[Paper Review] Pelvis Surface Estimation From Partial CT for Computer-Aided Pelvic Osteotomies

Robert B. Grupp, Yoshito Otake+4 more|arXiv (Cornell University)|Sep 23, 2019

Pelvic and Acetabular Injuries3 citations

TL;DR

This paper proposes a smooth extrapolation method using a statistical shape model (SSM) to reconstruct complete pelvic surfaces from partial CT scans, significantly reducing surface error compared to conventional cut-and-paste techniques. By leveraging the acetabulum and superior 5% of the iliac crest as priors, the method achieves an average 1.31 mm improvement in RMS surface error and 3.16 mm in maximum error, enabling accurate computer-aided pelvic osteotomies with reduced radiation exposure.

ABSTRACT

Computer-aided surgical systems commonly use preoperative CT scans when performing pelvic osteotomies for intraoperative navigation. These systems have the potential to improve the safety and accuracy of pelvic osteotomies, however, exposing the patient to radiation is a significant drawback. In order to reduce radiation exposure, we propose a new smooth extrapolation method leveraging a partial pelvis CT and a statistical shape model (SSM) of the full pelvis in order to estimate a patient's complete pelvis. A SSM of normal, complete, female pelvis anatomy was created and evaluated from 42 subjects. A leave-one-out test was performed to characterise the inherent generalisation capability of the SSM. An additional leave-one-out test was conducted to measure performance of the smooth extrapolation method and an existing "cut-and-paste" extrapolation method. Unknown anatomy was simulated by keeping the axial slices of the patient's acetabulum intact and varying the amount of the superior iliac crest retained; from 0% to 15% of the total pelvis extent. The smooth technique showed an average improvement over the cut-and-paste method of 1.31 mm and 3.61 mm, in RMS and maximum surface error, respectively. With 5% of the iliac crest retained, the smoothly estimated surface had an RMS surface error of 2.21 mm, an improvement of 1.25 mm when retaining none of the iliac crest. This anatomical estimation method creates the possibility of a patient and surgeon benefiting from the use of a CAS system and simultaneously reducing the patient's radiation exposure.

Motivation & Objective

• To reduce surface reconstruction error in incomplete pelvic CT scans for computer-aided pelvic osteotomies.
• To address the limitation of discontinuous transitions in existing cut-and-paste extrapolation methods.
• To enable accurate patient-to-CT registration in CAS systems using only partial preoperative CT scans.
• To minimize radiation exposure for adolescent and childbearing-age patients by avoiding full pelvic CT scans.
• To improve the accuracy and reliability of 3D surgical planning and navigation in periacetabular osteotomy (PAO).

Proposed method

• A statistical shape model (SSM) of the normal female pelvis was constructed from 42 valid pelvic surface meshes derived from 70 full CT volumes using deformable intensity-based volumetric registration.
• The SSM was trained using leave-one-out cross-validation to assess generalization error and guide extrapolation.
• Two extrapolation methods were compared: (1) cut-and-paste, which directly copies SSM-predicted regions onto known priors, and (2) smooth extrapolation, which uses a Thin Plate Spline (TPS) to ensure continuity between known and estimated regions.
• The TPS transformation was computed using common regions between the partial prior and SSM estimate, ensuring smooth interpolation across the boundary.
• Surface error was quantified using RMS and maximum surface error metrics in the extrapolated regions across varying prior sizes (0% to 15% of superior iliac crest).
• The method was validated by simulating missing CT slices and comparing reconstruction accuracy using different prior regions and extrapolation techniques.

Experimental results

Research questions

• RQ1 Can smooth extrapolation using a statistical shape model reduce surface reconstruction error compared to the cut-and-paste method in incomplete pelvic CT scans?
• RQ2 How does the inclusion of superior iliac crest anatomy (as a prior) affect the accuracy of pelvic surface reconstruction?
• RQ3 To what extent does the smooth extrapolation method improve registration accuracy in computer-aided surgical systems for pelvic osteotomies?
• RQ4 What is the impact of varying prior surface regions (e.g., 0% to 15% of superior iliac crest) on the accuracy of SSM-based pelvic surface estimation?
• RQ5 Can the proposed method enable accurate 3D surgical planning with significantly reduced radiation exposure?

Key findings

• The smooth extrapolation method reduced the average RMS surface error by 1.31 mm compared to the cut-and-paste method across all prior configurations.
• The smooth method reduced the average maximum surface error by 3.16 mm compared to the cut-and-paste method, with the minimum improvement still at 0.68 mm in RMS error.
• When the acetabulum and 5% of the superior iliac crest were used as prior, the RMS error in the extrapolated region was 2.21 mm, only 0.60 mm above the SSM’s generalization error of 1.61 mm.
• The inclusion of even 5% of the superior iliac crest significantly reduced surface error in the smooth method, while the cut-and-paste method showed less improvement with added prior.
• The smooth method produced visually continuous and anatomically plausible transitions between known and estimated regions, unlike the discontinuities seen in the cut-and-paste method.
• The SSM demonstrated strong generalization capability, with a root mean square (RMS) surface error of 1.61 mm when using complete anatomy as prior in leave-one-out testing.