Claron Technology Inc.

CT studies of the spine are common procedures, especially in facilities where MRI is not easily available or affordable. They are performed to evaluate injury, assess pre/post spine surgical anatomy, diagnose vertebral spine tumors, disc herniations, nerve root pinching, osteoporosis and others. Interpretation of CT spine studies is made much easier by viewing the spinal canal, disc spaces and vertebrae using images reformatted to accurately fit the local spine anatomy, rather than the general body axes. Generating such images currently requires a substantial amount of manual interaction to position and label the reformatted images. Claron's AutoSpine engine aims to eliminate most or all of the manual interaction time currently required to produce optimized and labeled reformatted spine images.

Functionality

The engine is a software module designed to be "plugged in" a host PACS or advanced-visualization software system. When a new study arrives, the host activates the engine which, in a completely automated fashion, identifies the spine and all its key anatomical landmarks. It can then either export the landmarks' geometrical descriptors to the host system or render and save new images reformatted and labeled according to the user's preferred clinical protocol provided by the host. Typical automated processing time on a recent model PC is 30 seconds (not including new series rendering and saving).

For an annotated live demonstration of the results, please visit YouTube and search for "Auto Spine" or click here.

The AutoSpine engine receives as input a DICOM CT image series containing some part, or all, of the vertebral column, and generates either geometrical descriptors or renders and saves new DICOM images representing:
Sagittal and coronal curved reformats along the centerline of the spinal cord

Automation Robustness

The engine is highly robust due to its use of the anatomical analysis modules in Claron's Withinsight Framework (WIF). These mask out all non-body regions (such as the scanning table), then register the study to a full-body atlas. Following the registration, the region containing the spine is reliably isolated for further analysis.

In clinical testing involving detailed expert assessment of AutoSpine results when processing a typical mix of spine studies, the engine's success rate was assessed at 97%. In other words, in only 3 out of 100 cases the automated results were deemed unacceptable, requiring substantial manual interaction to correct.