Chest x-ray images are commonly available. Based on a single chest radiograph image, deep learning can predict the risk of future cardiovascular events independent of cardiovascular risk factors and with similar performance to the established and guideline-recommended ASCVD risk score.

---

Encouraging research found that an AI model could use a single chest X-ray to estimate a patient's 10-year probability of mortality from a heart attack or stroke.

Healthcare practitioners traditionally utilise an algorithm/risk score calculator to determine a patient's 10-year risk of cholesterol and fat buildup in the arterial walls. Typically, this method needs a cardiologist to measure a patient's blood pressure and do many tab tests.

A team of researchers currently asserts that an advanced AI model can utilise chest X-ray pictures to forecast an individual's 10-year risk of cardiovascular mortality with the same precision as the conventional risk estimator. Chest X-rays are already often used for the detection of several diseases. If an AI model can take advantage of this common imaging technology, it may assist in the identification of people at high risk for heart disease who would not have otherwise seen a cardiologist.

The study's lead author, a radiologist affiliated with the Cardiovascular Imaging Research Center at Massachusetts General Hospital and the AI in Medicine program at Brigham and Women's Hospital in Boston, stated that such patients could take a statin or blood pressure medication to reduce their risk of suffering a heart attack or stroke. In November, the study was presented at a Radiological Society of North America conference.

The prediction model is not intended to replace the standard risk calculator. If approved, however, the approach might be used to forecast health consequences for individuals who would otherwise go overlooked.

With this methodology, clinicians would be able to identify these individuals and inform them that their chance of having a stroke or heart attack within the next decade is elevated. Please see your cardiologist to determine whether you qualify for risk-reduction medication, such as a statin or blood pressure medicine.

---

Educating a machine to make heart disease predictions

Deep learning is a sophisticated artificial intelligence technique. For this work, researchers trained a deep learning model to identify cardiovascular event risk by giving it more than 147,000 chest X-rays from more than 40,000 individuals and indicating which of these patients died of heart disease after a 10-year period. The data originated from the National Cancer Institute's multi-centre, randomised, controlled Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial.

In this method, scientists provide the machine with a set of rules at the beginning and a set of outcomes at the conclusion, and the computer will figure out the steps in between. The notion is that the computer may detect indicators of cardiac problems that cardiologists may miss. Some individuals have expressed worries about artificial intelligence being a "black box" due to the fact that this is done somewhat mathematically and not in a manner that I believe is simple for humans to understand. What goes in must come out, but what happens in between is beyond your comprehension.

To verify that the model works with data it has never seen before, researchers gave the model new photos from the second set of 11,430 Mass General Brigham hospital patients who underwent regular chest X-rays.

There was a strong link between the AI model's risk projections and the actual outcomes for the approximately 10% of patients who suffered a severe adverse cardiac event in the 10.3 years that followed. One-fifth of the patients had sufficient information in their medical records to assess their 10-year risk of dying from cardiovascular disease using the ASCVD method.

Traditional approach and deep learning model fared equally in forecasting patients' 10-year chances of cardiovascular disease-related mortality. A larger, more varied sample of patients should be examined using the model in a randomised, controlled experiment.

---

Implications of AI for the future of medicine

Utilising artificial intelligence techniques in medicine involves a number of ethical problems. In the absence of varied training data, computer models may have biases. Additionally, the medical profession must evaluate how to deploy AI properly in health systems.

The capacity of computers to take in vast quantities of data, analyse it, and output something useful has exceeded that of humans. Utilising this only makes sense if we want to improve our patient care. Does this imply that these prediction models will replace doctors? It is similar to a tool in a doctor's arsenal for patient management.

X-rays are two-dimensional pictures that are simple for artificial intelligence to analyse. As technology advances, scientists may be able to examine three-dimensional cross-sectional pictures, such as those from CT and MRI scans. The use of machine learning in the early diagnosis of lung disease2 and other malignancies has shown promise.

If AI were to be used in the medical field in the near future, a single X-ray might potentially be used to determine a patient's risk of developing many diseases.

---

Some important considerations

1. Pulmonary oedema may appear as either interstitial oedema (septal lines) or alveolar oedema (airspace shadowing/consolidation), thus it's important to be on the lookout for them if the heart is enlarged since they indicate heart failure.
2. It is occasionally possible to tell whether a particular heart chamber is enlarged when the heart is enlarged;
3. The heart contour may be altered owing to cardiac or pericardial illness.
4. The heart contour may be masked by surrounding lung disease.

Signs of heart failure

1. Cardiomegaly Cardio-Thoracic Ratio (more than 50%)
2. Septal (Kerley B) lines – a marker of interstitial oedema.
3. Upper zone vascular enlargement – an indication of pulmonary venous hypertension.
4. Acute peri-hilar (bat's wing) distribution of alveolar oedema.
5. Blunt costophrenic angles– owing to pleural effusions.

Disclaimer- The views and opinions expressed in this article are those of the author and do not necessarily reflect the official policy or position of M3 India.

About the author of this article: Dr Monish Raut is a practising super specialist from New Delhi.