Fair cardiovascular disease diagnosis and prognosis through machine learning and fractal-based features

Identifier:  TFM:1884

Handle: https://hdl.handle.net/20.500.11797/TFM1884

Authors:  Colmenero Gomez Cambronero, Carlos

Abstract:
Given the high number of deaths caused by cardiovascular diseases, innovative methods are essential to mitigate their detrimental effects. Fractal analysis offers a detailed representation of complex patterns, like those in cardiovascular conditions. This thesis evaluates the fairness and predictive performance of ML models using fractal features from CMR for diagnosing and prognosing cardiovascular diseases. No significant differences were found between the best fractal and radiomics models. After mitigation, fractals were superior to radiomics. Considering their equivalent predictive performance, reduced bias after mitigation, and fewer features, fractals are proposed as an alternative to radiomics.