Robustness Assessment of Long Axial Field of View PET Performance for Simulated Detector Failures: Insights from a Multiphantom Evaluation

Long axial field of view (LAFOV) PET scanners may exhibit greater tolerance to failing detectors due to the exceptionally high number of detectors distributed along an extended axial FOV. An understanding of the number and location of failing detectors that can be sustained without impairing diagnostic accuracy, is critical to minimize downtime from extensive recalibration or hardware replacement. This study assessed potential degradation of image quality and quantitative accuracy under various simulated detector failure conditions using phantom experiments tailored to the Biograph Vision Quadra LAFOV-PET/CT (Siemens Healthineers). Randomly distributed detector block failures ($\mathbf{n} \boldsymbol{=} \mathbf{1}, \mathbf{8}, \mathbf{1 6}, \mathbf{3 2}, \mathbf{4 8}, \mathbf{6 4}$) and clustered failures simulating a full detector unit loss (16 adjacent blocks) were introduced. A 140 cm long cylindrical phantom, the NEMA IEC body phantom (at center and 45 cm axial offset), and a micro hollow sphere phantom ($\varnothing=7.86 ~\text{mm}$) positioned at 24 locations across a quarter FOV were used to evaluate uniformity, quantification, image noise, and contrast recovery. Despite a true event loss of up to $\mathbf{1 0. 4 \%}$ ($\mathbf{6 4}$ blocks), uniformity was preserved with a maximum activity deviation of 4.8 % in each slice. Activity recovery for the 10 mm IEC phantom sphere showed a maximum reduction of $\mathbf{2. 4 \%}$ at off-center position, while image noise (coefficient of variation) increased moderately from $\mathbf{8. 3 \%}$ to $\mathbf{8. 5 \%}$, and was negligible ($<0.2 \%$) for a single detector unit failure. Clustered failures induced more localized degradation but maintained minimal contrast recovery variation for the micro hollow sphere phantom across the FOV (maximum contrast recovery coefficient difference of $\mathbf{0. 0 5}$). In summary, random failures of up to $\mathbf{6 4}$ blocks and clustered failures of a full detector unit resulted in only minor degradation in image quality and quantification, suggesting that LAFOV-PET systems can tolerate substantial detector block failures without immediate system intervention.