CAC Scoring from NCCT Using DL With External Validation
This work introduces an automatic CAC scoring method that uses multi-atlas segmentation for whole heart segmentation (WHS) and a DL model as a supervised classifier for correcting false positives (FP).
- The repository provides a DL-based FP (of CAC) model.
- The model is developed by using the Stanford AIMI COCA dataset which is publicly available for research purpose
- We used the multi-atlas segmentation pipeline implemented by the Biomedical Imaging Group Rotterdam (BIGR)
- Our work was externally validated on the Rotterdam Study
Generate labeled patches with annotated images
python3 patch_prep.py -patch_size 45
Split the patch data into non-overlapping 5 folds w.r.t subjects
python3 k-fold_prep.py -normalize
Evaluate binary classification performance and save the trained models
python3 fp_classifier_train_subject_fold.py -batch_size 32 -n_epochs 100 -lr 1e-4
Compute CAC scores
python3 coca_internal_eval.py -trained_model 'fp_vgg_trained_model_3.pth'
Assess the agreement between computed scores and reference scores
python3 coca_score_agreement.py
Mo, Hyunho, Daniel Bos, Maryam Kavousi, Maarten JG Leening, and Esther E. Bron. "Coronary Artery Calcium Scoring from Non-contrast Cardiac CT Using Deep Learning with External Validation." In International Workshop on Statistical Atlases and Computational Models of the Heart, pp. 122-131. Cham: Springer Nature Switzerland, 2024.
Bibtex entry ready to be cited
@inproceedings{mo2024coronary,
title={Coronary Artery Calcium Scoring from Non-contrast Cardiac CT Using Deep Learning with External Validation},
author={Mo, Hyunho and Bos, Daniel and Kavousi, Maryam and Leening, Maarten JG and Bron, Esther E},
booktitle={International Workshop on Statistical Atlases and Computational Models of the Heart},
pages={122--131},
year={2024},
organization={Springer}
}
This work is part of the project MyDigiTwin with project number 628.011.213 of the research programme "COMMIT2DATA - Big Data \& Health" which is partly financed by the Dutch Research Council (NWO). Furthermore, this work used the Dutch national e-infrastructure with the support of the SURF Cooperative using grant no. EINF-7675.