A study of aortic dissection screening method based on multiple machine learning models

Aortic dissection (AD) is a very rare clinical emergency, the pathogenesis of which is that the blood of the aorta enters the aortic wall under the pressure of the aorta, then a dissecting hematoma is formed in the wall of the aorta, and the longitudinal axis of the aorta is extended to form a “double lumenal aorta” (1). This is a very dangerous cardiovascular disease of which the death rate is 1–2% per hour in the first 24 hours of the onset of the disease and up to 60–70% in one week (2). Most patients who are not treated will die within a year (3). Although AD is an acute disease in urgent need of surgical treatment (4,5), the rate of misdiagnosis is relatively high (6). The clinical misdiagnosis rate of AD described in (7-9) is between 35% and 45%. At present, the golden criterion of AD diagnosis is CTA (computer tomography angiography), and its sensitivity is over 90% and specificity is close to 100% (10). Meanwhile, due to the rarity of AD, many doctors lack clinical experience for this disease, they usually don’t suspect that the patients have this disease. In fact, most patients with AD are found because they have obvious symptoms such as severe chest and/or back pain and undergo CTA on the advice of a doctor. Few people are directly suspected of having this disease. Patients with no obvious symptoms are difficult to detect because doctors will not think about sending these people to do CTA. Once a doctor cannot tell from the symptoms that the patient has an AD, the patient will not be able to receive proper follow-up diagnosis and treatment. In summary, the current screening of patients is mainly through the subjective identification of symptoms judged by clinicians, and in most cases, it is not directly suspected of this disease. So from the current clinical perspective, a basic, cheap, and fast early screening method for AD is needed urgently. If we can detect the majority of patients at high risk by screening and then recommend that they have a CTA diagnosis, we can greatly increase the detection rate for the disease