Automatic coronary artery calcium scoring calculation on non-ECG-gated chest CT with novel deep learning system
Principal Investigator
Name
Long Jiang Zhang
Degrees
MD, Ph.D
Institution
Jinling Hospital, Medical School of Nanjing University
Position Title
Head of the department of Diagnostic Radiology
Email
About this CDAS Project
Study
NLST
(Learn more about this study)
Project ID
NLST-839
Initial CDAS Request Approval
Sep 24, 2021
Title
Automatic coronary artery calcium scoring calculation on non-ECG-gated chest CT with novel deep learning system
Summary
Rationale:
Clinical value of coronary artery calcium (CAC) in risk assessment for adverse events in patients with coronary artery disease (CAD) has been proven, CAC scan was rarely routinely performed in china. Routine chest CT checkup exams are preferred; however, CAC evaluation is underscored. Thus, we aimed at developing a deep learning model to calculate automatically CAC scores on the routine chest CT images. The deep learning model was trained using our internal validation with more than 6000 patients with both non-ECG-gated unenhanced chest CT exams and ECG gated CAC CT performed. The train set was completely end-to-end with a stronger reference standard.
Then we will work on the validation of the non-gated model with chest CT in different clinical scenarios: (i) Calculation of the accuracy of the new non-gated model; (ii) Validation of pre-test probability of obstructive CAD with coronary CTA as reference standard in our ongoing prospective multicenter study; (iii) Prognostic potential of the new non-gated model in medical examination population and lung cancer screening population with chest CT.
We would like access to the NLST data so that we can apply the whole-process automatic deep learning model on chest CT in a real-world population to automatically calculate CAC scores for confirmation of the feasibility and prognostic potential with the end-to-end model in lung cancer screening population referred to different clinical scenarios.
Clinical value of coronary artery calcium (CAC) in risk assessment for adverse events in patients with coronary artery disease (CAD) has been proven, CAC scan was rarely routinely performed in china. Routine chest CT checkup exams are preferred; however, CAC evaluation is underscored. Thus, we aimed at developing a deep learning model to calculate automatically CAC scores on the routine chest CT images. The deep learning model was trained using our internal validation with more than 6000 patients with both non-ECG-gated unenhanced chest CT exams and ECG gated CAC CT performed. The train set was completely end-to-end with a stronger reference standard.
Then we will work on the validation of the non-gated model with chest CT in different clinical scenarios: (i) Calculation of the accuracy of the new non-gated model; (ii) Validation of pre-test probability of obstructive CAD with coronary CTA as reference standard in our ongoing prospective multicenter study; (iii) Prognostic potential of the new non-gated model in medical examination population and lung cancer screening population with chest CT.
We would like access to the NLST data so that we can apply the whole-process automatic deep learning model on chest CT in a real-world population to automatically calculate CAC scores for confirmation of the feasibility and prognostic potential with the end-to-end model in lung cancer screening population referred to different clinical scenarios.
Aims
1) Develop and validation the novel deep learning model on non-gated chest CT in lung cancer screening population to validate the prognostic value of whole-process automatic CAC calculation;
2) Broaden our model from real-world medical examination population to those found in the National Lung Screening Trial to help better application of the non-gated model with routine chest CT exams and identify risk of coronary adverse outcome.
Collaborators
1) Chun Xiang Tang, Ph.D, Jinling Hospital, Medical School of Nanjing University;
2) Chun Yu Liu, MS, Jinling Hospital, Medical School of Nanjing University.