Machine algorithm uses 85 variables to calculate risk in individuals Formulas used by humans include a handful of variables and have modest accuracy
By repeatedly analysing 85 variables in 950 patients with known six-year outcomes, an algorithm “learned” how imaging data interacts. It then identified patterns correlating the variables to death and heart attack with more than 90% accuracy.
Machine learning, the modern bedrock of artificial intelligence (AI), is used every day. Google’s search engine, face recognition on smartphones, self-driving cars, Netflix and Spotify recommendation systems all use machine learning algorithms to adapt to the individual user.
Study author Dr Luis Eduardo Juarez-Orozco, of the Turku PET Centre, Finland, said: “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.”
Doctors use risk scores to make treatment decisions. But these scores are based on just a handful of variables and often have modest accuracy in individual patients. Through repetition and adjustment, machine learning can exploit large amounts of data and identify complex patterns that may not be evident to humans.
Dr Juarez-Orozco explained: “Humans have a very hard time thinking further than three dimensions (a cube) or four dimensions (a cube through time). 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.”
The study enrolled 950 patients with chest pain who underwent the centre’s usual protocol to look for coronary artery disease. A coronary computed tomography angiography (CCTA) scan yielded 58 pieces of data on presence of coronary plaque, vessel narrowing, and calcification. Those with scans suggestive of disease underwent a positron emission tomography (PET) scan which produced 17 variables on blood flow. Ten clinical variables were obtained from medical records including sex, age, smoking and diabetes.
During an average six-year follow-up there were 24 heart attacks and 49 deaths from any cause. The 85 variables were entered into a machine learning algorithm called LogitBoost, which analysed them over and over again until it found the best structure to predict who had a heart attack or died.
Dr Juarez-Orozco said: “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.”
The predictive performance using the ten clinical variables alone (similar to current clinical practice) was modest, with an area under the curve (AUC) of 0.65 (where 1.0 is a perfect test and 0.5 is a random result). When PET data were added, AUC increased to 0.69. The predictive performance increased significantly (p=0.005) when CCTA data were added to clinical and PET data, giving an AUC 0.82 and more than 90% accuracy.
Dr Juarez-Orozco said: “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.”
The Latest on: Machine learning predictions
via Google News
The Latest on: Machine learning predictions
- MSP Cybersecurity Connection: Using Prediction To Get To Predictableon October 17, 2019 at 2:50 pm
Webroot has spent a decade developing its sixth-generation machine learning and artificial intelligence solutions to offer partners a reputable and proven cybersecurity option. By leveraging The ...
- Clinical-learning versus machine-learning for transdiagnostic prediction of psychosis onset in individuals at-riskon October 17, 2019 at 10:42 am
To date, there is no empirical research comparing the prognostic accuracy of these two methods for the prediction of psychosis onset. In a first experiment, no improved performance was observed when ...
- Sports and Machine Learning? Is there a connection?on October 15, 2019 at 5:08 pm
Machine learning helps to get to the context of the sports or the game Machine Learning is making it possible for the teams and the staff to make predictions Machine Leaning is revolutionizing sports ...
- The future of artificial intelligence and machine learning in hiring: Five predictionson October 15, 2019 at 3:49 pm
Valerie Freeman is a DBJ Lifetime Achievement Award winner, Women’s Business Enterprise HOF inductee, and Houston Distinguished Alum. Artificial intelligence, together with machine learning and big ...
- Merck turning to machine learning to prevent drug shortageson October 15, 2019 at 2:24 pm
By using analytics and machine learning, the software can improve predictions and help drugmakers better match drug demand. The software could also save drugmakers hundreds of millions of dollars ...
- Drugmaker to Test Machine Learning to Prevent Drug Shortageson October 14, 2019 at 5:03 pm
To sharpen its predictions, the company’s health-care division plans to start testing a ... hospitals and wholesale distributors. TraceLink is now developing machine-learning algorithms that will be ...
- Implementing Automated Machine Learning (AutoML)on October 11, 2019 at 8:53 am
Machine learning is now deployed across many kinds of applications, and more often than not, we are seeing machine learning used to make recommendations and predictions across multiple use cases.
- Google Stadia Will Use Input Prediction to Reduce Game Lagon October 9, 2019 at 8:18 am
It seems likely button predictions will work extremely well on games where there's very little input from the player, or all input occurs within a defined area and follows a rhythm. For a first-person ...
- New Algorithm Could Improve Accuracy of Machine Learning Modelson October 9, 2019 at 7:43 am
As machine learning continues to secure its footing in the healthcare landscape ... When a predictive model is set to zero on the algorithm’s scale, its predictions are the most accurate but also the ...
- Antimicrobial Resistance Prediction for Gram-Negative Bacteria via Game Theory-Based Feature Evaluationon October 9, 2019 at 2:15 am
In order to obtain similar classification results, identity thresholds as low as 53% were required when using BLASTp. In this paper we presented a machine learning method for prediction of ...
via Bing News