Not long ago, the ability of radiologists to mentally assemble 2D images into a 3D jigsaw reigned supreme. Computer tools for doing so were available, but the general feeling was why bother? CT changed all that. Spurred by slice overload, the flat world of radiology has begun to embrace 3D, not because it can but because it must. And this may be only the beginning.
Not long ago, the ability of radiologists to mentally assemble 2D images into a 3D jigsaw reigned supreme. Computer tools for doing so were available, but the general feeling was why bother? CT changed all that.
Spurred by slice overload, the flat world of radiology has begun to embrace 3D, not because it can but because it must. And this may be only the beginning.
Reconstructing three-dimensionally offers enormous advantages in that the data themselves are placed in 3D space for anyone and, maybe in the future, anything to evaluate. The "thing" would be software - pattern matching algorithms to be exact.
These algorithms are already creating a whole new world of volumetric processing, one that for now has little to do with radiology but could form the basis for extraordinary possibilities in the future. This world is exemplified by Polar Rose, a Swedish startup that may soon introduce a Web-based search engine that can match people viewed three-dimensionally with 2D photographs.
Remarkably, this search engine creates 3D models of faces from 2D photographs. The computer calculates where the camera was positioned in relation to the person and then does the geometry. This ability to seek and find common ground between 2D and 3D objects may open the door to possibilities in medicine, as it demonstrates an entirely new way of thinking about and handling data.
At present, volumetric reconstructions are seen as a shorthand means for homing in on pathologies contained in 2D slices. Data storage limitations often lead sites to dump these volumetric scans. But what if storage weren't an issue?
Polar Rose's Web-based search engine looks for volumetric objects rather than text. The development of this engine is based on the assumption that storage constraints will one day be a thing of the past.
Security is one problem driving this technology. Face recognition algorithms built into a search engine might be used to tell friendlies from unfriendlies, a critically important consideration at increasingly busy airports and mass transit hubs. Another is city planning, in which architects today spend a lot of time building 3D models from 2D images of buildings. But the challenge found in medical imaging may be the most intriguing of all, as the medical community has come only recently to recognize the true potential of 3D.
Freed by data storage technologies developed to handle other concerns, medical imaging might develop databases of 3D whole-body scans searchable for specific pathologies, reconstructions that serve as volumetric snapshots of the health or disease of patients. These 3D models could contain the quantitative indicators of disease, such as progressive measurements of tumor volume or the composition of plaques, as well as raw data that could be reanalyzed when new tools or medical knowledge allowed their interpretation. Patterns not apparent one day might be the next. It's really not that far a stretch.
Radiology was founded on this kind of logic, when film images were established as repositories from which information could be gleaned during a second read or an expert consultation. Computer-aided detection grew naturally from the digital revolution, breeding algorithms that made sense of subtle 2D patterns.
A search engine capable of reading 3D patterns may be the next logical step in a world where 3D is king.