Page 237 - Artificial Intelligence for Computational Modeling of the Heart

210 Chapter 6 Additional clinical applications

model. The performance is statistically equivalent to that of the
computational hemodynamics model, while the inference time is
reduced to milliseconds.
It is important to note that the accuracy of the proposed
method compared with invasive measurements of quantities of
interest depends on the accuracy of the computational model
used to generate the training samples. Validation of the method
could be based on the extensive validation of the WBC model on
patient-specific datasets with invasive measurements; followed by
re-training of the neural networks on a newly generated training
database.

6.4 Summary

This section presented three clinical applications of AI and
computational models. First, we showed how a virtual heart
model could be used to plan and guide cardiac resynchronization
therapy. Then, AI based alternatives to computational hemody-
namics models were presented for decision making in the cathlab.
These AI-enabled modeling solutions can be extended to other
applications and opportunities are possible: some natural exten-
sions of use for the virtual heart model are for instance the plan-
ning and guidance of ablation therapy for complex arrhythmias,
planning of valve surgery, etc. Indeed, the approaches presented
in this book could extend to the modeling of other organs and bi-
ological systems.
While many challenges still need to be addressed (e.g. noise
in the data and uncertainty quantification, validation studies,
among others), this technology already shows great potential to-
wards the future of medicine.