ABSTRACT
INTRODUCTION: Visualising the course of a complex perianal fistula on imaging can be difficult. It has been postulated that three-dimensional (3D) models of perianal fistulas improve understanding of the perianal pathology, contribute to surgical decision-making and might even improve future outcomes of surgical treatment. The aim of the current study is to investigate the accuracy of 3D-printed models of perianal fistulas compared with magnetic resonance imaging (MRI). METHODS: MRI scans of 15 patients with transsphincteric and intersphincteric fistulas were selected and then assessed by an experienced abdominal and colorectal radiologist. A standardised method of creating a 3D-printed anatomical model of cryptoglandular perianal fistula was developed by a technical medical physicist and a surgeon in training with special interest in 3D printing. Manual segmentation of the fistula and external sphincter was performed by a trained technical medical physicist. The anatomical models were 3D printed in a 1:1 ratio and assessed by two colorectal surgeons. The 3D-printed models were then scanned with a 3D scanner. Volume of the 3D-printed model was compared with manual segmentation. Inter-rater reliability statistics were calculated for consistency between the radiologist who assessed the MRI scans and the surgeons who assessed the 3D-printed models. The assessment of the MRI was considered the 'gold standard'. Agreement between the two surgeons who assessed the 3D printed models was also determined. RESULTS: Consistency between the radiologist and the surgeons was almost perfect for classification (κ = 0.87, κ = 0.87), substantial for complexity (κ = 0.73, κ = 0.74) and location of the internal orifice (κ = 0.73, κ = 0.73) and moderate for the percentage of involved external anal sphincter in transsphincteric fistulas (ICC 0.63, ICC 0.52). Agreement between the two surgeons was substantial for classification (κ = 0.73), complexity (κ = 0.74), location of the internal orifice (κ = 0.75) and percentage of involved external anal sphincter in transsphincteric fistulas (ICC 0.77). CONCLUSIONS: Our 3D-printed anatomical models of perianal fistulas are an accurate reflection of the MRI. Further research is needed to determine the added value of 3D-printed anatomical models in preoperative planning and education.
