Initial Performance of a Long Axial FOV PET with TOF and DOI capabilities: IMAS system

This paper presents the design, construction, and initial performance evaluation of the IMAS system, a long axial field-of-view PET prototype that uniquely combines time-of-flight and depth-of-interaction capabilities, demonstrating sub-4 mm spatial resolution and improved tumor identification in pilot clinical results despite some performance limitations attributed to data transfer bottlenecks.

The Gist

Imagine you are trying to take a photograph of a bustling city at night. A standard camera (a traditional PET scanner) can only take a picture of one neighborhood at a time. To see the whole city, you have to move the camera, take a picture, move again, and stitch the images together. This takes time, and the more you move, the more likely you are to miss a fleeting moment or get a blurry shot.

Now, imagine a super-camera that is so long it can photograph the entire city in a single snapshot. This is the IMAS system described in the paper. It is a new type of medical scanner called a "Total-Body PET" that can see a patient from head to toe all at once.

Here is a breakdown of how it works and why it's special, using simple analogies:

1. The "Super-Long" Lens (The Axial Field of View)

Most PET scanners are like short tunnels, about the length of a car (25–30 cm). You have to slide the patient in and out to scan different body parts.
The IMAS system is a giant tunnel, 71 cm long (about the height of a tall person).

• The Analogy: Think of a standard scanner as a flashlight that only illuminates one room. The IMAS is a floodlight that illuminates the whole house at once.
• The Benefit: Because it sees everything at once, it is incredibly sensitive. It can detect tiny signals (like a whisper in a noisy room) that other scanners miss. This means doctors can use less radioactive dye, scan patients faster, or see very small tumors that were previously invisible.

2. The "3D Glasses" (Depth-of-Interaction or DOI)

This is the IMAS system's secret superpower.
In a standard scanner, the "film" (detectors) is a grid of tiny crystals. When a particle hits the edge of the scanner, the system gets confused about exactly where it hit, kind of like looking at a reflection in a funhouse mirror. This causes the image to get blurry at the edges.
The IMAS system uses semi-monolithic slabs (thick blocks of special crystal) instead of tiny tiles. It has a special "brain" (software) that can tell exactly how deep a particle penetrated the block.

• The Analogy: Imagine trying to guess where a raindrop hit a thick window pane. A standard scanner just guesses the surface. The IMAS system has 3D glasses that tell it exactly how deep the drop went.
• The Result: Even if a tumor is near the edge of the scanner (far from the center), the image remains sharp and clear. The paper shows that without this feature, the system would be "blind" to the exact location of a tumor by several millimeters—a huge difference when looking for cancer.

3. The "Speed Trap" (Time-of-Flight or TOF)

PET scanners work by detecting two particles flying in opposite directions. To find out where they came from, the scanner measures the tiny fraction of a second it takes for them to hit the detectors.

• The Analogy: Imagine two people shouting "Hello!" at the same time. If you are standing in the middle, you hear them at the same time. If you are closer to one, you hear that one first. The IMAS system is so fast it can measure that split-second difference with incredible precision (about 560 picoseconds).
• The Benefit: This acts like a speed trap for the particles, allowing the computer to pinpoint exactly where the signal originated, making the image much clearer and reducing "noise" (static).

4. The "Traffic Jam" (The One Weakness)

The system is amazing at seeing things, but it has a slight bottleneck in how it moves data.

• The Analogy: Imagine a Ferrari engine (the detector) that can go 200 mph, but it's connected to a tiny, narrow pipe (the data transfer cable) that can only handle 50 mph. The engine is working hard, but the data gets stuck in traffic.
• The Reality: The paper admits that when the scanner is very busy (high activity), the data transfer slows down, and the system can't count as many events as it theoretically should. However, the authors say this is a temporary "traffic jam" they are fixing by adding more lanes (better computers) to the system.

5. The "Real-World Test" (Clinical Results)

The team tested this new scanner on a real patient and compared it to a standard, high-end scanner.

• The Result: The IMAS scanner found tumors that the standard scanner missed or showed as blurry blobs. Specifically, it clearly identified a cluster of lesions near the patient's armpit that looked like a fuzzy smudge on the other machine.
• The Takeaway: It's like upgrading from a standard definition TV to a 4K Ultra HD TV. The details are sharper, the contrast is better, and the doctor can make a more confident diagnosis.

Summary

The IMAS system is a prototype "Total-Body" PET scanner that combines three major upgrades:

1. Longer View: Sees the whole body at once.
2. 3D Depth: Knows exactly how deep particles hit, keeping images sharp everywhere.
3. Super Speed: Measures time so precisely it cuts through the noise.

While it currently has a minor data traffic issue, it represents a massive leap forward in medical imaging, promising to find diseases earlier, with less radiation, and with much greater clarity than ever before.

Technical Summary

Here is a detailed technical summary of the paper "Initial Performance of a Long Axial FOV PET with TOF and DOI capabilities: IMAS system."

1. Problem Statement

Long Axial Field of View (LAFOV) or Total-Body (TB) PET scanners offer significant clinical advantages over conventional Whole-Body (WB) PET systems, including increased sensitivity, reduced patient dose, shorter scan times, and the ability to image multiple organs simultaneously. However, existing commercial TB-PET systems (e.g., Siemens Vision Quadra, United Imaging uExplorer/Panorama) typically rely on pixelated crystal arrays that lack Depth-of-Interaction (DOI) capabilities.

The absence of DOI leads to parallax errors, causing spatial resolution to degrade significantly at the edges of the Field of View (FOV). While DOI is standard in pre-clinical imaging, its implementation in clinical TB-PET has been limited. Furthermore, achieving high timing resolution (Time-of-Flight or TOF) in large, complex detector geometries remains a technical challenge. The IMAS project aims to address these gaps by developing a prototype TB-PET scanner that simultaneously integrates TOF and DOI capabilities using a novel semi-monolithic scintillator design.

2. Methodology

System Design and Geometry

• Scanner Architecture: The IMAS system features a 71 cm axial FOV with an 82 cm bore diameter. It consists of 5 "super-rings" (SR), each 10 cm long, separated by 5 cm gaps.
• Detector Modules: The system uses LYSO semi-monolithic scintillator slabs (3 mm × 25 mm × 20 mm). Arrays of 1×8 slabs are coupled to 8×8 Silicon Photomultiplier (SiPM) arrays (Hamamatsu S13361-3050AE-08).
• Signal Reduction: A proprietary readout scheme reduces the 64 signals from each SiPM array to 16 outputs (8 row + 8 column channels) without significantly compromising 3D positioning or timing accuracy. This reduces the total channel count to 30,720.
• Electronics: Data acquisition is handled by PETsys electronics (TOFPET2 ASICs). To manage high count rates, a global energy threshold was implemented at the FPGA level to filter out low-energy noise before data transmission, and a second DAQ board was added to increase data-handling capacity.

Calibration and Reconstruction

• Positioning Algorithms: To determine the $y$ (monolithic) and $z$ (DOI) coordinates, the system employs Neural Networks (MLPs). Due to the time-consuming nature of calibrating every module individually, a reference module was scanned with a slit collimator and $^{22}$Na sources. The data was used to train 24 different MLP categories to account for signal variations (e.g., dead channels, gain differences) across the system.
• Timing Calibration: An iterative algorithm was used to correct intrinsic timing offsets (skew correction) for each detector element, converging when centroid errors fell below 10 ps.
• Image Reconstruction: A GPU-based List-Mode Ordered Subset Expectation Maximization (OSEM) algorithm was used (5 iterations, 5 subsets). The reconstruction incorporates DOI correction and TOF information (discretized into 7 bins with an effective 750 ps FWHM).

Performance Evaluation

The system was evaluated following the NEMA 2018 protocol, measuring:

• Spatial resolution (at various radial offsets up to 30 cm).
• Sensitivity and Noise Equivalent Count Rate (NECR).
• Coincidence Time Resolution (CTR) and Energy Resolution.
• Image quality using a torso phantom.
• Clinical Pilot: A comparative study was conducted on a 45-year-old female patient with multiple lesions, comparing IMAS images against a commercial Philips Gemini TF-64 PET/CT.

3. Key Contributions

• First Simultaneous TOF and DOI in TB-PET: IMAS is the first reported TB-PET prototype to successfully integrate both TOF and DOI capabilities in a clinical setting.
• Novel Detector Geometry: The use of semi-monolithic slabs with a 64:16 signal reduction readout allows for cost-effective manufacturing (using gaps between rings) while maintaining high spatial uniformity.
• Machine Learning Calibration: The implementation of category-specific MLPs allows for robust position reconstruction across the entire system, compensating for hardware imperfections without requiring exhaustive individual calibration of every module.
• Clinical Validation: The paper provides the first pilot clinical results demonstrating the diagnostic advantages of this specific architecture.

4. Key Results

• Spatial Resolution: The system achieved a homogeneous spatial resolution across the entire 71 cm FOV.
  • At the center: 3.3 ± 0.5 mm (radial).
  • At 30 cm off-radial (near the edge): 3.53 mm (radial).
  • Significance: Without DOI correction, the resolution at 30 cm degrades significantly (showing a 6 mm shift). With DOI, the degradation is minimal (<0.2 mm), outperforming existing commercial systems which typically degrade to 7–9 mm at similar offsets.
• Sensitivity: Measured at 56.54 cps/kBq (peak sensitivity 7.6%), which is 4–5 times higher than conventional WB-PET scanners.
• Timing Resolution (CTR):
  • With a small $^{22}$Na source: 560 ps FWHM.
  • At the NECR peak (3.26 kBq/mL): 690 ps FWHM.
  • Note: The authors note that individual detector blocks achieved ~277 ps CTR, suggesting further system-wide improvements are possible with refined calibration.
• Count Rate Performance (NECR): The peak NECR was 79 kcps at 3.26 kBq/mL.
  • Limitation: This value is significantly lower than other LAFOV systems (which often exceed 1000 kcps). The authors attribute this to a data transfer bottleneck between the system and the acquisition workstation, rather than a fundamental detector limitation.
• Clinical Results: In the pilot patient study, IMAS demonstrated superior tumor identification compared to the commercial Philips scanner. Specifically, a multi-centric lesion in the axilla was clearly resolved by IMAS but appeared blurred in the conventional scan. A lesion 16 cm off-center showed a 3 mm mispositioning in the non-DOI reconstruction, which was corrected by IMAS.

5. Significance and Conclusion

The IMAS system represents a significant proof-of-concept for the next generation of Total-Body PET scanners. By successfully combining TOF and DOI capabilities, it solves the parallax error problem that plagues current long-axial scanners, ensuring high spatial resolution even at the extreme edges of the FOV.

While the current prototype faces limitations in count rate performance due to data bandwidth bottlenecks, the authors propose a roadmap to distributed acquisition to resolve this. The clinical pilot results confirm that the homogeneous spatial resolution and high sensitivity of the IMAS system lead to improved diagnostic confidence, particularly for small or off-center lesions. This work validates the viability of semi-monolithic scintillators and multiplexed readouts for cost-effective, high-performance total-body imaging.