Impact of CT scan parameters on deformable image registration accuracy using deformable thorax phantom

Ryutaro Ikeda, Noriyuki Kadoya, Y. Nakajima, Shin Ishii, Takayuki Shibu, Keiichi JinguJournal of Applied Clinical Medical Physics 2023 · DOI

TL;DR

This study investigated how CT scan parameters, specifically tube voltage and reconstruction algorithms, affect the accuracy of deformable image registration (DIR) in thoracic imaging. They found that these parameters can moderately impact DIR accuracy, with reconstruction algorithms showing a more pronounced effect, highlighting the need to consider acquisition settings for reliable DIR in radiotherapy.

What Problem Does This Paper Address?

Deformable Image Registration (DIR) is a crucial technique in modern radiotherapy, enabling applications like dose accumulation and automatic contour propagation. However, the accuracy of DIR can vary significantly depending on the software used and the institution's practices. While guidelines exist for DIR quality assurance, they often focus on software configuration and virtual phantoms. This paper addresses the gap in understanding how fundamental CT acquisition parameters, which are easily adjustable and impact image intensity, influence DIR accuracy. Specifically, they aim to quantify the effect of tube voltage and reconstruction algorithms on DIR performance in a realistic thoracic scenario.

Methods

The researchers utilized a custom-developed dynamic deformable thorax phantom ("Dephan") designed to simulate lung motion and anatomy. This phantom features an acrylic cylinder representing the chest wall and an inner component with deformable sponges containing Lucite beads and nylon cross-wires to mimic vascular and bronchial structures.

To evaluate DIR accuracy, they acquired peak-exhale and peak-inhale CT images of the phantom using a specific CT scanner (Aquilion LB). They systematically varied two key CT parameters:

  • Tube Voltage: Tested at four levels (80, 100, 120, and 135 kV).
  • Reconstruction Algorithms: Employed six different algorithms (FC11, FC13, FC15, FC41, FC44, and FC52), each with different characteristics (e.g., sharpness, intended for specific anatomical regions).

Intensity-based DIR was performed between the peak-exhale and peak-inhale image pairs using MIM Maestro software. The primary metric for evaluating DIR accuracy was the Target Registration Error (TRE), calculated using fiducial markers (beads and cross-wires) embedded in the phantom. They also analyzed the standard deviation (SD) of CT values within a region of interest to assess image quality. Statistical analysis (ANOVA) was used to determine the significance of differences in TRE.

Datasets

The study did not use pre-existing datasets. Instead, they generated their own dataset by acquiring CT images of a custom-built deformable thorax phantom under various CT scan parameter settings. This phantom, with its embedded fiducial markers, served as a novel benchmark for evaluating DIR accuracy in a controlled, reproducible manner.

Key Insights and Ideas

  • CT Scan Parameters Matter for DIR: The study definitively shows that seemingly minor adjustments to CT acquisition parameters can have a measurable impact on DIR accuracy. This is a critical finding for anyone relying on DIR in clinical workflows.
  • Reconstruction Algorithms are Key Drivers: The research highlights that reconstruction algorithms, which directly influence image sharpness and noise, have a more significant impact on DIR accuracy than tube voltage in this specific thoracic phantom setup.
  • Image Noise and Blurring as DIR Hindrances: The findings suggest that increased image noise (higher SD values) and blurring, often associated with certain reconstruction algorithms, can degrade DIR performance by making it harder to accurately identify corresponding anatomical points.
  • Phantom Design for Realistic Evaluation: The use of a dynamic, deformable phantom with fiducial markers provides a more realistic end-to-end test of DIR accuracy compared to purely virtual phantoms or static phantoms.

Key Results

  • The Target Registration Error (TRE) varied across different combinations of tube voltage and reconstruction algorithms, with a range from 3.96 mm to 6.96 mm, a difference of up to 3.0 mm.
  • When varying tube voltage (at 120 kV), the average TRE ranged from 4.44 mm to 5.69 mm.
  • When varying reconstruction algorithms (at 120 kV), the average TRE ranged from 4.26 mm to 4.59 mm.
  • Statistical analysis indicated a significant difference in TRE among tube voltages (p = 0.002) but not among reconstruction algorithms (p = 0.609) when considered independently. However, the interaction between these parameters was crucial, as evidenced by the overall TRE range.
  • The FC52 reconstruction algorithm, designed for lung regions, resulted in the largest TRE, potentially due to a trade-off between higher resolution and increased image noise.

Limitations and Future Work

The authors acknowledge several limitations:

  • Single CT Scanner: The study was conducted on only one CT scanner, and results might differ with other scanner models.
  • Single DIR Software: Only one DIR software (MIM Maestro) was evaluated. Different DIR algorithms and software packages may yield different results.
  • Thoracic Phantom Specificity: The phantom primarily simulates high-contrast thoracic tissues. The impact of CT parameters on DIR accuracy for low-contrast tissues (e.g., parotid gland) remains to be investigated.
  • Limited Parameter Scope: Only tube voltage and reconstruction algorithms were tested. Other parameters like pitch and slice thickness could also influence DIR accuracy.

Future work will involve investigating these other CT parameters and evaluating DIR accuracy on different CT scanners and with various DIR software. They also plan to explore the impact of these parameters on DIR accuracy for low-contrast tissues.

Summary updated

Abstract

The purpose of this study was to evaluate the deformable image registration (DIR) accuracy using various CT scan parameters with deformable thorax phantom. Our developed deformable thorax phantom (Dephan, Chiyoda Technol Corp, Tokyo, Japan) was used. The phantom consists of a base phantom, an inner phantom, and a motor-derived piston. The base phantom is an acrylic cylinder phantom with a diameter of 180 mm, which simulates the chest wall. The inner phantom consists of deformable, 20 mm thick disk-shaped sponges with 48 Lucite beads and 48 nylon cross-wires which simulate the vascular and bronchial bifurcations of the lung. Peak-exhale and peak-inhale images of the deformable phantom were acquired using a CT scanner (Aquilion LB, TOSHIBA). To evaluate the impact of CT scan parameters on DIR accuracy, we used the four tube voltages (80, 100, 120, and 135 kV) and six reconstruction algorithms (FC11, FC13, FC15, FC41, FC44, and FC52). Intensity-based DIR was performed between the two images using MIM Maestro (MIM software, Cleveland, USA). Fiducial markers (beads and cross-wires) based target registration error (TRE) was used for quantitative evaluation of DIR. In case with different tube voltages, the range of average TRE were 4.44-5.69 mm (reconstruction algorithm: FC13). In case with different reconstruction algorithms, the range of average TRE were 4.26-4.59 mm (tube voltage: 120 kV). The TRE were differed by up to 3.0 mm (3.96-6.96 mm) depending on the combination of tube voltage and reconstruction algorithm. Our result indicated that CT scan parameters had moderate impact of TRE, especially for reconstruction algorithms for the deformable thorax phantom.
BibTeX
@article{ikeda2023impact,
  author = {Ryutaro Ikeda and Noriyuki Kadoya and Y. Nakajima and Shin Ishii and Takayuki Shibu and Keiichi Jingu},
  title = {{Impact of CT scan parameters on deformable image registration accuracy using deformable thorax phantom}},
  journal = {Journal of Applied Clinical Medical Physics},
  year = {2023},
  doi = {10.1002/acm2.13917},
}
Figure