PRO-TRACT: Prostate Tractography for Optimized Nerve-Sparing - A Pilot Study

In Switzerland and Western countries, prostate cancer (PCa) is the most frequently diagnosed solid cancer in men and represents the second leading cause of cancer-related deaths. Similarly, in Geneva, between 1995 and 2021, PCa has been the most common cancer with 8,373 new cases reported.

For patients with prolonged life expectancy harboring high-risk and intermediate-risk disease, radical prostatectomy (RP) is one of the gold standard treatments. While RP is a safe and effective surgical procedure leading to a high rate of cancer control, the genito-urinary function can be severely impaired after surgery since the prostate is closely surrounded by neurovascular fibers that are critical for urinary continence and erectile function. In 1982, Walsh developed the nerve-sparing (NS) procedure to preserve the neurovascular bundles (NVBs) . Nevertheless, accurately identifying the NVBs still remains a challenge nowadays. Despite the advancements made in the area of precision surgery with the advent of 3D robotic-assisted RP (RARP) as well as preoperative planning (particularly with Multiparametric magnetic resonance imaging (mpMRI) providing information about tumor localization and extracapsular extension), erectile dysfunction (ED) is currently the most common post-operative sequela occurring in up to 74.7% of patients. One of the key reasons is linked to the fact that the NVBs are not clearly visible during surgery and are in contact with the prostatic capsule.

Novel intraoperative imaging modalities have emerged to solve this unmet clinical need. First, it has been suggested that assistance can be provided by intraoperative transrectal ultrasound (TRUS) eventually combined with a robotic manipulator. TRUS can identify key anatomical landmarks. However, it is rarely used due to the poor spatial resolution. Second, electrical nerve stimulation offers fast functional feedback for nerve mapping. Although this solution is deliverable, its inconsistent clinical response reduces its efficacy and usefulness. Third, fluorescent imaging allows for a real-time assessment, but its clinical use is limited due to, among others,the toxicity of dyes.

Tractography, mainly used in brain surgery, is an advanced non-invasive imaging technique relying on diffusion tensor imaging (DTI) and algorithms to reconstruct three-dimensional nerve fiber pathways by analyzing directional movement of water molecules in tissues The accuracy of DTI Tractography heavily relies on sophisticated protocols and highly developed technical parameters, which also includes the post-processing aspects. For instance, the number of gradient diffusion directions, b-values affecting the signal-to-noise ratio (SNR) and tract fiber length play a key role, as studied in Nordbrøden's thesis focusing on providing an optimized protocol for periprostatic fiber nerve tracking. Due to inherent complexity, despite being a promising tool, application of tractography for NVB mapping preoperatively has not been well studied yet. To improve functional outcomes, surgeons should be able to rely on a patient-tailored anatomical map to provide personalized nerve-sparing surgery.

Moreover, three-dimensional virtual model (3DVM), reconstructed from classical preoperative MRI done during the diagnostic pathway, are gaining popularity in RARP. Various formats are described; for instance, Virtual Reality (VR) uses a fullly digital immersive environment, whereas Augmented Reality (AR) overlays the 3D model onto the intraoperative view. However, integration of such a tool into clinical practice faces many challenges at each step of the procedure. First, when generating the model, suboptimal image quality and the difficult segmentation task needed for surface rendering (SR) could generate anatomical inaccuracy. Secondly, in AR, the 3DVM needs to be registered to the patient's anatomy (landmarks) using manual or semi-automatic techniques, with fully-automatic registration being the subject of ongoing research . Nonetheless, Schiavina et al (2021) using a manual registration accurately managed a real-time identification of the index lesion with the downside of having a dedicated 3DVM manipulator changing the NS plan in 38.5%.