Application of Gated Recurrent Units for CT Trajectory Optimization

TL;DR

The paper addresses CT trajectory optimization in dual-robot systems by applying a Gated Recurrent Unit (GRU) to select projection views based on projection-derived metrics. It integrates absorption, pixel intensity, contrast-to-noise ratio, and data-completeness considerations, including Tuy's data sufficiency condition and unit-sphere coverage, to drive sequential projection selection with a binary cross-entropy loss. In simulations, the GRU-optimized trajectory outperformed a circular trajectory, increasing SSIM from 0.381 to 0.491 and CNR from 6.97 to 9.08 for a center VOI, demonstrating improved image quality with fewer projections. The work suggests GRU-based trajectory optimization can enable more efficient, higher-quality robotic CT imaging, though broader validation and metric-weighting studies are needed.

Abstract

Recent advances in computed tomography (CT) imaging, especially with dual-robot systems, have introduced new challenges for scan trajectory optimization. This paper presents a novel approach using Gated Recurrent Units (GRUs) to optimize CT scan trajectories. Our approach exploits the flexibility of robotic CT systems to select projections that enhance image quality by improving resolution and contrast while reducing scan time. We focus on cone-beam CT and employ several projection-based metrics, including absorption, pixel intensities, contrast-to-noise ratio, and data completeness. The GRU network aims to minimize data redundancy and maximize completeness with a limited number of projections. We validate our method using simulated data of a test specimen, focusing on a specific voxel of interest. The results show that the GRU-optimized scan trajectories can outperform traditional circular CT trajectories in terms of image quality metrics. For the used specimen, SSIM improves from 0.38 to 0.49 and CNR increases from 6.97 to 9.08. This finding suggests that the application of GRU in CT scan trajectory optimization can lead to more efficient, cost-effective, and high-quality imaging solutions.

Figures (6)

• Figure 1: A great circle formed by normal vectors of all planes containing the X-ray path $SP$liu2012completeness.
• Figure 2: Scheme of a single GRU unit. ${x_t}$ denotes the input data at time step $t$, ${r_t}$ represents the reset gate at time step $t$, and ${z_t}$ corresponds to the update gate at time step $t$. ${h_t}_{ - 1}/{h_t}$ indicates the hidden state at time step $t-1/t$, while ${{h'}_t}$ represents the current memory of the unit at time step $t$kostadinov2017understanding.
• Figure 3: Diagram of the selection process for the next projection through the GRU network
• Figure 4: Illustration of the data processing in GRU network.
• Figure 5: Test specimen overview. (a) Carbon specimen illustration. (b) Simulated projection highlighting the VOI in red.