Purpose or Learning Objective
To evaluate and compare the efficiency of Grayscale Softcopy Presentation State (GSPS) versus Secondary Capture (SC) for storing and visualizing AI outputs in DICOM format, with a focus on storage efficiency, clinical usability, and scalability across modalities.
Methods or Background
In clinical AI workflows, visual outputs like ROIs, annotations, and labels are typically stored using SC or GSPS. SC creates a new image series by duplicating the original and embedding visual overlays, increasing storage load and occupying full viewports in PACS viewers. GSPS, however, references the original DICOM images and stores overlays as lightweight, toggleable vector layers. This study compares the storage and transfer impact of both formats across four modalities by simulating AI outputs for 1000 studies per type. Transfer time per study was calculated using a 0.5 Gbps network assuming sequential upload of a single output file.
Results or Findings
GSPS provides significant benefits over SC in both storage and transfer times. Storage is reduced by up to 49.5% for X-ray, 44.4% for CT and MRI, and 16.7% for Mammography. Transfer times are drastically faster, with GSPS reducing transfer time by up to 99.9%, compared to SC’s seconds-long transfers for larger studies like CT and MRI.
Conclusion
GSPS is a highly optimized format for storing and transmitting AI-generated overlays in radiology. Compared to SC, it significantly reduces PACS storage requirements and improves network performance—even at lower bandwidths. It also enhances usability by enabling overlays to be toggled on/off without affecting the primary diagnostic image or taking up additional viewports. These benefits make GSPS the preferred choice for modern AI-driven radiology environments.
