Automated Cross-Sectional Co-Registration

Purpose

Accurate co-registration of imaging data sets both within and between patient studies.

Tag(s)

Non-Interpretative

Panel

Reading Room

Define-AI ID

19120003

Originator

Elliot Wasser
Lead Elliot Wasser

Panel Chair

Ben Wandtke
Non-Interpretative Panel Chairs: Alexander J Towbin, Adam Prater  

Panel Reviewers

Reading Room Subpanel

License

Creative Commons 4.0 
Status Public Commenting
                      

Clinical Implementation


Value Proposition

Cross-sectional imaging exams result in hundreds of images for radiologist review. Numerous sequences acquired for MRI, and multiple contrast time-points for CT, require constant mental comparison and task switching during the process of interpretation. Such activity may result in medical error and operator exhaustion.

The ability to accurately co-register cross-sectional data sets could substantially improve the efficiency of interpretation for large imaging data sets. By collating sequences and contrast time points together into a single 'meta series', keystrokes or other inputs could be used to rapidly compare series and enhancement dynamics without shifting attention.

Further, sequence attributes could be superimposed as a color-overlay allowing a single series to provide all information required for accurate interpretation. Such a method could also be applied for purposes of temporal comparison, both for identical and different modalities.

Narrative(s)


Multisequence contrast-enhanced MR acquisition in the abdomen for liver lesion characterization. Using a single collated 'meta sequence', all relevant lesion attributes - including T1/T2 characteristics, intra-lesional fat, enhancement pattern and diffusion - are available for overlay upon keystroke, allowing for rapid and accurate diagnosis.

Workflow Description


Primary use case would likely be integration of multiple MR sequences. Learning data set could be composed of MR exams acquired on a single anatomic location, for example, MR of the abdomen. Sequences identified by acquisition parameters and imaging plane.  Anatomic localization 'anchors' or other relatively consistent and identifiable aspects of anatomy could be used as co-registration points with morphing of images to account for differences in slice thickness, matrix, patient position, breathing and other factors affecting accurate co-registration.

Considerations for Dataset Development


  • Different vendors, pulse sequences, etc

  • Different anatomical confounders like patient size, obesity, ascites, scoliosis, etc

  • Adult/pediatrics

Technical Specifications


Inputs

Cross-Sectional DICOM Datasets

Procedure

CT, MR, PET/CT

Views

N/A

Data Type

DICOM

Modality

CT, MR, PET/CT

Body Region

All

Anatomic Focus

All


Outputs 


Unified Meta-Series

Definition

Co-registered series for each available imaging plane which may be presented as a single master series with color overlays to indicate relevant signal characteristics / enhancement kinetics, or traditional series with ability to rapidly switch between co-registered acquisitions.

Data Type

DICOM Images with overlay

Value Set

N/A

Units

N/A

Future Development Ideas


  • Methods to manually correct/modify alignments