Can artificial intelligence diagnose cardiovascular disease using stool samples?

Is it unrealistic to train machines to “read” stool samples and help diagnosing cardiovascular disease? No, according to a recent study which found this original approach to be almost as effective as existing diagnostic techniques and, more importantly, much less time-consuming.

Created 15 December 2020
Updated 29 November 2021

About this article

Created 15 December 2020
Updated 29 November 2021

Cardiovascular disease (CVD) is the world’s number one cause of death. By 2030, CVD-related deaths are expected to peak at 23.6 million. Its diagnosis currently involves a series of time-consuming and costly examinations (clinical tests, ECG, chest X-rays, echocardiogram). An alteration (dysbiosis) of the gut microbiota has been linked to several types of CVD, including hypertension, heart failure and atherosclerosis. So why not use artificial intelligence to design a diagnostic test for CVD based on gut microbiota composition?

CVD “signatures” present in the stool

Machine learning is a branch of artificial intelligence that involves inputting data into a computer so that it can learn how to solve a problem. In healthcare, it has been successfully used to diagnose and predict various diseases, such as cancer, diabetes mellitus and inflammatory bowel disease. To test its usefulness for diagnosing CVD, a team of researchers compared different analytic algorithms and sought to identify characteristic “signatures” for the disease in stool samples obtained from 478 patients with CVD and 473 healthy subjects. They found significant differences between the two groups in the relative intestinal abundance of 39 bacteria.

Strong diagnostic capacity

The researchers identified a specific algorithm which, by targeting 25 bacterial families within the gut microbiota, could discriminate between the two groups with 70% accuracy. This level of accuracy is only slightly below that of the conventional approach, which is able to diagnose 76% of CVD patients, but requires an array of clinical data (age, gender, smoking status, blood pressure, cholesterol levels, etc.). According to the authors, the use of machine learning to identify intestinal dysbiosis characteristic of cardiovascular disease has very promising diagnostic potential in the context of routine check-ups.

Old sources

Sources:

Aryal S, Alimadadi A, Manandhar I, et al. Machine Learning Strategy for Gut Microbiome-Based Diagnostic Screening of Cardiovascular Disease. Hypertension. 2020 Nov;76(5):1555-1562.

en_view en_sources

    See also