Machine learning-based pulse wave analysis for early detection of abdominal aortic aneurysms using in silico pulse waves

This repository contains all codes and data relevant to the study 'Machine learning-based pulse wave analysis for early detection of abdominal aortic aneurysms using in silico pulse waves' by Wang et al. [1]. Please find attached for details.

'generating input files' and 'extracting output files' contain the Matlab scripts developed by Charlton [2]. These scripts can be used to generate the input files of the original database of in silico pulse waves (PWs) [3] and extract the output files obtained by using Nektar1D [4], respectively. The only modification of the original codes is changing line 892 of 'ConvertHistoryFiles.m' into "curr_beat_onset = curr_min;". 'baseline subjects and parameter sensitivity analysis', 'database of in silico pulse waves', and 'machine learning-based pulse wave analysis' contain the Matlab scripts, input files, other relevant codes, and simulated results of the new database of in silico PWs created by this study.

There are scripts to (i) run the parameter sensitivity analysis described in [1] ('baseline subjects and parameter sensitivity analysis.zip'), (ii) generate the input files of the three subsets in which the database is divided ('generating input files.zip' and 'database of in silico pulse waves.zip'), (iii) extract the output files ('extracting output files.zip'), and (iv) run the machine learning-based pulse wave analysis ('machine learning-based pulse wave analysis.zip'). The machine learning architecture developed by this study refers to the previous code developed by Jin [5]. Each folder contains a 'README.txt' file with a brief guideline about how to use the scripts to achieve the goal.

The database of in silico PWs created by this study contains three subsets: (i) baseline subset ('database of in silico pulse waves-simulated results-sim_AAA0.zip'), (ii) increased global stiffness subset ('database of in silico pulse waves-simulated results-sim_AAA0_PWVinc.zip'), and (iii) AAA subset. The first two subsets consist of subjects without AAAs, while the third subset contains subjects with different AAA sizes: AAA with maximum diameters of 30 mm ('database of in silico pulse waves-simulated results-sim_AAA30.zip'), 40 mm ('database of in silico pulse waves-simulated results-sim_AAA40.zip') and 50 mm ('database of in silico pulse waves-simulated results-sim_AAA50.zip'). Further details are provided in Wang et al. [1].

[1] https://doi.org/10.20944/preprints202103.0745.v1
[2] https://github.com/peterhcharlton/pwdb/tree/master/pwdb_v0.1
[3] http://haemod.uk/ageing
[4] http://haemod.uk/nektar
[5] https://github.com/WeiweiJin/Estimate-Cardiovascular-Risk-from-Pulse-Wave-Signal