You have probably read this title before, and, if you have, you have already encountered a recommendation algorithm. In the most basic terms, a recommendation algorithm will determine what item you are most likely to purchase or like, based on some information it has about you. Most likely you will find that the algorithm represents you as a vector of different properties such as the current product you are viewing, the last item you bought, your age, your gender, and probably more (alarmingly private) information. This is what my role in this project will be about- only am I not looking to recommend a product that a patient would buy, but a treatment that will terminate atrial fibrillation.
There is a lot of relevant information that can be encoded about a patient and his or her disease. Encoding a patient as a vector may therefore be composed of the time since atrial fibrillation was first diagnosed, geometrical information of the atria, electrophysiological characteristics, fibrosis patterns, and many other biological markers. Based on these biomarkers different treatment options must be considered, and the appropriate one needs to be recommended while providing an explanation alongside the recommendation. The explanation is the most challenging part here. You will find a lot of publications showing impressive results obtained by machine learning algorithms, however, the applicability of such algorithms in practice is very limited if they cannot provide an explanation. The algorithm should not prescribe a treatment, but support cardiologists in their decision-making process. It is therefore not exactly a recommendation system, but a decision support system. The general plan is to combine existing knowledge, and obtain additional knowledge from data, to produce novel guidelines for the treatment of atrial fibrillation. I will tell you all about that in the next blog post, so stay tuned!
Finally, I would like to conclude this blog post with a traditional Valencian dish: Paella Valenciana. It is cooked in a shallow pan, and best enjoyed on the beach during the lunch break at work. 🙂
Input your search keywords and press Enter.