Diagnosis of thoracic aortic aneurysms and pathological pulmonary trunk dilation using chest computed tomography and artificial intelligence: modern approaches and prospects (a review)

Full Text

INTRODUCTION

According to the World Health Organization, cardiovascular diseases (CVDs) and related complications are among the leading causes of death worldwide.¹ Each year, they account for 17.9 million deaths [1].

These diseases represent a major social problem in healthcare and remain highly relevant in modern society. CVDs are associated with high economic costs, loss of working capacity, and an increased risk of fatal outcomes for patients [2]. Among CVDs, diseases of the aorta and pulmonary artery are of particular importance. Mortality associated with these conditions continues to rise and is projected to increase by 42% by 2030. Among vascular diseases, aortic aneurysms and pulmonary hypertension play an important role [3].

Aneurysms are characterized by localized pathologic dilatation of the aortic lumen that exceeds its normal diameter by 1.5 times [4]. Absolute threshold values are also used as diagnostic criteria: a diameter of ≥ 50 mm for the ascending thoracic aorta and ≥ 40 mm for the descending aorta [5]. Aneurysms are often asymptomatic and detected incidentally during imaging. Because of the sudden onset of complications and the high mortality associated with rupture, they are often referred to as silent killers [4].

The diagnosis is often established only at advanced stages, when the aneurysm reaches critical dimensions and complications occur, such as aortic dissection or aortic rupture. Scientific evidence shows that complications caused by aortic aneurysms have severe consequences and result in fatal outcomes in 94%–100% of cases in the absence of treatment, which requires substantial healthcare resources [6, 7].

In addition, diseases leading to the development of pulmonary hypertension also have a great influence and represent a major challenge for healthcare. Pulmonary hypertension is accompanied by pathologic dilatation of the main pulmonary artery and its branches and may be caused by various conditions, such as primary pulmonary hypertension, chronic thromboembolic pulmonary hypertension, chronic lung diseases, and heart failure [8]. Without timely diagnosis and treatment, pulmonary hypertension inevitably progresses, resulting in right ventricular failure and increasing the risk of death. Fatal outcomes occur due to complications such as thromboembolism, acute heart failure, and arrhythmias [9, 10].

Timely detection of thoracic aortic aneurysms and pathologic dilatation of the pulmonary trunk is an essential step that ensures comprehensive diagnostic evaluation and appropriate treatment selection, which positively affects clinical prognosis. This approach enables initiation of necessary treatment before the disease reaches a critical stage, significantly reducing the risk of severe complications and death. Implementation of modern methods for early detection of aortic and pulmonary artery diseases, such as computed tomography (CT) with automated image analysis using artificial intelligence (AI) technologies within opportunistic screening, opens new perspectives for subsequent use of specialized diagnostic methods, including computed tomography angiography (CT angiography), and treatment.

SEARCH METHODOLOGY

For this scientific review, articles were selected from PubMed, Google Scholar, and Scopus published between 1983 and 2024 in Russian and English. The search was conducted using the following keywords: aneurysm of the thoracic aorta, pulmonary artery dilatation, computed tomography, artificial intelligence in radiology, and early diagnosis of cardiovascular diseases. Original studies, reviews, and meta-analyses related to the diagnosis of thoracic aortic aneurysms and pathologic dilatation of the pulmonary trunk using CT and AI technologies were considered. Articles published in non–peer-reviewed journals, as well as studies involving animals, were excluded. The total number of sources identified was 538, of which 121 publications were included in the final review.

AORTIC ANEURYSM

Statistics show that the global mortality rate from aortic aneurysms between 1990 and 2019 increased by 81.6%—from 94,968 to 172,427 cases. The disease predominantly affects men [11–13].

In Russia, the prevalence of ascending thoracic aortic aneurysm varies from 0.16 to 1.06% [14]. For example, an analysis of autopsy findings at City Clinical Hospital No. 15 named after O.M. Filatov for 1991–2001 showed that thoracic aortic aneurysm was identified in 0.8% of cases [15]. In 2020, more than 6000 individuals in the United States of America (USA) died from aortic aneurysm [16]. The incidence of thoracic aortic aneurysm is approximately 10 cases per 100,000 individuals per year, with a rupture rate of about 1.6 cases per 100,000 individuals [17]. Complications of this condition include aortic dissection and rupture in 3.7% and 3.6%, respectively [18]. In Hong Kong (China), thoracic aortic aneurysm is diagnosed in 7.5% of individuals with hypertension [19]. More than 1900 hospitalizations and approximately 350 deaths associated with aortic aneurysm and dissection are reported each year in this large city. A persistent upward trend in incidence has been noted [20].

Regional differences highlight the need for global analysis, as data from other countries indicate the significance of thoracic aortic aneurysm as a major medical and social problem worldwide. For example, in Canada, the prevalence increased from 3.5 to 7.6 per 100,000 individuals between 2002 and 2014 [21]. In Western European countries, thoracic aortic aneurysm ranks second in prevalence among aortic diseases after atherosclerosis [5, 22]. In South Korea, this condition is diagnosed in 36.5% of men with hypertension [23]. In Japan, thoracic aortic aneurysm was identified in 6.5% of those examined [24]. In Iran, the frequency of >45 mm aneurysms is 1.2% [25].

Variability in statistical data on aortic aneurysm may be associated with the use of different diagnostic methods, such as CT, ultrasound, and magnetic resonance imaging (MRI), which differ in accuracy and sensitivity, as well as differences in imaging protocols and reporting standards established in different countries. Additionally, aortic size is influenced by the individual’s race and ethnicity [26].

Importance of Early Detection of Thoracic Aortic Aneurysms

Individuals with thoracic aortic aneurysms frequently have other CVDs and concomitant conditions. Early detection enables optimized management and timely surgical treatment. Elective surgery for thoracic aortic aneurysms markedly increases survival and allows patients to achieve an expected life span comparable to that of healthy individuals of the same age. The 5-year survival rate after elective surgery for aortic aneurysm is approximately 85%, whereas after emergency procedures it decreases to 37% [20, 27]. The maximum normal diameter of the thoracic aorta is up to 40 mm for the ascending aorta and aortic arch, and up to 30 mm for the descending aorta. Surgical intervention for ascending aortic aneurysm is indicated for individuals who have Marfan syndrome and a ≥ 50 mm aneurysm (this threshold is reduced to ≥ 45 mm if additional risk factors are present). In other cases, the threshold aneurysm diameter is ≥ 55 mm. Similar recommendations apply to aneurysms of the aortic arch and descending aorta (≥ 55 mm) [5]. In addition, when the diameter of the thoracic aorta exceeds 6 cm, the risk of life-threatening complications becomes almost unavoidable. The aneurysm growth rate may reach 0.1–0.3 cm per year [28].

The main diagnostic methods for thoracic aortic aneurysm include the following (Table 1):

• chest X-ray [29, 30];
• transthoracic echocardiography [31–42];
• transesophageal echocardiography [34, 37, 42];
• CT [3, 16, 33, 43–53];
• MRI [3, 54–62];
• positron emission tomography [63–67];
• direct aortography [5, 59, 68–71].

 

Table 1. Diagnostic methods for thoracic aortic aneurysm

Advantages	Limitations	Visualization
Chest radiography
• simplicity; • wide availability	• low specificity; • detection of aneurysm in 30%–40% of cases; • limited information on aortic size and configuration	• abnormalities of aortic contours and dimensions; • widening of the mediastinal shadow; • tracheal displacement
Transthoracic echocardiography
• noninvasive method; • absence of ionizing radiation; • wide availability and low cost; • real-time visualization	• operator dependent; • limited visualization of aortic segments (arch and descending aorta); • sensitivity of 55%–60%; • difficulties in visualization in patients with obesity or chest wall deformities	• aortic wall and lumen; • ascending aorta; • aortic root; • partial visualization of the aortic arch
Transesophageal echocardiography
• semi-invasive method; • more accurate visualization of the aortic root and arch; • advantage in patients with limitations for transthoracic echocardiography	• operator dependent; • potential underestimation of aortic dimensions; • patient discomfort due to semi-invasiveness; • risk of esophageal injury	• aortic wall and lumen; • aortic root; • aortic arch; • ascending aorta
Computed tomography
• high spatial resolution; • high acquisition speed; • availability of 3D reconstructions; • gold standard for the diagnosis of aortic aneurysm	• exposure to ionizing radiation; • risk of renal impairment associated with contrast administration; • allergic reactions to contrast agents; • potential motion- or metal-related artifacts	• detailed assessment of aortic diameter, wall, and lumen; • size and location of the aneurysm; • evaluation of adjacent structures; • detection of complications (dissection, thrombosis)
Magnetic resonance imaging
• absence of ionizing radiation; • high soft-tissue contrast; • availability of 3D reconstructions; • contrast-free techniques available	• longer examination time; • limitations in the presence of metallic implants; • lower availability compared with computed tomography	• detailed assessment of aortic wall and lumen; • detection of intramural hematoma and thrombus; • detailed evaluation of aortic anatomy and its relationships with surrounding structures
Positron emission tomography
• detection of aortic wall inflammation and infection; • ability to combine with computed tomography for improved diagnostic accuracy	• high cost; • limited use in routine clinical practice; • relatively low spatial resolution	• metabolic activity of the aortic wall; • detection of inflammation, infection, or malignant processes
Direct aortography
• accurate visualization of the aortic lumen; • assessment of functional characteristics of valves and the left ventricle; • possibility of therapeutic endovascular interventions	• invasive method with risk of complications; • aortic wall is not visualized	• assessment of aortic lumen and diameter; • functional state of the aortic valve and left ventricle

 

Predictors of Risk for Acute Aortic Events

It has been established that aortic diameter depends on age, sex, body weight, and height, with the strongest correlation observed with age. An increase in diameter is noted with age at all levels of the thoracic aorta; therefore, the established normal values should be age-appropriate. Sex affects only the diameter of the descending aorta, with an average difference of 1.99 mm, which is not considered clinically relevant. Body mass index is also significant at most aortic levels: each unit increase is associated with a mean diameter increase of 0.27 mm. This requires adjustment based on individual anthropometric characteristics [72].

To improve the accuracy of assessing aortic dilatation and identify the risk of aneurysm development, the aortic size index (ASI) was proposed, which incorporates aortic diameter and body surface area (BSA). It demonstrated higher prognostic value for adverse events compared with maximum aortic diameter. In 2006, Davies et al. [73] proposed the use of the aortic size index to normalize aortic diameter to body surface area.

According to these data, individuals with an aortic size index < 2.75 cm/m² were classified as having a low risk of complications, whereas values > 4.25 cm/m² indicated the need for surgical intervention. However, the method has limitations because body surface area and body weight change with age and do not always reflect aortic physiology. Therefore, more reliable indicators are needed to assess the degree of aortic dilatation [74].

The ratio of aortic cross-sectional area to its height was proposed to assess the risk of acute aortic events in individuals with genetic aortic diseases, but this parameter has limitations. Studies have shown that individuals with a low index had reduced long-term survival; however, when overall mortality was used instead of acute aortic events, its prognostic value was not demonstrated. Consequently, it did not gain widespread clinical use [74].

In addition, the length of the ascending aorta should be considered when assessing the risk of acute aortic events. It has been shown that aortic length increases with age independently of body surface area, accompanied by increased tortuosity, which can be assessed using the aortic tortuosity index [74].

A combined assessment of aortic diameter and length with calculation of aortic volume demonstrates higher sensitivity in predicting acute aortic events compared with the use of a single parameter. Volumetric aortic analysis is a quantitative assessment method with high informative value during preoperative planning and postoperative follow-up. This parameter enables the detection of important changes that may be missed during conventional diameter-based assessment. Studies have shown that an increase in aortic volume is a sensitive predictor of complications after endovascular procedures on the aorta. Therefore, volumetric assessment may become an important tool for monitoring and prognosis; however, further studies are required to confirm its value [74].

Thus, relevant parameters for assessing the risk of acute aortic events include aortic diameter, length, and volume, which account for an individual’s body habitus. Combined measurement methods, such as volumetry, provide higher sensitivity and accuracy in risk prediction and aortic monitoring, contributing to timely diagnosis and treatment [74].

PATHOLOGIC DILATATION OF THE PULMONARY TRUNK

Pulmonary hypertension affects approximately 1% of the global population, with its prevalence reaching 10% among individuals aged 65 years or older. The main causes of pulmonary hypertension include heart and lung diseases [75]. In the United Kingdom, the incidence of pulmonary hypertension is 97 cases per 1 million population, with men affected 8 times more often than women. In the United States, mortality from this condition ranges from 4.5 to 12.3 cases per 100,000 individuals. The most common causes of pulmonary hypertension are diseases affecting the left heart chambers as well as pulmonary pathological processes accompanied by hypoxia [9]. In Western Australia, 326 cases of pulmonary hypertension per 100,000 population were recorded, of which 250 were associated with left heart disease and 37 with lung disease [76].

In Russia, the prevalence of pulmonary arterial hypertension ranges from 15 to 60 cases per 1 million individuals, whereas the incidence reaches 10 cases per 1 million individuals [11, 77]. According to Aliev et al. [78], dilatation of the pulmonary trunk (≥ 29 mm) was identified in 189 of 511 individuals with COVID-19. In a study involving 7164 individuals with chronic obstructive pulmonary disease, the prevalence of pulmonary hypertension among those with respiratory symptoms was 21.8%, whereas in the overall sample it was 15.3% [79, 80].

Methods for Early Detection of Pulmonary Hypertension

The diagnosis of pulmonary hypertension requires coordinated work within a multidisciplinary team, including cardiologists, radiologists, and pulmonologists [9].

At early stages, symptoms of pulmonary hypertension may be minimal and occur only during physical exertion, which complicates early diagnosis. Dyspnea and fatigue may be mistakenly attributed to other, less serious conditions. As the disease progresses, clinical manifestations become more pronounced [9].

Pulmonary hypertension is one of the main causes of pulmonary trunk dilatation. With increasing use of noninvasive imaging methods, the likelihood of incidental detection of pulmonary artery enlargement is rising. Early detection of pulmonary hypertension is crucial for initiating specific treatment and improving prognosis. However, growing awareness of pulmonary hypertension does not always ensure timely diagnosis because its symptoms closely resemble those of other, more common respiratory or cardiovascular diseases [81–84].

Among the diagnostic methods for identifying pulmonary trunk dilatation as a sign of pulmonary hypertension, the following are distinguished (Table 2):

• echocardiography [9, 35, 36, 85];
• chest X-ray [9, 85, 86];
• CT [9, 85–88];
• MRI [9, 85, 86];
• right heart catheterization [9, 86].

 

Table 2. Diagnostic methods for pulmonary artery trunk dilatation

Advantages	Limitations	Visualization
Echocardiography (transthoracic echocardiography)
• noninvasive method; • assessment of systolic pulmonary artery pressure; • visualization of cardiac structures and assessment of cardiac function; • safety and wide availability	• operator dependent	• maximum tricuspid regurgitation velocity; • cardiac chamber dimensions; • right ventricular myocardial hypertrophy; • atrial septal defects and other congenital heart defects; • anomalous pulmonary venous return; • ventricular diastolic function
Chest radiography
• availability and simplicity; • detection of cardiomegaly and pulmonary artery dilatation; • assessment of potential causes of pulmonary hypertension (e.g., interstitial lung disease)	• low specificity; • limited accuracy; • poor sensitivity at early stages of pulmonary hypertension	• cardiomegaly (enlargement of right heart chambers); • dilatation of central pulmonary arteries
Computed tomography
• high spatial resolution; • high acquisition speed; • availability of 3D reconstructions	• ionizing radiation exposure; • contrast agent administration required	• assessment of pulmonary vascular system and lung parenchyma; • thrombi and vascular wall thickening; • right ventricular hypertrophy and dilatation
Magnetic resonance imaging
• absence of ionizing radiation; • high soft-tissue contrast; • accurate assessment of right ventricular volume and mass; • phase-contrast magnetic resonance imaging for assessment of pulmonary arterial blood flow	• lower availability compared with computed tomography; • prolonged examination time; • contraindications in some patients	• structural and functional changes of the right ventricle; • right ventricular hypertrophy and dilatation; • pulmonary arterial flow velocity; • vascular abnormalities (thrombi, wall changes)
Right heart catheterization
• gold standard for the diagnosis of pulmonary hypertension; • direct measurement of pressure in the pulmonary artery and right ventricle; • assessment of pulmonary vascular resistance	• invasive method; • risk of complications	• direct measurement of pressure in the right atrium, right ventricle, and pulmonary artery

 

Features of Diagnosing Pulmonary Trunk Dilatation Using Computed Tomography

One of the key diagnostic signs of pulmonary hypertension on CT is dilatation of the pulmonary trunk. According to studies, a pulmonary trunk diameter greater than 29 mm indicates its presence with high accuracy and demonstrates a positive predictive value of 97% for detecting this condition [87, 88]. A >29 mm increase in the pulmonary artery diameter on CT is considered an important biomarker of pulmonary hypertension, associated with elevated pulmonary vascular resistance [8, 89, 90]. However, the updated 2022 guidelines of the European Society of Cardiology/European Respiratory Society (ESC/ERS) revised the criteria indicating the presence of pulmonary hypertension: a pulmonary artery diameter of 30 mm or more on CT is considered diagnostically significant. In addition, in October 2024, the Ministry of Health of the Russian Federation also updated clinical guidelines to include this diameter; however, it should be highlighted that three parameters must be assessed, which can only be adequately measured in contrast-enhanced studies:

• ≥ 30 mm pulmonary artery diameter;
• ≥ 6 mm thickness of the anterior wall of the right ventricle;
• ≥ 1 ratio of right-to-left ventricular sizes [9, 86].

There is a statistically significant association between the ratio of the pulmonary trunk diameter to the ascending aorta diameter and an increased risk of death, regardless of the presence of ischemic heart disease. It has been proposed to measure the ratio between the maximum diameters of the pulmonary trunk and the ascending thoracic aorta. The normal value of the index is ≤ 1, whereas values > 1 indicate pulmonary trunk dilatation. According to the 2022 ESC/ERS guidelines and the 2024 clinical guidelines of the Ministry of Health of the Russian Federation, the threshold value of the index has been revised from 1 to 0.9 [9, 86]. This ratio may assist in the clinical assessment and prediction of the course of pulmonary hypertension [90–94]. An example of measuring the aortic and pulmonary artery diameters on CT is shown in Fig. 1.

 

Fig. 1. Measurement of the pulmonary artery trunk–to–ascending aorta diameter ratio. On axial computed tomography images obtained at the level of the pulmonary artery bifurcation, the diameters of the pulmonary artery trunk and the ascending aorta were measured. Electronic calipers were used to record the maximum vessel diameters perpendicular to the long axis of the main pulmonary artery: a, with contrast enhancement; b, without contrast enhancement.

 

The pulmonary trunk is usually measured at its bifurcation, perpendicular to the long axis on an axial slice; the diameter of the ascending aorta is measured at the same level when calculating their ratio [90]. Imaging biomarkers such as pulmonary trunk diameter and secondary signs of heart failure, including enlargement of the inferior vena cava, pleural effusion, and ground-glass opacity, are also instrumental for identifying pulmonary hypertension. A pulmonary trunk–to–ascending aorta diameter ratio greater than 1 has a specificity of 92% for detecting elevated mean blood pressure above 20 mm Hg [88].

In addition to conventional diagnostic methods, the use of four-dimensional phase-contrast MRI provides clinicians with the possibility of noninvasive assessment of blood flow characteristics in the aorta and pulmonary artery. This method is becoming increasingly available due to improved technologies and the development of data-processing software. It allows noninvasive visualization of blood flow features and measurement of hemodynamic parameters in the aorta and pulmonary artery. Four-dimensional phase-contrast MRI helps better understand the mechanisms of aneurysm development, for example in individuals with a bicuspid aortic valve. However, widespread implementation of this technology in clinical practice requires standardization of the methodology and improved accessibility [3, 95].

OPPORTUNISTIC SCREENING

Opportunistic screening refers to the analysis of imaging studies performed for other clinical indications. This approach enables identification not only of the target disorder but also of additional risk factors and conditions that may potentially lead to serious complications, without the need for repeated examinations of the same anatomical region. This method reduces radiation exposure and optimizes the diagnostic process [96, 97].

An example of this approach is the use of chest and abdominal CT data for opportunistic screening of aortic aneurysms. Achieving high effectiveness requires screening a large number of patients because it is difficult to narrow down a specific risk group for this condition [41].

AI technologies can be used during opportunistic screening to automatically analyze disease-specific imaging biomarkers on CT when assessing the aorta and the pulmonary artery trunk [98].

These biomarkers include:

• dilatation of the ascending aorta ranging from 40 to 49 mm, which is associated with aneurysm development;
• pulmonary trunk dilatation of 29 mm or greater, which may indicate pulmonary hypertension [97, 99].

The use of AI technologies for detecting such biomarkers may help reduce diagnostic errors [88].

In addition, after verification, imaging biomarkers can be used to assess treatment effectiveness and serve as endpoints in clinical trials. Their use increases diagnostic efficiency and supports the implementation of a personalized approach to CVD management, including pulmonary hypertension [88].

Artificial Intelligence and Its Role in Opportunistic Screening Based on Computed Tomography

The term artificial intelligence was coined in 1956, and since then AI technologies have become widespread and applied in many areas of life, including clinical practice. In medicine, driven by increased computational capacity and the growing number of digital radiologic images, AI helps reduce physicians’ routine workload by improving abnormality visualization and accelerating the diagnostic process [100]. For example, an AI system was developed in Russia during the COVID-19 pandemic to identify signs of pneumonia on CT images [101].

The integration of AI into medical imaging has substantially changed the diagnosis of CVDs. Deep learning methods used in AI systems have substantially improved the accuracy and efficiency of image interpretation [1]. The use of AI technologies in combination with radiologic imaging enables a more comprehensive analysis of structural and functional features of the cardiovascular system. These technologies support segmentation, disease classification, risk prediction, and clinical decision-making, underscoring their importance in addressing CVDs [102, 103].

CT data serve as a basis for opportunistic screening, which makes it possible to detect additional disorders in other organs without the need for repeated examinations. This approach has proved effective in identifying osteoporosis based on CT findings among the Moscow population. During the COVID-19 pandemic, more than 90,000 chest CT examinations were performed, allowing identification of signs of osteoporosis in more than 29,000 patients using an AI system based on neural networks [95, 104].

AI systems have also been used for the early diagnosis of aortic and pulmonary trunk dilatation. For instance, Mets et al. [105] discuss the feasibility of using noncontrast CT in screening programs for the early detection of aortic dilatation. This approach would enable timely application of preventive surgical methods aimed at reducing aneurysm rupture risk and saving patients’ lives. Incorporating thoracic aortic CT into comprehensive screening examinations that allow assessment of the lungs, coronary calcium, and vertebral bone mineral density would substantially expand diagnostic capabilities and enhance the effectiveness of preventive measures.

The active development of AI was kick-started during the COVID-19 pandemic, as evidenced by the emergence of early studies on its application. For example, AI-based analysis of CT data revealed the effect of COVID-19 on the biomechanical properties of the ascending aorta. A study conducted in Wuhan demonstrated that more than 50% of 38 adult patients with COVID-19 showed an increase in the diameter of the ascending aorta, which was accompanied by pronounced inflammation and myocardial injury. In children with COVID-19 and multisystem inflammatory syndrome, reduced indices of wall stress and distensibility of the ascending aorta were reported [106–108]. These findings underscore the importance of timely identification of high-risk patients and the need to consider these data in the post–COVID-19 period.

According to a study by Eltorai et al. [109], opportunistic screening using CT is feasible in the primary healthcare setting because the obtained data facilitate detection of disorders and may contribute to reduced morbidity and mortality. However, the authors expressed concerns regarding the use of AI technologies for automated detection of abnormalities on CT scans. They emphasized that findings identified by AI must be verified by a radiologist to minimize potential errors and build trust in this technology. A total of 71 physicians participated in the survey: 74.6% were aware of the concept of AI, whereas only 8.5% used it in clinical practice, and only 4.2% were familiar with CT-based opportunistic screening. For successful implementation of AI technologies and CT-based opportunistic screening, physicians emphasized the need for additional training and informational support [109].

Given the large volume of accumulated CT data, opportunistic screening using AI technologies may contribute to the early detection of aortic dilatation and aneurysms in the general population [97].²

One of the key areas of AI application is automation of CT image processing, which offers several substantial advantages.

First, AI enables retrospective analysis of large data sets, significantly accelerating the diagnostic process. Second, it improves reproducibility and measurement accuracy, thereby reducing uncertainty for medical experts. Thus, AI is becoming a valuable tool for radiologists, enabling more accurate and efficient diagnosis [110].

These findings underscore the importance of further research into the integration of AI technologies into clinical practice for disease diagnosis, particularly in settings requiring high accuracy and rapid analysis. For instance, Mori et al. [111] analyzed 5662 chest CT examinations and incidentally detected aortic dilatation in 2.1% of cases, including 3.2% in men and 0.9% in women, predominantly among individuals older than 50 years. These data confirm the need for screening for thoracic aortic dilatation and aneurysm in men older than 50 years. However, women should also be considered, because a high mortality rate has been reported among female patients undergoing surgery for thoracic aortic aneurysm. In addition, more rapid aneurysm progression and a higher risk of aortic dissection have been observed in women compared with men [112], as also demonstrated by Cheung et al. [113]. These data support the feasibility of opportunistic screening.

Accuracy of aortic diameter measurement is an important aspect of diagnosis; however, it may vary substantially among radiologists. It creates an additional risk of erroneous conclusions. Cayne et al. [114] identified a mean difference of 4 mm when measuring the maximum aortic diameter on CT images. For this reason, accurate diagnosis requires comparison of identical images and direct manual measurements performed by the examiner [114, 115]. Sedghi Gamechi et al. [116] noted that manual aortic measurement is labor intensive and observer dependent; therefore, automated segmentation and diameter analysis are preferable for both screening and clinical practice. Automatic aortic segmentation on noncontrast CT is challenging because of the lack of contrast between blood pool and surrounding tissues, unlike CT angiography, for which automated solutions already exist. Nevertheless, as a part of the Moscow experiment, AI-based tools enable vessel annotation using data obtained from noncontrast CT examinations [102].

In addition, Monti et al. [117] evaluated the performance of a commercial AI-based software platform (AI-Rad Companion^®; Siemens Healthineers, Germany) in the analysis of 250 chest CT examinations (noncontrast enhanced) and CT angiography studies across various pathologic conditions. In addition, Monti et al. [117] evaluated the performance of a commercial AI-based software platform (AI-Rad Companion^®; Siemens Healthineers, Germany) in the analysis of 250 chest CT examinations (noncontrast enhanced) and CT angiography studies across various pathologic conditions.

The results demonstrated that the system measured thoracic aortic diameter with high accuracy, yielding values comparable to those obtained by expert readers. Bland–Altman analysis revealed a difference of 1.5 mm between measurements performed by the system and by expert readers. Despite this relatively small difference, these findings indicate the need for human involvement in data verification and interpretation to ensure maximum diagnostic accuracy and reliability.

Supporting these findings, Hamelink et al. [118] conducted a study including 240 patients and demonstrated that automated measurements of thoracic aortic diameter using AI technologies on low-dose noncontrast CT scans were comparable with manual assessments. Bland–Altman analysis revealed no bias, and the mean difference between measurements was approximately 2 mm.

The obtained results are also supported by the study by Pradella et al. [16]. The authors reported that use of the DL-prototype AI-Rad Companion^® software (Siemens Healthineers, Germany) incorporating AI technologies on a dataset of contrast-enhanced and noncontrast-enhanced chest CT images demonstrated high accuracy in assessing thoracic aortic diameter. According to radiologist reports, the AI-based correction system correctly identified the presence or absence of thoracic aortic dilatation in 17,691 cases (97%), including 452 previously undetected cases of dilatation; moreover, the results were independent of contrast enhancement. Thus, AI enabled detection of previously missed dilatation in 2.6% of cases (452 of 17,691). These findings confirm its value as a supportive tool for improving the quality and efficiency of radiologic reporting.

Kim et al. [119] investigated the use of a deep learning model based on a 3D U-Net architecture for automatic segmentation of pulmonary arteries on CT images. This approach enabled 3D reconstruction and quantitative analysis of structural changes in pulmonary arteries in chronic thromboembolic pulmonary hypertension and pulmonary arterial hypertension.

In addition, studies focusing on segmentation of the heart and vessels have demonstrated the feasibility of comprehensive cardiovascular assessment. Dwivedi et al. [88] applied AI to analyze the heart and major vessels on CT images for pulmonary hypertension phenotyping. The authors demonstrated that U-Net–based models achieved Dice similarity coefficients approaching 0.9, which were comparable to or exceeded the quality of manual segmentation performed by experienced radiologists. These methods not only simplify the annotation process but also provide the opportunity to automatically detect and assess a wide range of pulmonary hypertension–related disorders, including through the use of texture analysis and classification methods aimed at accurate prediction of the disease course.

MOSCOW EXPERIMENT

Since 2020, Moscow has been conducting the largest study worldwide on the use of AI technologies for medical image analysis as a part of a project aimed at introducing innovative computer vision technologies into the city healthcare system [102, 103].² In 2022, more than 647,000 chest CT examinations were performed, and in 2023, more than 470,000. These data enable opportunistic screening for pathologic dilatation of the thoracic aorta and pulmonary trunk [97].

Nevertheless, systematic (organized) screening for thoracic aortic aneurysm and dilatation is lacking [111]. An example of AI system performance within the Moscow Experiment is shown in Fig. 2.

 

Fig. 2. Example of performance of a Russian artificial intelligence–based system: a, diameters of the ascending and descending thoracic aorta as well as the pulmonary artery trunk are indicated by green lines (within normal limits). A pulmonary lymph node is highlighted by a red box, with its size and volume indicated; b, the diameter of the ascending aorta is marked by a yellow line (dilatation), and the diameter of the pulmonary artery trunk by an orange line (pathologic dilatation); absence of measurements of the descending aorta indicates incorrect performance of the artificial intelligence system. Suspected lung parenchymal consolidation (pneumonia) is outlined in orange, and pleural effusion is outlined in yellow.

 

CONCLUSION

Early diagnosis of thoracic aortic aneurysm and pathologic dilatation of the pulmonary trunk is crucial for preventing serious complications. In the absence of timely treatment, these conditions may lead to aortic dissection or rupture, which often result in death. Pulmonary hypertension also poses a serious threat: without adequate therapy, it may progress to right ventricular heart failure, accompanied by an increased risk of death due to thromboembolism, arrhythmia, or pulmonary edema. Thoracic aortic dilatation and pathologic dilatation of the pulmonary trunk represent key imaging biomarkers indicating the presence of these conditions. Early detection of such predictors enables initiation of treatment at stages when it is most effective, thereby reducing morbidity and mortality.

AI is one of the drivers of higher effectiveness of diagnostic and monitoring strategies. Its use contributes to increased accuracy and reproducibility of measurements, including during longitudinal patient follow-up, reduces the likelihood of human-related errors, and accelerates medical data processing. AI enables more precise and objective monitoring of disease progression, which is essential for selecting optimal treatment strategies.

Integration of AI technologies into the diagnostic process in combination with opportunistic screening substantially increases the accuracy and efficiency of CT image analysis, reduces physician workload, and decreases the probability of errors. The development of these technologies and their implementation in clinical practice have significant potential to improve diagnostic quality and reduce CVD-related mortality. Nevertheless, some experts emphasize the need for mandatory oversight of AI-generated results by physicians.

ADDITIONAL INFORMATION

Author contributions: A.V. Solovev, A.V. Vladzymyrskyy: conceptualization, writing—original draft, writing—review & editing; V.E. Sinitsyn: conceptualization, writing—review & editing (final revisions); A.P. Pamova: writing—review & editing, consultation support. All the authors approved the version of the manuscript to be published and agreed to be accountable for all aspects of the work, ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Ethics approval: Not applicable.

Funding sources: This article was part of the research project Opportunistic Screening for Socially Significant and Other Common Diseases (Unified State Information Accounting System No. 123031400009-1), in accordance with Order No. 1196 dated December 21, 2022, On Approval of State Assignments Funded by the Budget of the City of Moscow for State Budgetary (Autonomous) Institutions Under the Jurisdiction of the Moscow City Health Department for 2023 and the Planned Period of 2024–2025, issued by the Moscow City Health Department.

Disclosure of interests: The authors have no relationships, activities, or interests for the last three years related to for-profit or not-for-profit third parties whose interests may be affected by the content of the article.

Statement of originality: No previously published material (text, images, or data) was used in this work.

Data availability statement: The editorial policy regarding data sharing does not apply to this work.

Generative AI: No generative artificial intelligence technologies were used to prepare this article.

Provenance and peer review: This paper was submitted unsolicited and reviewed following the standard procedure. The peer review process involved three external reviewers and the in-house science editor.

 

¹ The top 10 causes of death; [approximately 12 pages]. In: World Health Organization [Internet]. Geneva: World Health Organization; 2024–2024.

Available at: https://www.who.int/ru/news-room/fact-sheets/detail/the-top-10-causes-of-death Accessed on May 12, 2024.

² Certificate of state registration of database No. 2023621254, dated April 18, 2023. Bull. No. 4. Vladzimirskyy A. V., Andreichenko A. E., Solovev A. V., et al. MosMedData: Computed Tomography Scans With and Without Signs of Pulmonary Trunk Dilation. Available at: https://www.elibrary.ru/item.asp?edn=dbiddw Accessed on May 12, 2024.

About the authors

Alexander V. Solovev

Research and Practical Clinical Center for Diagnostics and Telemedicine Technologies; Morozov Children's Municipal Clinical Hospital

Author for correspondence.
Email: atlantis.92@mail.ru
ORCID iD: 0000-0003-4485-2638
SPIN-code: 9654-4005

MD

Russian Federation, Moscow; Moscow

Valentin E. Sinitsyn

Research and Practical Clinical Center for Diagnostics and Telemedicine Technologies; Lomonosov Moscow State University

Email: vsini@mail.ru
ORCID iD: 0000-0002-5649-2193
SPIN-code: 8449-6590

MD, Dr. Sci. (Medicine), Professor

Russian Federation, Moscow; Moscow

Anton V. Vladzymyrskyy

Research and Practical Clinical Center for Diagnostics and Telemedicine Technologies

Email: VladzimirskijAV@zdrav.mos.ru
ORCID iD: 0000-0002-2990-7736
SPIN-code: 3602-7120

MD, Dr. Sci. (Medicine)

Russian Federation, Moscow

Anastasia P. Pamova

Research and Practical Clinical Center for Diagnostics and Telemedicine Technologies

Email: PamovaAP@zdrav.mos.ru
ORCID iD: 0000-0002-0041-3281
SPIN-code: 5146-4355

MD, Cand. Sci. (Medicine)

Russian Federation, Moscow

References

Supplementary files