Cardiovascular diseases are the first cause of lethality in the world today. CVD accounts for 45% of all deaths in Europe, 46 times the number of deaths caused by AIDS, tuberculosis, and malaria combined. There are 2,353 deaths from CVD each day in the U.S., based on 2017 data.
The General Practitioner (GP) is the first physician to diagnose heart diseases based on chest pain symptoms. An independent study shows that the GP’s diagnosis yields 69% of sensitivity and 89% specificity. The next physician in line for diagnosis is the cardiologist. He/She can perform several tests, including electrocardiogram (ECG), exercise stress tests, X-rays, echocardiogram, blood tests, coronary angiography, radionuclide tests, MRI scans, CT scans. ECG, stress tests, and fluoroscopy are standard exams performed on patients suspected of CVD.
Some of these test results, combined with patient information and symptoms, constitute a data set from which we intend to extract an early CVD diagnosis using Quaartz. The dataset is composed of 303 patients and includes the following 13 variables:
So far, the diagnosis of CVD has been mostly dependent on the traditional approaches, using trained professionals’ expertise, particularly cardiologists.
Quaartz’ Machine Learning platform can help automate, in part, the CVD risk prediction. Thus General Practitioners and Cardiologists can get indications extracted from the patient information and test results.
It poses the following cardiology question:
Can cardiovascular disease prediction identify a person at risk?
Quaartz predictive models’ results reach an 83% accuracy based on real data for cardiovascular disease prediction. The model’s MCC (Matthews Correlation Coefficient) is 65%, and its AUC (Area Under Curve) is 86%.
Among the thirteen variables mentioned earlier and available in the dataset, Quaartz has selected and identified four main criteria to predict CVD. They bear different weights relative to their global influence on the final decision:
Quaartz obtains a 83% accuracy in the identification of CVD cases. The artificial intelligence tool distinguishes patients at high risk of CVD with excellent efficiency. It can also help the General Practitioner and the Cardiologists understand how each element measured impacts the diagnosis. It shows the dependency between the ST depression induced by exercise and CVD chances.
In one day, cardiologists gained significant support in their cardiovascular disease diagnosis and their fight against strokes by: