Generative Adversarial CT Volume Extrapolation for Robust Small-to-Large Field of View Registration

Andrei Puiu,Sureerat Reaungamornrat,Constantin Suciu,Florin Cristian Ghesu,Lucian Mihai Itu,Tommaso Mansi,Thomas Pheiffer

doi:10.3390/app12062944

Abstract

Intraoperative Computer Tomographs (iCT) provide near real time visualizations which can be registered with high-quality preoperative images to improve the confidence of surgical instrument navigation. However, intraoperative images have a small field of view making the registration process error prone due to the reduced amount of mutual information. We herein propose a method to extrapolate thin acquisitions as a prior step to registration, to increase the field of view of the intraoperative images, and hence also the robustness of the guiding system. The method is based on a deep neural network which is trained adversarially using self-supervision to extrapolate slices from the existing ones. Median landmark detection errors are reduced by approximately 40%, yielding a better initial alignment. Furthermore, the intensity-based registration is improved; the surface distance errors are reduced by an order of magnitude, from 5.66 mm to 0.57 mm (p-value = 4.18×10−6). The proposed extrapolation method increases the registration robustness, which plays a key role in guiding the surgical intervention confidently.

Highlights

Over the past years, the use of medical imaging in computer aided interventions has become more and more popular, supporting clinicians in their workflow and reducing the procedural associated risks [1].This paper is focused on increasing the trustworthiness of liver needle therapies such as Radiofrequency Ablation (RFA) or biopsy, where real time imaging plays a main role in guiding the intervention confidently
We found that our method reduced the median landmark detection error by a very large margin, leading to a superior feature-based registration
We proposed a method for improving the performance of intraoperative image registration by expanding the field of view of thin slabs, enhancing the context information required for the matching process

Summary

Introduction

The use of medical imaging in computer aided interventions has become more and more popular, supporting clinicians in their workflow and reducing the procedural associated risks [1]. This paper is focused on increasing the trustworthiness of liver needle therapies such as Radiofrequency Ablation (RFA) or biopsy, where real time imaging plays a main role in guiding the intervention confidently. It is well known that there is a trade-off between radiation dose, acquisition time and image quality, during such surgical interventions all procedures must be carried out as quickly and accurately as possible. Registration is a technique used to align two images with respect to the patient’s internal structures. Having a reference and a template image R, T : Rd → R, registration objective is to find a transformation φ : Rd → R such that R ≈ T ◦ φ [3]

Methods

Discussion

Conclusion