Blackford comes from outer space, almost literally. Their origins are in the astrophysics world, and their break-through technology is called MOPED:
Blackford Analysis’s core technology is MOPED, an algorithm developed in astrophysics to tackle analysis of immense datasets. The patented approach involves compressing the huge datasets while retaining all information needed to solve a problem – allowing speedups of many orders of magnitude over traditional techniques.So what does this have to do with imaging? It seems that we can apply the algorithm to matching volumes, such as two CT scans!
The technology comes from astronomical surveys, where instruments capture gigabytes of images per hour. This information is generally interpreted by comparison with models, essentially complex formulae driven by a set of parameters, which reproduce the observations.
Parameters might be the mass of a galaxy, or the distance it is from Earth, and there will be some combination of parameters that produces a modelled image that is very close to that actually seen through the telescope. Situations like this are called ‘parametric modelling problems’.
MOPED’s particular ability when solving such problems is that it speeds up the step that determines how well a given combination of parameters recreates the image.
After the initial compression, the time taken for each combination changes from being set by the number of pixels to being set by the number of parameters. If 10 parameters were to be determined by an image taken by a modern digital camera with 12 million pixels, the calculation would be more than one million times faster.
This means that problems that were too slow become possible, often solvable in real-time. As datasets become larger, and the cost of the hardware resources required to tackle them rocket, the case for MOPED is even more compelling: the powerful algorithm vastly reduces that hardware cost.
Developed by Blackford Analysis, the medical imaging technology makes it possible for radiologists to anatomically link small features such as lung nodules between studies for the first time within the PACS.A video is worth at least 10,000 words:
While the radiology imaging software will also align CT and MR from any part of the body, instant anatomical alignment in the chest is a major breakthrough, given respiratory movement and the requirement for a deformable registration.
Blackford Analysis’ technology greatly reduces the time it takes to compare current and prior studies, a drain of radiologist’s time as volumetric datasets increase in size and complexity and become ever more commonplace.
A key advantage of the software is that it designed for integration in the existing PACS environment so radiologists can use it without having to interrupt their natural review processes by moving to another workstation.
Crucially, the alignment is achieved without any alteration of the raw slice data so radiologists don't need to worry about the authenticity of what they are reviewing.
Blackford is thinking outside the box. In all of the other registration programs I've seen, an attempt is made to match the entire volume of the old study to the new. Because patients are not rigid (their bodies aren't, anyway) this doesn't work so well. Some software will attempt to distort the data-set to achieve a fit, which could conceivably distort the findings as well. Blackford takes the novel and proper approach of an instantaneous point-to-point mapping, finding the exact spot on the old study that I'm seeing in the new exam. Brilliant! That's really all we need in the end, isn't it? And it does appear to work quite well. And to have it actually integrated into the PACS viewer would be incredible.
I hope to meet with the folks from Blackwell at RSNA and see the thing live and in action.
PACS vendors: you WANT this in your product. You really do. Jump on it now.
Contact r
info@blackfordanalysis.com
B
2 comments :
Non-rigid registration is certainly not "novel". Fusion7D was doing this in 2004. "Some software will attempt to distort the data-set to achieve a fit, which could conceivably distort the findings as well" doesn't describe what deformable registration does.
Non-rigid registration methods generate a transformation between the source and target images, which is a "point-to-point mapping" (also known as a displacement field). You can then use that transformation to deform the source image such that it overlays the target image, deform the target image so that it overlays the source, transform findings on the source to their new shape on the target - or just do crosshair correlation.
Mirada XD has been doing exactly what this product has been doing for several years - and much more. There is no "alteration of the raw slice data"; if you want crosshair correlation through non-rigid transformation, we can give you that. We can even do it cross-modality, across ten timepoints and multiple MR sequences. Come and see us at RSNA and we'll show you the state of the art.
David McCabe
Mirada Medical Ltd
Dr. Dalai – Thank you for your review! You’ve really captured the essence of what we’re trying to achieve and I look forward to meeting you in person at RSNA. It’s been a long journey from astronomy technology to medical imaging, and it’s thanks to massive amounts of interaction with radiologists like you that we’ve been able to really understand the clinical applications that drive our development and focus.
David - I don’t think anyone would argue with you that Mirada has some great technology.
If readers of the blog would like to find out how our approach is different, I look forward to meeting with them at RSNA – please get in touch via our website, http://www.blackfordanalysis.com/applications/medical/, to arrange a meeting.
Post a Comment