Improvement of RARP Outcomes Via 3D Printed/Virtual Prostate Models (RARP-3D)

Improvement of Robotic-assisted Radical Prostatectomy (RARP) Outcomes Via Automatedly Segmented 3D Printed and Virtual Prostate Models: a Feasibility Study

Study the effect 3D printed or 3D virtual prostate models of a patient, when manipulated by surgeons during RARP, has on positive surgical margins and functional outcomes of patients. Our main hypothesis is that there is a reduction of positive resection margins and functional outcomes of patients undergoing RARP when surgeons are presented with 3D printed or 3D virtual patient-specific prostate models during surgery. Specifically, we hypothesize that the anatomical knowledge of surgeons that results from the manipulation of 3D printed/virtual models constructed from automated segmentations reduces positive resection margins and functional outcomes.

This is a parallel group research feasibility study consisting of two intervention arms (3D printed and 3D virtual models) and a control group (standard practice). Intervention groups are prospective; control group is retrospective. Prospective patients, complying with the inclusion criteria, will randomly be allocated to only one intervention group. Primary outcomes Study the effect of two-intervention arms (3D printed and virtual prostate models) have on the improvement of positive resection margins after RARP, validate the accuracy of automated methods when identifying masks of the prostate gland and, cancer lesions urethra, neurovascular bundles, and external sphincter, and validate the effectiveness of an automated deployment pipeline with the goal of setting groundwork in preparation for a randomise control trial in a subsequent study. Secondary outcomes Study the effect of two-intervention arms have on functional outcomes, surgeons' and patients' perspectives on using 3D prostate models. A total of 162 cases will be considered in this feasibility study stratified into 3 cohorts: Control group. The control group will consist of 54 retrospective case-matched dataset whereby mp-MRI, positive resection margins, and functional outcomes will be collected and used as a baseline. Automated segmentation of prostate gland and lesions will be done on mp-MRI. Intervention arm 1 - 3D printed models. This cohort will consist of 54 prospective cases whereby patient-specific 3D printed models will be available to the surgeon during RARP for manipulation. Intervention arm 2 - 3D virtual models. This cohort will consist of 54 prospective cases whereby 3D virtual models will be available to the surgeon during RARP for manipulation using Innersight Labs platform.

A total of 108 patients will be recruited prospectively, i.e. 54 patients for intervention arm 1 (3D printed models), and 54 patients for intervention arm 2 (3D virtual models). An additional set of 54 patients will be obtained retrospectively.

This cohort will consist of 54 prospective cases whereby patient-specific 3D printed models will be available to the surgeon during RARP for manipulation.

This cohort will consist of 54 prospective cases whereby 3D virtual models will be available to the surgeon during RARP for manipulation using Innersight Labs platform.

3D patient-specific printed models are given to the surgeon before the start of RARP on a given patient. These 3D models are generated using automatic segmentation in MONAI followed by validation by a radiologist, and then post-processed for 3D printing.

3D patient-specific virtual models loaded into Innersight Labs platform are given to the surgeon before the start of RARP on a given patient. These 3D models are generated using automatic segmentation in MONAI followed by validation by a radiologist, and then post-processed for 3D visualisation.

Automated segmentation metrics in relation to the accuracy of predicted masks of prostate gland and lesions

Assessed after mp-MRI of patient is available and before surgery (RARP). Reported at end of study (6 months).

Patient recruitment rate will be captured throughout the study to measure effectiveness in preparation for a follow-up randomised control trial.

Percentage of cases that led to successful model deployment to the theatre will be captured throughout the study to measure effectiveness in preparation for a follow-up randomised control trial.

These are captured via questionnaires (EPROM) answered by patients in relation to urinary incontinence measured by pad weights.

These are captured via questionnaires (ICIQ-LUTS-QoL) answered by patients in relation to quality of life resulting from urinary incontinence.

These are captured via questionnaires (IIEF, and erectile hardness score) answered by patients in relation to erectile dysfunction.

Surgeon's perspectives on the use of patient-specific 3D printed and virtual prostate models during surgery will be captured after surgery

Assessed only once after surgery and once surgeon has participated in both arms. Reported at end of study (6 months)

Patient's perspectives on the use of their patient-specific 3D printed and virtual prostate model after surgery will be captured after surgery during clinic review

Assessed after surgery and after first clinical follow-up at 6 weeks. Reported at end of study (6 months).

Accuracy metrics of a model that will automatically identify neurovascular bundles, urethra, and external sphincter based on manually identified masks on mp-MRI

Assessed after mp-MRI of patient is available and before surgery (RARP). Reported at end of study (6 months).

Surgical phase and action recognition accuracy metrics of RARP endoscopic videos to understand the actions done leading to the reported complications

Inclusion Criteria: Eligible for RARP after assessment of mp-MRI during multi-disciplinary team meetings at Guy's Hospital T2b-T3 prostate cancer patients Gleason's score>=3+4 . Exclusion Criteria: prior treatment for prostate cancer patients with pre-existing urinary incontinence problems patients where mp-MRI scans are not possible patients participating in other studies investigating functional outcomes after surgery will be excluded. This is to avoid other studies influencing our secondary endpoints.