Supplementary materials for: Constitutive neural networks for main pulmonary arteries: Discovering the undiscovered

This dataset in KU Leuven's research data repository is the supplementary materials for the research paper entitled "Constitutive neural networks for main pulmonary arteries: Discovering the undiscovered".

Authors: Thibault Vervenne (KU Leuven), Nele Famaey (KU Leuven), Ellen Kuhl (Stanford University) and Mathias Peirlinck (TU Delft).

Citation: Vervenne, T., Peirlinck, M., Famaey, N. et al. Constitutive neural networks for main pulmonary arteries: discovering the undiscovered. Biomech Model Mechanobiol (2025). https://doi.org/10.1007/s10237-025-01930-1

The dataset contains the source data, code, results, and an explanatory readme.txt file.

readme.txt

data\
stretch (F) and nominal stress (P) for n = 8 samples, in axial (11) and circumferential (22) directions.

code\
CANN4Artery_main_loop.py: main code sample-specific runs of neural network, all fiber angles, fixed and fitted fiber direction
CANN4Artery_main_all-in-one.py: main code for cross-sample feature selection, all fiber angles, fixed and fitted fiber direction
ModelsArtery_p.py: build the invariant strain energy densitity function + stress calculations, as well as the reduced and Holzapfel-Gasser-Ogden models, no regularization and L1 regularization
PlotArtery_p.py: plot the results for sample-specific runs of neural network
PlotArtery_all.py: plot the results for cross-sample feature selection

Python: 3.9.18
tensorflow: 2.10.0
pandas: 2.2.2
numpy: 1.26.4
matplotlib: 3.8.4
sklearn: 1.5.1

results\
results-sample-specific.xlsx: spreadsheet containing results for sample-specific runs of neural network, all samples, all fiber angles, fixed and fitted fiber direction, no regularization and L1 regularization
results-cross-sample.xlsx: spreadsheet containing results neural network for cross-sample feature selection, all fiber angles, fixed and fitted fiber direction, no regularization and L1 regularization
results-reduced.xlsx: spreadsheet containing neural network results for reduced constitutive equation, all fiber angles, fixed and fitted fiber direction
results-hgo.xlsx: spreadsheet containing neural network results for Holzapfel-Gasser-Ogden constitutive equation, all fiber angles, fixed and fitted fiber direction, + classical Holzapfel-Gasser-Ogden material parameter fitting results