Patients who had complained of chest pains were subjected to a battery of scans and test before being treated using traditional methods, this data was later added to train the LogitBoost algorithm on risks and during the 6 years follow up was found to have had a 90% success rate at predicting 24 heart attacks and 49 deaths from any cause.
“These advances are far beyond what has been done in medicine, where we need to be cautious about how we evaluate risk and outcomes. We have the data but we are not using it to its full potential yet.” says Dr. Luis Eduardo Juarez-Orozco of the Turku PET Centre.
Doctors use risk scores to make treatment decisions based on a handful of patient variables. According to the researchers through repetition and adjustment machines use larger amounts of data to identify complex patterns that may not be evident to humans.
“Humans have a very hard time thinking further than three dimensions or four dimensions.The moment we jump into the fifth dimension we’re lost.Our study shows that very high dimensional patterns are more useful than single dimensional patterns to predict outcomes in individuals and for that we need machine learning.”
950 patients were enrolled in this study with chest pain who were given the centre’s usual protocols to look for coronary artery disease, and CCTA scans gathered 58 pieces of data on potential risks of heart attack including presence of calcification, coronary plaque, and vessel narrowing. Those with scans suggesting disease had PET scans taken which produced 17 variables on blood flow. 10 clinical variables were obtained from medicals records such as age, sex, diabetes, and smoking habits. 85 variables were then entered into LogitBoost which repeatedly analysed them until it determined the best structure to predict who had a heart attack or who had died; results were presented at The International Conference on Nuclear Cardiology and Cardiac CT.
“The algorithm progressively learns from the data and after numerous rounds of analyses, it figures out the high dimensional patterns that should be used to efficiently identify patients who have the event – the result is a score of individual risk. Doctors already collect a lot of information about patients – for example, those with chest pain. We found that machine learning can integrate these data and accurately predict individual risk. This should allow us to personalise treatment and ultimately lead to better outcomes for patients.”
