferior mesenteric artery origin, the right and left HNs at the level of their origin, the IHP due to medial tenting as the rectum is retracted medially, and the IHP at the level of the seminal vesicles [
58]. The first pilot cases of stereotactic navigation during transanal TME are already reported [
4]. During a transanal approach, its application is supposed to prevent iatrogenic damage to the urethra and nerves and to help maintain the appropriate plane of dissection. A future application of individual 3D nerve maps in stereotactic navigation seems to be the logical next technological advancement to further improve the quality of care. To facilitate this, two projects are currently underway. Algorithms are currently developed to assist with semiautomatic nerve segmentation, thus facilitating the individualized preoperative mapping of the patient’s pelvic innervation. In addition, the positional changes and deformity of the pelvis are determined between scanning and different operating positions in order to facilitate the use of stereotactic navigation. The accuracy of neoadjuvant radiotherapy is also expected to improve by these developments facilitating image-guided radiotherapy when MRI is coupled to a Tele-radiotherapy unit [
59]. Finally, a comparison between the preoperative and postoperative nerve statuses on MRI is valuable from a scientific perspective, providing insight into the influence of iatrogenic nerve damage on functional outcome. From a training perspective, potential benefits include a better understanding of the relative location of structures and a shorter learning curve.
In short, the current study shows that pelvic nerves at risk of damage during pelvic visceral surgery are visible on 3-T MRI. A specific knowledge of their course along anatomical landmarks and dedicated scanning protocols allows for 3D mapping. This opens up new opportunities such as its application in stereotactic navigation during pelvic visceral surgery.