Cardiomegaly, a pathologic enlargement of the heart that is usually a sign of another condition, affects an estimated 18 million people aged 20 and older in the U.S.
Rothenberg
In clinical practice, the condition is generally diagnosed on CT using a subjective visual assessment, which lacks standardization. As a result, the disease often goes undiagnosed and may progresses to a future cardiovascular disease (CVD) event such as myocardial infarction, heart failure, stroke, arrhythmias or death.
However, thanks to advancements in AI, this could soon change.
According to a study conducted at the University of Alabama at Birmingham (UAB) and presented during a Sunday session, a fully automated AI algorithm has the potential to opportunistically screen for cardiomegaly on CT images of the chest or abdomen.
"Early detection of cardiomegaly would allow for appropriate workup and management to potentially reduce CVD events, improving individual patient and population health," said Steven Rothenberg, MD, assistant professor, cardiopulmonary imaging section at UAB and the study's lead author.
Algorithm Predicts Cardiovascular Disease Events
The retrospective study included a cohort of 14,299 consecutive adult patients with chest or abdominal CT exams obtained during the first half of 2016. The AI algorithm was trained on external data to select the slice where the heart was the largest, then segment the heart and inner chest in 2D and extract the linear cardiothoracic ratio (CTR), with a CTR of > 0.56 indicating severe cardiomegaly.
"We trained the algorithm on chest and abdominal CT exams without and with contrast from multiple institutions, including technically difficult cases, and then validated it on a separate dataset before deploying at UAB," Dr. Rothenberg explained.
What researchers found was that when used to quantify CTR, the fully automated AI algorithm was very good at predicting major CVD events and death. Specifically, the AI found that 60.4% of patients had normal heart size (CTR < 0.50), while 29.3% had mild cardiomegaly (CTR between 0.50 – 0.56). 10.3% of the patient cohort had severe cardiomegaly (CTR >= 0.56).
But what really surprised researchers was the fact that a majority of the patients with severe cardiomegaly were unmanaged.
"This is particularly concerning considering the algorithm showed that, compared to patients with a normal sized heart, those with severe cardiomegaly are 4.2 times more likely to have a myocardial infarction," Dr. Rothenberg said.
They are also at least twice as likely to have arrhythmia or a stroke, not to mention being over three times more likely to experience heart failure.
Reducing The Risk Associated With Severe Cardiomegaly
The heart is imaged in nearly 50 million annual chest and abdominal CT exams in the U.S., and unmanaged cardiomegaly may be present in more than 2 million of these patients. According to Dr. Rothenberg, by using AI to opportunistically screen for undiagnosed and unmanaged cardiovascular disease, radiologists can help reduce the risks associated with unmanaged severe cardiomegaly.
"Opportunistic screening for cardiomegaly on routine chest and abdominal CT scans adds no patient cost or radiation," he concluded. "Followed by appropriate care coordination and management with early behavior and lifestyle interventions, it has the potential to markedly reduce the public health burden of CVD."
Access the presentation, "Opportunistic Screening for Cardiomegaly on Chest or Abdominal CT Using Fully Automated AI," (S2-SSCA01-3) on demand at Meeting.RSNA.org..
