Skip to main content
. 2024 May 29;104:105164. doi: 10.1016/j.ebiom.2024.105164

Fig. 1.

Fig. 1

Illustration of the machine learning pipeline used in this work. It starts by extracting relevant patient data (cleaned and filtered) according to the clinical question. The data is then passed through 3 separate feature extraction algorithms feeding their respective models. Variable feature extraction parameters allow further model optimization. In the next stage, the baseline and CNN models are trained. These are done sub-sequentially to evaluate performance improvements. Lastly, each model is separately validated using 5-fold cross-validation before collecting the final performance metrics.