Insights

Insights May 2012

Moore’s Law in Radiology? Technology Makes Radiologists’ Work Slower Not Faster - but more Valuable

Benjamin Fine, Sean Symons and Mark Prieditis

Gordon Moore, co-founder of Intel, once opined that the number of transistors that can be crammed onto a microchip will double every 2 years1. As transistor density increases, cost per transistor decreases.

Will Falk proposes in these pages2 that Moore’s Law applies to radiology (later echoed by the Drummond report, and now a favourite line of Deb Matthews): radiologists are cramming more image interpretation into the same day (really, same year – thus increased total billing). As a result, he argues, price per image interpretation should come down.

Fortunately, it has. Accelerating imaging technology has made each CT or MR examination much more powerful in answering diagnostic questions. Radiologists interpret these increasingly complex studies – more images for the same price – slowing their work, but benefiting patients.

To illustrate, consider a radiologist reading head CTs (one of the workhorse studies of a radiology department) in 1999, the year one of our hospitals switched from the lightbox to its digital equivalent, PACS. The number of images this radiologist had to examine to interpret the study was 20. For the same study in 2012, the radiologist has to examine 65 images3. The fee for this service has risen less than inflation4. Therefore, the cost to the taxpayer per image interpretation has declined to one third of the 1999 level; each radiology report summarizes 3 times the number of images for the same cost. Running the same scenario with abdominal CT and head MRIs yields similar results5. And these figures are low-ball estimates which exclude complex studies which can have over 1000 images today6. 

While not Moore’s Law - the number of images interpreted per government dollar doubled in about 8 years, not 2 – the government has benefitted from an exponential increase return on investment. 

Advancing MRI and CT technology has enabled this cost per image reduction. A CT scanner now produces more images as it slices the human body into thinner slices; increased computing power (real Moore’s Law in action) allows rapid reconstruction of images in three dimensions. MRI spins protons in ever-fantastical new ways to allow sets of images (sequences is the technical term) to depict recent strokes, blood or flow in vessels. These technologies have increased the diagnostic power of imaging, but also the number of images a radiologist must interpret for the same fee. 

While cost per image has decreased benefitting taxpayers, the total number of images demanded by referring physicians grows at a near Moore’s Law rate. At our site, the total number of images the radiology department must interpret on a daily basis has doubled every 4.5 years.

That growing demand is what is driving rising diagnostic imaging costs despite a cost-per-image decrease. Diagnostic imaging, with all these technology-driven images, has provided referring physicians with immense amount of diagnostic power. As a result, referring physicians are demanding more images. The number of CT and MRI scans ordered by surgeons, family physicians, and other specialists has skyrocketed. Just as Intel sold a greater number of chips as a result of cheaper computing power, radiologists are working longer hours reading more CT and MRIs to meet growing demand for cheaper interpretation.

While not quite Falk’s Moore’s Law, diagnostic imaging is experiencing exponential decline in cost per image. Clinicians and patients have benefitted from the ability of new technology to provide diagnostic clarity with at least three times the number of images for the same price as in 1999 – akin to more transistors on a microchip. As a result, clinicians are demanding more microchips. The challenge in Ontario is how to pay for these more advanced chips – or perhaps how to appropriately manage exponentially growing demand.

Notes

  1. Gordon Moore, co-founder of Intel, first opined that transistors per integrated circuit would double every year thanks to improving technology, then revised the number to two years – a trend that has largely held for nearly 50 years.
  2. https://www.longwoods.com/content/22475
  3. Data taken from Sunnybrook PACS comparing one week in 1999, the year PACS was introduced, and one week in 2012.  Sunnybrook is a large academic hospital with trauma and cancer centers. 
  4. Costs of imaging studies from MOH schedule of benefits for physician services 1999-2011.
  5. Abdominal CTs and head MRIs both now routinely include over 400 images. The cost per image interpreted has dropped 3 and 3.5 fold drop in cost per image over time 12 years. Image per dollar doubled in 8 and 7 years.
  6. Number of images per study estimated from lower of median or mode number of images. The purpose was to exclude massive studies (eg CT perfusion performed on a stroke has over 3000 images) that would inflate numbers and more accurately reflect practice across the province (less fancy stuff). As a result, these numbers are low-ball estimates. Singh et al paper shows overall images per study have increased 5 fold for CT and 7 fold for MRI.

About the Author(s)

Ben Fine is a radiology resident in Toronto. Sean Symons and Mark Prieditis are radiologists in Toronto. All 3 hold degrees in engineering.

References

See additional essays and commentaries on this subject at these links:

Prescription for Sustainable Medical Imaging in Canada? Ask the Experts: The Radiologists by Deljit Dhanoa

 

Comments

Jane Sauder wrote:

Posted 2012/05/24 at 04:42 PM EDT

Sounds about right.

 

Anton Hart wrote:

Posted 2012/05/28 at 01:42 PM EDT

Another response. Prescription for Sustainable Medical Imaging in Canada? Ask the Experts: The Radiologists

Deljit Dhanoa
http://www.longwoods.com/content/22897

 

Note: Please enter a display name. Your email address will not be publically displayed