To Pre-fetch or Not to Pre-fetch, That is the Question

Optimizing medical image display speed has always been important, and is critical in today's world of radiologist shortages and extreme production pressure.

The fastest possible way to present an exam group (an imaging exam plus relevant priors) is to have it on your computer's local hard drive before you access the exam. PACS systems since the late 1980's implemented pre-fetching for this reason. A number of strategies can be used to anticipate which exams will be read by which users so that the proper exams are pre-fetched to the proper computer. Rules-based image routing was first used decades ago; more recently, automated exam assignment algorithms have guided pre-fetching. If you are reading on a PACS today, access an exam, and the current plus prior images all appear in a few seconds, the images had likely been pre-fetched.

In addition to optimal speed, the other big benefit of local caching is load balancing, because each user's local computer performs most of the work of image display. The historical downside of local caching is that PHI is sent to remote locations, posing a security risk. Recent advances in encryption and the way browsers manage memory can eliminate this risk.

The next fastest way of displaying images, server-side rendering, was introduced about 20 years ago.  In this method, exam groups are rendered on a remote local or web server.  With this method, much smaller rendered image files (usually JPEGS or PNGs) are streamed. This method is slower than viewing pre-fetched exam groups, but faster that client-side rendering of exams that are accessed before they are pre-fetched. Another downside of server-side rendering is that all the image processing is done on the server, which has to be carefully scaled as simultaneous users are added.

Wouldn't it be great to have the optimal speed and load-balancing of client-side rendering with pre-fetching, but automatically switch to server-side rendering when the exams have not been pre-fetched?  We call this "hybrid viewing" and it's available now.  Suddenly, the answer to the question, "To Pre-cache or Not to Pre-cache?", is "Let us automatically select the optimal solution for the current requested images."