Sudden chest pain in emergency settings presents a critical diagnostic challenge — determining whether it is caused by a heart attack or an aortic dissection.

Both are life-threatening cardiovascular emergencies marked by sudden onset, high mortality and overlapping symptoms such as chest pain and shortness of breath. However, their treatments differ fundamentally, meaning a misdiagnosis can have fatal consequences.

A joint research team from the People's Hospital of Xinjiang Uygur Autonomous Region and Xinjiang University has developed an artificial intelligence-powered diagnostic method that analyzes blood samples to distinguish between these easily confused conditions. Using just a single drop of blood and delivering results in about 10 minutes, the method offers a new approach for rapidly and accurately identifying the cause of acute chest pain.

The study was recently published in Engineering Applications of Artificial Intelligence, a journal that focuses on practical uses of AI across engineering fields. The journal ranks in the top quartile in areas such as AI, automation and electronic engineering.

A heart attack, or acute myocardial infarction, occurs when a blood clot blocks a coronary artery, cutting off blood supply to the heart muscle and causing tissue damage or death. It requires immediate treatment to dissolve the clot and restore blood flow.

An aortic dissection, by contrast, involves a tear in the inner wall of the aorta, the body's main artery. Blood then flows between the layers of the vessel wall, weakening it and risking rupture. This condition typically requires emergency surgery, and the use of clot-dissolving drugs is strictly avoided.

Administering such drugs to a patient with aortic dissection can trigger fatal internal bleeding, while failing to promptly treat a heart attack can cause a patient to miss the critical window for lifesaving care.

Yang Yining, director and a professor at the hospital and one of the study's corresponding authors, said the current diagnosis of aortic dissection relies on contrast-enhanced computed tomography scans. While accurate, these scans are time-consuming, costly and often impractical in ambulances or primary care settings, highlighting the need for faster and more accessible tools.

The research team turned to blood analysis instead of imaging. They combined two techniques — infrared spectroscopy and Raman spectroscopy — which examine how molecules in a sample interact with light to produce distinct "fingerprints". These fingerprints can reveal subtle biochemical differences between diseases.

Infrared spectroscopy helps identify molecular structures and functional groups, while Raman spectroscopy detects how chemical bonds vibrate. Together, they provide complementary information about the biological sample.

To make sense of this complex data, the team developed a deep learning model — a type of AI system trained to recognize patterns — that can rapidly classify diseases.

"This method turns complex data into accurate diagnoses and has shown strong results in clinical validation, with an accuracy of 94.06 percent and a specificity of 97.03 percent in distinguishing aortic dissection, myocardial infarction and noncritical cases. The cost is also much lower than that of a contrast-enhanced CT scan," Yang said.

Lyu Xiaoyi, a professor at the School of Software at Xinjiang University and another corresponding author, said the AI model — known as a multimodal attention fusion network — analyzes the two types of spectral data simultaneously, similar to examining the same sample from two different but complementary perspectives.

The system uses an "attention" mechanism, a feature in modern AI that mimics how humans focus on the most relevant details. It automatically highlights the most informative parts of the data and combines them for analysis.

Unlike traditional methods that simply merge datasets before processing, the model allows the two types of data to interact, uncovering relationships between them. This approach leads to more comprehensive and accurate results.

According to Yang, the method is expected to complement existing imaging tests. In emergencies outside of hospital settings and initial screenings, it enables rapid blood testing without the need for CT equipment, helping doctors quickly determine the appropriate treatment and avoid improper medication or unnecessary transfers.

In major hospitals, it could provide additional molecular-level evidence for patients waiting for CT scans or those unable to undergo imaging.

The equipment is also significantly less expensive than high-end CT machines and easier to operate, making it a promising option for county-level chest pain centers and primary care institutions seeking to improve diagnostic capabilities.

Yang said the method remains in the clinical research stage. The team plans to conduct multicenter studies and develop a portable device based on the technology, with the goal of accelerating its use in real-world medical settings.

Contact the writers at fangaiqing@chinadaily.com.cn