Calcium scoring in non-contrast CT showing the heart
- CT Image (Any CT image)
- Generic Overlay (An overlay of unknown type. Legacy, please use alternative interfaces.)
- Results JSON File (A collection of results of unknown type. Legacy, if possible please use alternative interfaces.)
This algorithm performs calcium scoring in non-contrast-enhanced computed tomography (CT) scans. Additional to a total score for coronary calcium, it provides scores per anatomical location, i.e. left anterior descending, left circumflex, right coronary artery and thoracic aorta. More information on the data and performance of the method can be found in our publication.
This is a fully automatic deep learning based calcium scoring method. The method consists of two consecutive convolutional neural networks (CNN). The first one has a large field of view due to dilated convolutions and predicts lesion candidates: voxels that are likely calcifications. This CNN also provides anatomical labels. The second CNN predicts whether candidate voxels are truly clacifications.
The final output consists of volume and Agatston scores for total coronary calcium and for each label separately.
For more information please refer to the publication "Deep Learning for Automatic Calcium Scoring in CT: Validation Using Multiple Cardiac CT and Chest CT Protocols"
Validation and Performance
This method has been evaluated using a large set of CT scans with various imaging protocols including cardiac CT, PET attenuation correction CT, diagnostic chest CT, radiotherapy treatment planning CT and calcium screening CT.
|CT type (n)||CAC||LAD||LCX||RCA||Aorta|
|Cardiac CT (529)||Sensitivity||0.93||0.93||0.86||0.91||0.86|
|PET attenuation correction CT (200)||Sensitivity||0.92||0.94||0.78||0.92||0.84|
|Diagnostic chest CT (291)||Sensitivity||0.92||0.93||0.73||0.90||0.90|
|Radiotherapy treatment planning CT (841)||Sensitivity||0.86||0.95||0.92||0.57||0.95|
|Calcium screening CT (2479)||Agatston ICC||0.94|
Uses and Directions
This algorithm was developed for research purposes only. This algorithm is intended to be used in non-contrast-enhanced CT scans that show the heart.
Standardly scans with 3 mm slice thickness and 1.5 mm increment are used for calcium scoring. Therefore, scans that have different slice thickness and increment will be resampled. If this information is not available from the dicom (or other) header, 3 mm slice thickness and 1.5 mm increment is assumed and no resampling is performed.
Common Error Messages
Information on this algorithm has been provided by the Algorithm Editors, following the Model Facts labels guidelines from Sendak, M.P., Gao, M., Brajer, N. et al. Presenting machine learning model information to clinical end users with model facts labels. npj Digit. Med. 3, 41 (2020). 10.1038/s41746-020-0253-3