TY - GEN
T1 - Differentiating between giant cell arteritis and atherosclerosis on [18F]FDG-PET
T2 - 36th IEEE International Symposium on Computer-Based Medical Systems, CBMS 2023
AU - Vries, H. S.
AU - Van Praagh, G. D.
AU - Nienhuis, P. H.
AU - Bouhali, O.
AU - Slart, R. H.J.A.
AU - Alic, L.
N1 - Publisher Copyright:
© 2023 IEEE.
PY - 2023
Y1 - 2023
N2 - Background This work aims to investigate the feasibility of an explainable machine learning model based on radiomics features to differentiate between giant cell arteritis (GCA) and atherosclerosis in aortic [18F]FDG-PET scans. Method Twenty [18F]FDG-PET scans (ten of patients with GCA, ten with atherosclerosis) were retrospectively included. The aorta was delineated into four segments (ascending, arch, descending, and abdominal aorta). In total, 93 radiomic features and two quantitative features were extracted from each of the 80 segments. Four different feature selection methods and four classifiers were used to identify important features for the machine learning model and determine the probability. The model's performance was evaluated using accuracy and AUC. To enhance explainability of the model, feature importance was determined, and an occlusion sensitivity map of the aorta was created. Results The combination of the first-order skewness, GLDM dependence non-uniformity, and GLRLM run entropy features showed the highest accuracy and AUC of, 0.90±0.08 and 0.960±0.029, respectively. Conclusion This study demonstrated the potential of an explainable radiomics-based machine learning model for the differentiation between GCA and atherosclerosis in P8F]FDG-PET scans.
AB - Background This work aims to investigate the feasibility of an explainable machine learning model based on radiomics features to differentiate between giant cell arteritis (GCA) and atherosclerosis in aortic [18F]FDG-PET scans. Method Twenty [18F]FDG-PET scans (ten of patients with GCA, ten with atherosclerosis) were retrospectively included. The aorta was delineated into four segments (ascending, arch, descending, and abdominal aorta). In total, 93 radiomic features and two quantitative features were extracted from each of the 80 segments. Four different feature selection methods and four classifiers were used to identify important features for the machine learning model and determine the probability. The model's performance was evaluated using accuracy and AUC. To enhance explainability of the model, feature importance was determined, and an occlusion sensitivity map of the aorta was created. Results The combination of the first-order skewness, GLDM dependence non-uniformity, and GLRLM run entropy features showed the highest accuracy and AUC of, 0.90±0.08 and 0.960±0.029, respectively. Conclusion This study demonstrated the potential of an explainable radiomics-based machine learning model for the differentiation between GCA and atherosclerosis in P8F]FDG-PET scans.
KW - [18F]FDG-PET
KW - atherosclerosis
KW - explainable machine learning
KW - giant cell arteritis
KW - radiomics
UR - https://www.scopus.com/pages/publications/85166476630
U2 - 10.1109/CBMS58004.2023.00334
DO - 10.1109/CBMS58004.2023.00334
M3 - Conference contribution
AN - SCOPUS:85166476630
T3 - Proceedings - IEEE Symposium on Computer-Based Medical Systems
SP - 870
EP - 875
BT - Proceedings - 2023 IEEE 36th International Symposium on Computer-Based Medical Systems, CBMS 2023
A2 - Sicilia, Rosa
A2 - Kane, Bridget
A2 - Almeida, Joao Rafael
A2 - Spiliopoulou, Myra
A2 - Andrades, Jose Alberto Benitez
A2 - Placidi, Giuseppe
A2 - Gonzalez, Alejandro Rodriguez
PB - Institute of Electrical and Electronics Engineers Inc.
Y2 - 22 June 2023 through 24 June 2023
ER -