Info
Background:
Single-Photon Emission Computed Tomography (SPECT) is a cornerstone of preclinical research and diagnostic medicine in cardiology, neurology, and oncology; however, its imaging performance is often limited by inherent technical limitations, such as poor sensitivity, low spatial resolution, rigid gantry designs, and limited angular sampling. In parallel, the recent WHO’s WHA78.13 resolution (May 2025) highlighted the urgent need for safe and affordable medical imaging solutions to improve healthcare equity [1-2]. To improve global access, imaging technology should become patient-adaptive, compact, mobile, and user-friendly. The integration of robotics into medical imaging [3] offers a clear pathway to address both challenges: enabling point-of-care systems that improve accessibility, while also supporting the advancement of precision medicine and theranostics through enhanced imaging performance.
Hypothesis/Objective:
We propose a robot-assisted SPECT system designed for adaptive imaging [4-5]. By integrating a compact CZT-based gamma camera with a collaborative robot, the compact and gantry-free architecture facilitates deployment in non-traditional settings like ICUs and operating rooms. The system’s flexibility allows its use both in clinical and preclinical applications, particularly for large animals such as non-human primates, rabbits and dogs, for which dedicated SPECT systems are not available.
By optimizing the system’s configuration based on an initial low-resolution “scout image” of the 3D radiotracer distribution, the system aims to enhance diagnostic accuracy and spatial resolution while reducing acquisition times and radiation doses.
Methods:
The proposed system incorporates a 7-Degrees-Of-Freedom robotic manipulator (LBR MED 14, KUKA, Augsburg, Germany, Fig.A.1). A pixelated CZT detector panel (IDE-Gamma-Spect-3×3, IDEAS, Oslo, Norway, Fig.A.2) is mounted on its end-effector (EE) and it offers a total imaging area of 76.2 mm x 76.2 mm, with pixel pitch of 1.6mm. The detector presents an energy range up to 380 keV suitable for most clinical and preclinical diagnostic (Tc-99m) and theranostics (Lu-177g) radiotracers. A custom multi-pinhole tungsten collimator is designed to cover a 210mm field of view. To complete the system, a 3D structured light camera (2+MR60, Zivid, Oslo, Norway) is used for precise object localization, and a 4D LiDAR (L2, Unitree, Hangzhou, China) for environmental awareness.
Results:
To show the high flexibility of the robotic system, two scanning trajectories were planned around a Jaszczak phantom (Fig.B.1), and in case of pendant breast imaging (Fig.B.2). Energy calibration and uniformity correction of the CZT panel show an energy resolution of 4.9±0.4 keV at 122 keV. Finally, Eye-in-Hand calibration of the structure light camera and LiDAR object detection models enable accurate and fast object localization and environmental awareness.
Imaging simulations were performed to assess the benefits of this scanning flexibility by comparing the robotic system’s non-standard acquisition trajectories against conventional approaches. MLEM reconstruction of a Jaszczak phantom compare a standard cylindrical acquisition – consisting of 30 projections covering 360° – with a robotic-enabled spherical acquisition utilizing 30 projections distributed uniformly around the object. Both geometries maintained the detector-to-object at a constant radius of 252.25mm assuming a 30-minute acquisition time (1 min/position) and 10 mCi of Tc-99m. The system achieves a reconstructed spatial resolution of 4.8 mm (Fig.C.1) and 6.4 mm (Fig.C.2) in noiseless and noisy conditions, respectively. Our results show an increase in the Peak-to-Valley-Ratio (PVR) for the smallest distinguishable rods of about 71% in the noisy case (Fig.C.2).
Conclusion:
While small-animal SPECT imaging is well-established, a significant gap remains for point-of-care clinical systems and specialized platforms for larger preclinical models. We propose a robotic SPECT system that integrates a CZT detector panel to enable adaptive imaging acquisitions. This approach offers a clear pathway toward the innovative and cost-effective solutions requested by the WHO’s call. Our results demonstrate that a robotic-enabled spherical acquisition outperforms standard cylindrical approaches in terms of spatial resolution and quantification.
Image/Figure:
Click to view full size
Click to view full size
Image/Figure Caption:
The robot-assisted prototype in a position around the Flanged Ultra Deluxe Jaszczak Phantom (A.1); Detector panel, detector controller and radioactive source setup for energy calibration (A.2); Trajectory planning showing the robot home configuration and four positions around the Jaszczak phantom (B.1), and four positions in case of pendant breast imaging (B.2); Simulation results comparing image reconstructions of a Jaszczak phantom obtained via cylindrical (left) and spherical (right) acquisition geometries in noiseless (C.1) and noisy (C.2) conditions.
Author
Lorenzo Confalonieri
The University of Texas at Austin