BRIEFCASE | SEARCH | EMAIL THIS EVENT | LOGIN    
RSNA 2003 Scientific Papers > Evaluation of Parameters to Help Develop an Automated ...
 
  Scientific Papers
  SESSION: Radiology Informatics (PACS: Tools I)

Evaluation of Parameters to Help Develop an Automated Algorithm to Separate a Single CT Acquisition into Grouped Procedures

  DATE: Tuesday, December 02 2003
  START TIME: 10:52 AM
  END TIME: 11:14 AM
  LOCATION: Room S404CD
  CODE: G21-718
TOOLS
 
ADD TO BRIEFCASE
  PRINT
  EMAIL

PARTICIPANTS
PRESENTER
Eliot Siegel MD
Baltimore , MD
 
CO-AUTHOR
Bruce Reiner MD
 

Keywords
Computed tomography (CT)
Computed tomography (CT), computer programs
 
Abstract:
HTML

Introduction: One of the seven original Integrating the Healthcare Enterprise (IHE) protocols addressed the challenge presented by the need to perform a single acquisition of CT images that subsequently must be logically broken into multiple overlapping grouped procedures. Unfortunately this remains a labor intensive manual process. The purpose of this study was to determine whether a computationally simple algorithm based on analysis of image content could be used to automatically identify and separate various anatomic regions on a CT study. Materials and Methods: A total of 1,156 images were evaluated from multiple CT studies which included images from the head, neck, chest, abdomen, and pelvis. A histogram analysis was performed on each image to determine the relative proportions of fat, bone, and soft tissue. Additionally two indices were computed, bone-x and bone-y which are related to the physical distribution pixels that have values in the range of bone on a particular image (e.g. peripheral versus central). Results: Multiple regression analysis of the bone, fat, and soft tissue content of an individual image resulted in high levels of confidence (90-99% depending on anatomic region) for any given axial section. ROC analysis demonstrated that the most useful indices were: bone-y and tissue ratio for the head, tissue area for the neck, air ratio for the thorax, fat ratio for the abdomen, and bone-x and fat ratio for the pelvis. Discussion: We found a high level of accuracy of a combination of multiple indices in the determination of anatomic area for a given CT section. This does not take into account slice position and slice order which could be factored in to the algorithm to markedly increase accuracy through the knowledge that the best predictor of the anatomic region of a given slice is the anatomic location of the previous and subsequent slices. Conclusion: The data suggest that a computationally simple algorithm can be utilized to break a single CT acquisition into multiple grouped procedures. This would be likely to result in substantial time savings for technologists and radiologists in a digital imaging department.

 

 

Understand the presentation of grouped procedures IHE integration profile and its limitations. Describe strategies for automating the identification of an individual CT slice with regard to its anatomic location Be able to discuss the advantages of an automated algorithm which can sort individual CT sections into anatomic regions

 

 

 


Questions about this event email: esiegel@umaryland.edu