Why is medical image transfer still so slow?

By Pascal Looijé, CEO & Co-founder at Interoplab

Medical image transfer sometimes takes unacceptably long. A referral waits, a clinician waits, the patient waits. Yet the infrastructure has been there for a while. What goes wrong?

Push and pull: Two image transfer models

There are two basic models for image transfer:

• Push: the sender chooses which images to send and actively transfers them to the recipient
• Pull: the recipient requests what’s needed from the sender’s storage

In practice, both models are active — sometimes simultaneously. And that’s exactly where things can go wrong.

An image transfer in three steps

Retrieving images from an external institution involves three sequential transactions:

1. Fetch metadata — what studies are available?
2. Fetch the DICOM object — technical specifications of the image
3. Download the actual image data — the pixels themselves

Step 3 is by far the slowest. And step 3 contains most bottlenecks.

Possible causes of delay

• Network capacity — insufficient bandwidth at a given moment
• Adapter layer overhead — translation software between systems adds latency
• Compression and decompression — image compression reduces file size but costs processing time
• Deliberate throttling — some systems limit upload speed to protect other users
• Overload — a component in the chain is overloaded at peak times
• Inconsistent protocols — without agreements on retrieval method, every system works differently

What you can do

The solution is rarely purely technical. A holistic approach pays off: examine all layers of the interoperability model, test realistically (not just in ideal conditions), and adjust expectations based on what practice shows.

Interoplab conducts performance analysis and simulation testing to identify bottlenecks — without impacting the end-user experience during implementation.

Experiencing slow image transfer and need to find the bottleneck? Get in touch →