Anatomically Precise, Non-Biodegradable Whole Pelvic Phantom for Ultrasound-Guided Training in Hydrogel Spacer Injections

Abstract

*1st 2 authors are co-1st authors To limit toxicity, we commonly use hydrogel spacer injection to displace the rectum & bladder outside of high-dose radiation fields. This procedure is challenging as inaccurate needle placement can accidentally perforate nearby organs. Thus, there is a clinical need for trainees to improve needle insertion skills. Here, we propose a low-cost method to build a whole-pelvic, non-biodegradable training phantom with patient-specific anatomy and realistic textures/imaging properties. Our hypothesis is that this phantom can produce ultrasound (US) images that are comparable to patient scans using a multi-reader, multi-case (MRMC) study, non-inferiority study design. We manually contoured organs at risk (tumor, bladder, rectum, uterus & vagina) based on a patient’s clinical MRI (3D slicer software). Phantom was fabricated using 3D-printed positive molds and PVC plastisol. We created realistic tissue structures by changing the plastic softener/hardener ratio, and imaging properties by varying microbead density. To test our hypothesis, we generated “phantom images” by 1st placing an US transducer on the abdomen, then with a separate transducer into rectal cavity of the phantom. “Patient images” were obtained from clinical US scans. We asked 9 resident physicians (readers) to evaluate 10 US images (cases) and classify each as “patient” or “phantom” scan. No prior structured training was provided for US interpretation (other than routine clinic exposure). However, readers can review each image with and without pelvic structures labeled (to help visualize anatomy without obscuring key imaging characteristics). MRMC analysis (R software) was used to account for variability between cases & between readers. The null hypothesis is that phantoms scans are inferior; i.e. readers can correctly identify “patient” vs “phantom” scans at least 65% of the time (H0: p ≥ 0.65, HA: p < 0.65). We also asked readers to review 2 US videos and identify the hydrogel needle during insertion into the pelvic phantom (1 transabdominal & 1 transrectal view). Readers correctly identified “patient” vs “phantom” scans 53.3% ± 6.2% of the time (p = 0.054, t = -1.89, df = 6). The p-value is marginally >0.05, likely from limited sample size (power = 0.32). Readers correctly identified the hydrogel needle in US videos 100% of the time in transabdominal and 78% in transrectal view. In terms of cost, we can recreate this life-sized pelvic phantom (3.5L) at $60 using existing 3D molds; in comparison, commercial phantoms cost $400 (for smaller models) to $3000 (for whole-pelvic models). This 3D printed phantom is a cost-effective solution to help trainees improve hydrogel injection skills. We have shown that this phantom can produce US images that are near comparable to patient scans, though reader experience in image interpretation could be limited without structured ultrasound training. In the future, we will explore its use in brachytherapy training.