Health & Medical imaging
Encrypted PACS/DICOM + diagnostic AI
FPGA encrypts DICOM streams, CPU orchestrates HDS patient access, GPU runs segmentation/classification AI inference.
FPGA
DICOM encryption + anonymization
Workloads
- ▸ Encrypted DICOM Store / Query
- ▸ Inline de-identification
- ▸ JPEG2000 transcoding
Performance
10 Gbps imaging
CPU
HDS + FHIR orchestration
Workloads
- ▸ Patient identity vigilance
- ▸ Encrypted FHIR R4 API
- ▸ HDS journal
Performance
1 k requests / s
GPU
Diagnostic AI
Workloads
- ▸ Tumor segmentation (nnU-Net)
- ▸ Pathology classification
- ▸ Radiomic feature extraction
Performance
< 30 s per CT study
Multi-agent scenario
A CT scan arrives encrypted from the hospital: FPGA decrypts and anonymizes, CPU verifies patient consent, GPU segments the tumor and returns the re-encrypted AI report to the radiologist.