Skip to main content

Unsupervised machine learning based on clinical factors for the detection of coronary artery atherosclerosis in type 2 diabetes mellitus

Abstract

Background

Coronary atherosclerosis can lead to serious cardiovascular events. In type 2 diabetes (T2DM) patients, the effects of clinical factors on coronary atherosclerosis have not been fully elucidated. We used a clustering method to distinguish the population heterogeneity of T2DM and the differences in coronary atherosclerosis evaluated on coronary computed tomography angiography (CCTA) among groups and to facilitate clinical management.

Methods

Clinical data from 1157 T2DM patients with coronary atherosclerosis who underwent CCTA in our hospital from January 2018 to September 2021 were retrospectively collected. The coronary artery segment plaque type and stenosis, the number of involved vessels, the segment involvement score (SIS) and the segment stenosis score (SSS) were evaluated and calculated. Unsupervised clustering analysis based on clinical information was used (cluster 1: n = 463; cluster 2: n = 341; cluster 3: n = 353). The association of coronary plaque characteristics with cluster groups was evaluated.

Results

The clinical data among the three groups were different in several aspects: (1) Cluster 1 had the least male patients (41.7%), the lowest proportion of patients with smoking (0%) or alcohol history (0.9%), and the lowest level of serum creatinine (74.46 ± 22.18 µmol/L); (2) Cluster 2 had the shortest duration of diabetes (7.90 ± 8.20 years) and was less likely to be treated with diabetes (42.2%) or statins (17.6%) and (3) Cluster 3 was the youngest (65.89 ± 10.15 years old) and had the highest proportion of male patients (96.6%), the highest proportion of patients with smoking (91.2%) and alcohol (59.8%) history, the highest level of eGFR (83.81 ± 19.06 ml/min/1.73m2), and the lowest level of HDL-C (1.07 ± 0.28 mmol/L).

The CCTA characteristics varied with different clusters: (1) Cluster 1 had the largest number of segments with calcified plaques (2.43 ± 2.46) and the least number of segments with mixed plaques (2.24 ± 2.59) and obstructive stenosis (0.98 ± 2.00); (2) Cluster 1 had the lowest proportion of patients with mixed plaques (68%) and obstructive stenosis (32.2%); (3) Cluster 3 had more segments with noncalcified plaques than cluster 1 (0.63 ± 1.02 vs 0.40 ± 0.78, P < 0.05) and the highest proportion of patients with noncalcified plaques (39.9%) and (4) There was no significant difference in the extent of coronary plaques among the three clusters.

Conclusions

The unsupervised clustering method could address T2DM patients with heterogeneous clinical indicators and identify groups with different types of coronary plaque and degrees of coronary stenosis. This method has the potential for patient stratification, which is essential for the clinical management of T2DM patients with coronary atherosclerosis.

Background

Diabetes is a serious threat to public health. According to data from the International Diabetes Federation (IDF), the number of people with diabetes has reached 536 million worldwide, and it is estimated that this number will rise to 783 million by 2045 [1]. Type 2 diabetes mellitus (T2DM) is the most common type of diabetes. Cardiovascular disease is a common complication of T2DM and is the leading cause of morbidity and mortality among diabetes patients [2]. The presence of T2DM is associated with an increased risk of cardiovascular disease and cardiovascular events [3,4,5,6].

Coronary atherosclerosis is a major cardiovascular disease. The assessment of blood pressure, plasma lipid levels and other risk factors could be used to guide the management of atherosclerotic cardiovascular disease [2]. T2DM has traditionally been considered a risk factor for atherosclerosis and can accelerate the progression of coronary atherosclerosis [7, 8]. T2DM is often accompanied by other risk factors [1, 9]. Although the risk factors for cardiovascular disease have been elucidated separately [10,11,12], the different clinical features of T2DM patients have varying influences on coronary atherosclerosis [13,14,15]. The comprehensive effect of the clinical features of T2DM on coronary atherosclerosis still needs further study.

Machine learning methods have been widely used in cardiovascular disease research [16, 17]. An unsupervised machine learning approach has been applied to clarify the heterogeneity of coronary artery disease, indicating the feasibility of this method to identify important subpopulations based on clinical data [17]. However, few studies have focused on the exploration of unsupervised machine learning in distinguishing the clinical heterogeneity of T2DM and the relationships between the subgroups and coronary atherosclerosis characteristics. Accordingly, we used unsupervised machine learning to analyze the heterogeneity of T2DM patients based on clinical indicators and to clarify the comprehensive effect of clinical factors on coronary atherosclerosis characteristics detected on coronary computed tomography angiography (CCTA), which may facilitate individualized clinical management.

Methods

This retrospective study was approved by the Biomedical Research Ethics Committee of our hospital, and written informed consent was waived.

Study cohort

Between January 2018 and November 2021, T2DM patients with coronary plaque detected on CCTA in our hospital were retrospectively reviewed. The exclusion criteria were as follows: patients with a history of coronary artery bypass grafting or stenting before CCTA scanning; CCTA image quality that was too poor for coronary artery plaque assessment; incomplete clinical information; and severe renal failure [estimated glomerular filtration rate (eGFR) < 30 mL/min/1.73 m2]. Finally, 1157 patients with T2DM were included in the study.

CT scanning protocols

The CCTA examinations were performed using a GE CT scanner (Revolution CT, GE Healthcare, Waukesha, WI, USA) or Siemens CT scanner (SOMATOM Definition FLASH, Siemens Medical Solutions, Forchheim, Germany; or SOMATOM Definition, Siemens Medical Solutions, Forchheim, Germany). A bolus of nonionized contrast agent was intravenously injected. The CCTA scanning ranged from the tracheal bifurcation to 20 mm below the inferior cardiac apex. The parameters were as follows: For Siemens CT scanners, tube voltage of 100–120 kV, tube current of 220 mAs; collimation, 64/128 × 0.5 mm. For the GE Revolution CT, the tube voltage was set to 120 kV, the tube current was automatically adjusted, and the image slice thickness was reconstructed to 0.625 mm. A prospective or retrospective electrocardiogram-gated protocol was used for CCTA image acquisition.

CCTA analysis

The CCTA analysis included assessment of plaque type and stenosis of the coronary artery segment and calculation of the segment involvement score (SIS) and segment stenosis score (SSS). By visual evaluation, coronary plaques were classified into three types: calcified plaque, noncalcified plaque and mixed plaque (Fig. 1). Calcified plaque was defined as plaque containing only calcified components; noncalcified plaque was defined as plaque without any calcification that manifested as plaque with computed tomographic density lower than the contrast-enhanced coronary lumen; calcification with noncalcified components shown in a single plaque was defined as mixed plaque [18]. The stenosis degree was estimated according to the Coronary Artery Disease-Reporting and Data System (CAD-RADS) [19]: score 0 (absence of plaque), score 1 (luminal stenosis < 25%), score 2 (25–49% luminal stenosis), score 3 (50–69% luminal stenosis), score 4 (70–99% luminal stenosis), or score 5 (total occlusion). Obstructive stenosis was defined as any presence of stenosis > 50%, and obstructive disease was defined as the presence of obstructive stenosis. The SIS was defined as the number of coronary artery segments observed with plaques. The SSS was defined as the sum of the stenosis scores of the relevant stenosis grades of all segments. According to the modified American Heart Association standard [20], the assessment of the coronary artery included four main vessels (left main artery, left anterior descending coronary artery, left coronary circumflex artery and right coronary artery) and 16 segments. Two cardiovascular radiologists who were blinded to the clinical information of the patients evaluated the images independently. The two observers reached a consensus by discussion when there were divergences.

Fig. 1
figure 1

Representative coronary computed tomography angiography images of A calcified plaque, B noncalcified plaque, and C mixed plaque

Unsupervised machine learning

The K-prototypes algorithm with the elbow method was used for clustering of the 1157 T2DM patients based on clinical characteristics (Fig. 2). The elbow method was used for to determine the optimal number of clusters. The core idea of the elbow method is to minimize the sum of the squared error between the cluster center and the remaining points of the corresponding clusters. As the cluster number (K) increases, the separation of the model is more distinguished. When K increases beyond the optimal value, the sum of the squared error will not substantially change. K-prototype clustering is a method that combines K-means and K-modes for clustering objects mixed with continuous and categorical data [21]. The main steps of the K-prototypes algorithm are as follows: (1) Randomly select K points as initial cluster centers; (2) Calculate the distance (Euclidean distance for continuous data and Hamming distance for categorical data) between the center and the remaining points and assign the remaining points to the closest cluster centers; and (3) Compute the new cluster center by calculating the mean of all samples in each cluster. Steps 2 and 3 are repeated until cluster membership becomes stable. The unsupervised clustering method allowed the optimal number of clusters to explain the overall variance of the data to be determined (cluster group 1: n = 463, cluster group 2: n = 341, cluster group 3: n = 353). The differences in clinical variables and the characteristics of coronary atherosclerosis among the three cluster subgroups were analyzed and compared. The scikit-learn library (version 0.24.2) based on Python (version 3.7.0) was used for the K-prototypes clustering with elbow method.

Fig. 2
figure 2

Schematic for the main steps of this study

Statistical analysis

After the identification of the three cluster groups, clinical information and coronary atherosclerosis characteristics were compared among the cluster groups. All statistical analyses were performed using SPSS software (version 25.0; IBM, Armonk, New York, USA). Categorical variables are presented as numbers (%), and continuous variables are expressed as the mean ± standard deviation in this study. The comparison of clinical information and coronary atherosclerosis characteristics among cluster groups in categorical variables were compared using Fisher’s exact tests (when expected cell value ≤ 5) or the chi-square test followed by Bonferroni’s post hoc test. The Kruskal–Wallis rank test or one-way analysis of variance followed by Bonferroni’s post hoc test was used to compare continuous data among cluster groups. A two-tailed P value of less than 0.05 was considered indicative of statistical significance.

Result

Study population

A total of 1157 T2DM individuals were included in the study, of whom 65.7% (760/1157) were men, with an average age of 69.32 ± 9.89 years old. Based on the 27 clinical parameters of the 1157 T2DM participants, the unsupervised analysis identified three subgroups of T2DM patients with coronary atherosclerosis, which were named cluster group 1 (n = 463), cluster group 2 (n = 341) and cluster group 3 (n = 353). The main clinical characteristics of the participants in the three cluster groups are shown in Table 1.

Table 1 Baseline characteristics of the study cohort

Cluster group 1

The results showed that cluster group 1 had the fewest male patients (41.7%), the lowest proportion of T2DM patients with smoking (0%) or alcohol drinking history (0.9%), and the lowest level of serum creatinine (74.46 ± 22.18 µmol/L). Compared with cluster 2, patients in cluster 1 were more likely to have hypertension (84%), had a longer duration of diabetes (9.94 ± 7.29 years), and had a higher proportion of patients who underwent diabetes treatment. Compared with cluster 3, patients in cluster 1 had a lower proportion of patients with CAD family history (4.1%) and had lower levels of serum uric acid (318.97 ± 97.29 µmol/L).

Cluster group 2

The patients in cluster group 2 had the shortest duration of diabetes (7.90 ± 8.20 years) and had the lowest proportion of patients who underwent diabetes treatment (42.2%) or were treated with statins (17.6%). The patients in cluster group 2 had a lower proportion of hypertension (74.5%) than those in cluster 1, and had a lower level of diastolic blood pressure (77.60 ± 11.51 mmHg) and a higher level of pulse pressure (58.43 ± 17.66 mmHg) those in cluster 3.

Cluster group 3

The patients in cluster group 3 were the youngest (65.89 ± 10.15 years old), had the highest proportion of male patients (96.6%), had the highest proportion of patients with a history of smoking (91.2%) and alcohol drinking (59.8%), had the highest level of eGFR (83.81 ± 19.06 ml/min/1.73m2), and had the lowest level of HDL-C (1.07 ± 0.28 mmol/L).

Association of cluster identity with coronary artery atherosclerosis

The characteristics of coronary artery atherosclerosis among the three clustering groups are compared in Table 2 and Fig. 3.

Table 2 Comparison of coronary artery plaque characteristics of three cluster groups in T2DM
Fig. 3
figure 3

Characteristics of coronary artery atherosclerosis among the three clustering groups. A The number of coronary artery segments (mean value ± standard error [SE]) of different types of plaque; B percentage of T2DM patients with each type of plaque; C the number of coronary artery segments with obstructive and nonobstructive coronary artery stenosis (mean value ± SE); D the percentage of T2DM patients with obstructive coronary disease, ≥ 3 involved vessels, segment involvement score (SIS) ≥ 4, or segment stenosis score (SSS) ≥ 6; E the number of diseased vessels, SIS and SSS (mean value ± SE). *Denotes that the difference between cluster group 2 or cluster group 3 and cluster group 1 is statistically significant. #Denotes that the difference between cluster group 3 and cluster group 2 is statistically significant

In terms of plaque types (Fig. 3A), cluster group 1 had the largest number of segments with calcified plaques (2.43 ± 2.46) and the smallest number of segments with mixed plaques (2.24 ± 2.59). Cluster group 3 had more segments with noncalcified plaques than cluster group 1 (cluster group 3 vs. cluster group 1: 0.63 ± 1.02 vs. 0.40 ± 0.78, P < 0.05). At the patient level (Fig. 3B), cluster group 3 had the highest proportion of patients with noncalcified plaques (39.9%). Cluster group 1 had the lowest proportion of patients with mixed plaques (68.0%). Cluster group 3 had a lower proportion of patients with calcified plaques than cluster 1 (65.7% vs. 76.0%, P < 0.05).

In terms of the degree of coronary artery stenosis (Fig. 3C), cluster group 1 had the least number of segments with obstructive stenosis (0.98 ± 2.00). Cluster group 3 had more segments with moderate stenosis (0.91 ± 1.57 vs. 0.72 ± 1.48) and severe stenosis (0.40 ± 1.09 vs. 0.26 ± 0.91) than cluster group 1 (P values < 0.05). At the patient level, cluster group 1 had the lowest proportion of patients with obstructive stenosis (32.2%) (Fig. 3D).

There was no significant difference in involved vessels, SIS, SSS or proportions of involved vessels ≥ 3, SIS ≥ 4, and SSS ≥ 6 among the three cluster groups (all P values > 0.05) (Fig. 3E).

Discussion

In this study, an unsupervised machine learning method was used to explore the subgroups of T2DM patients with different clinical characteristics. The unsupervised machine learning method provides techniques to integrate various data to enable the discovery of new biomarkers without providing specifications about how to partition the data based on expertise [17, 22]. Our data demonstrated that unsupervised machine learning methods could be used to address heterogeneous clinical data and have the potential to distinguish among subgroups of T2DM patients with different plaque types and degrees of coronary artery stenosis.

Unsupervised machine learning for processing clinical data

It is known that aging, diabetes, hypertension, hyperlipidemia, or declining renal function alone is a risk factor for coronary artery disease [4, 23, 24]. Currently, the combined effects of these factors on coronary artery disease are gaining more attention. Previous research focused mainly on the combination of two or three factors for coronary artery disease [25,26,27]. Unsupervised machine learning for clustering does not rely on manual judgement; thus, this method may have the potential to distinguish a group of patients with similar clinical situations. As the clinical data grow rapidly, the clustering method may be more useful for processing the various and heterogeneous data in electronic clinical records.

Relationship of coronary plaque types with clusters

We observed that the clustering method could not only distinguish T2DM patients with different clinical contexts, but also indirectly identify the group with different types of coronary plaque. The results showed that cluster 3 had relatively more segments with mixed and noncalcified plaques. This may be explained by the fact that cluster group 3 had the highest proportion of males, tended to have the unhealthy habits, including smoking and alcohol drinking, and had the lowest level of HDL-C. Current smoking has been reported to be a risk factor for mixed and noncalcified plaques in coronary atherosclerosis [28]. Mild to moderate alcohol consumption may reduce cardiovascular risk, while heavy drinking may promote to coronary artery calcification [29]. Treatment to raise HDL-C levels regressed coronary plaque and reduced lipid content in plaque [30]. A previous study indicated that the presence of mixed plaques and noncalcified plaques had relatively higher risks of cardiovascular events than calcified plaques [31]. This result draws attention to the importance of proper management for T2DM patients with these risk factors for coronary plaques.

Obstructive coronary disease in clusters

Cluster 1 had the least number of obstructive coronary stenosis cases and the lowest proportion of patients with obstructive coronary disease in this study. This may in part be explained by the fact that cluster 1 had the lowest proportion of males and was less likely to smoke or drink alcohol. A previous study showed that males had a larger plaque volume than females [32]. A multicenter prospective CCTA cohort study showed a similar result: obstructive coronary disease was more prevalent in men than in women (42% vs. 26%) [33]. It has been reported that coronary artery stenosis in T2DM patients adds to the risk of an acute plaque event [34]. Another study also demonstrated that obstructive coronary artery disease was a predictor of cardiac events in diabetic patients [35]. We also noticed that patients in cluster 3 were the youngest. Traditionally, aging is considered to be a risk factor for coronary atherosclerosis. From this, we deduce that cluster 3 would have more obstructive disease when adjusted for age. It also emphasizes the necessity of early diagnosis and timely treatment for T2DM patients with obstructive coronary disease to reduce cardiac events.

Association of extent of coronary atherosclerosis with clusters

Although there were some differences in coronary plaque types and the degree of luminal stenosis, there was no significant difference in the extent of coronary atherosclerosis among the three cluster groups. a previous study demonstrated that the risk of mortality usually correlates with the extent of coronary atherosclerosis in individuals with and without T2DM [36]. Another study showed that atherosclerotic risk factors such as systolic blood pressure, LDL-C, and current smoking have heterogeneous impacts on arterial territories of different vascular diseases, including coronary ischemic and hemorrhagic stroke, abdominal aortic aneurysms, and peripheral arterial disease [37]. We may infer that the extent of coronary atherosclerosis was affected by various factors. Whether and how the different combinations of clinical data with similar coronary characterization affect prognosis are not entirely clear. Thus, the comprehensive effect of the risk factors on coronary atherosclerosis will be investigated in more detail in future studies to verify the benefits of the comprehensive assessment.

Limitations

There are some limitations of this study. First, as a single-center study, selection bias is inevitable, and multicenter studies are needed to verify the results in the future. Second, this was a retrospective study, and follow-up information was not included. The evolution of coronary atherosclerosis in T2DM patients requires further exploration. Third, coronary artery stenosis evaluated with CCTA in this study was not validated with coronary angiography. However, CCTA, a noninvasive examination, can be used to evaluate different plaque types as well as luminal stenosis and has been widely accepted in the evaluation of coronary plaque [38].

Conclusions

The unsupervised clustering method could address T2DM patients with heterogeneous clinical indicators, which may contribute to the stratification of T2DM patients with coronary atherosclerosis. Our findings show one potential way to identify T2DM patients with different types of coronary plaque and degrees of coronary stenosis based on clinical data, which is essential for the clinical management of T2DM patients with coronary atherosclerosis.

Availability of data and materials

The datasets used and analyzed during the current study are available from the corresponding author on reasonable request.

Abbreviations

T2DM:

Type 2 diabetes mellitus

CCTA:

Coronary computed tomography angiography

SIS:

Segment involvement score

SSS:

Segment stenosis score

eGFR:

Estimated glomerular filtration rate

HDL-C:

High-density lipoprotein cholesterol

References

  1. International Diabetes Federation. IDF Diabetes Atlas, 10th edn. Brussels, Belgium: 2021. https://www.diabetesatlas.org. Accessed 20 Jun 2022.

  2. American Diabetes Association: 10. Cardiovascular disease and risk management: standards of medical care in diabetes-2019 Diabetes care 2019;42(Suppl 1): S103-S123. https://doi.org/10.2337/dc19-S010

  3. Einarson TR, Acs A, Ludwig C, Panton UH. Prevalence of cardiovascular disease in type 2 diabetes: a systematic literature review of scientific evidence from across the world in 2007–2017. Cardiovasc Diabetol. 2018;17(1):83.

    Article  PubMed  PubMed Central  Google Scholar 

  4. Malakar AK, Choudhury D, Halder B, Paul P, Uddin A, Chakraborty S. A review on coronary artery disease, its risk factors, and therapeutics. J Cell Physiol. 2019;234(10):16812–23.

    Article  CAS  PubMed  Google Scholar 

  5. DeFronzo RA, Ferrannini E, Groop L, Henry RR, Herman WH, Holst JJ, et al. Type 2 diabetes mellitus. Nat Rev Dis Primers. 2015;1:15019.

    Article  PubMed  Google Scholar 

  6. Wright AK, Kontopantelis E, Emsley R, Buchan I, Mamas MA, Sattar N, et al. Cardiovascular risk and risk factor management in type 2 diabetes mellitus. Circulation. 2019;139(24):2742–53.

    Article  PubMed  Google Scholar 

  7. Raggi P, Cooil B, Ratti C, Callister TQ, Budoff M. Progression of coronary artery calcium and occurrence of myocardial infarction in patients with and without diabetes mellitus. Hypertension. 2005;46(1):238–43.

    Article  CAS  PubMed  Google Scholar 

  8. Poznyak A, Grechko AV, Poggio P, Myasoedova VA, Alfieri V, Orekhov AN. The diabetes mellitus-atherosclerosis connection: the role of lipid and glucose metabolism and chronic inflammation. Int J Mol Sci. 2020;21(5):1835.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  9. Climie RE, van Sloten TT, Bruno RM, Taddei S, Empana JP, Stehouwer CDA, et al. Macrovasculature and microvasculature at the crossroads between type 2 diabetes mellitus and hypertension. Hypertension. 2019;73(6):1138–49.

    Article  CAS  PubMed  Google Scholar 

  10. Hajar R. Risk factors for coronary artery disease: historical perspectives. Heart Views. 2017;18(3):109–14.

    Article  PubMed  PubMed Central  Google Scholar 

  11. Pencina MJ, Navar AM, Wojdyla D, Sanchez RJ, Khan I, Elassal J, et al. Quantifying importance of major risk factors for coronary heart disease. Circulation. 2019;139(13):1603–11.

    Article  PubMed  Google Scholar 

  12. Brown JC, Gerhardt TE, Kwon E: risk factors for coronary artery disease. in: StatPearls. treasure Island (FL): StatPearls Publishing; 2022. https://www.ncbi.nlm.nih.gov/books/NBK554410/

  13. Ceriello A, Prattichizzo F. Variability of risk factors and diabetes complications. Cardiovasc Diabetol. 2021;20(1):101.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  14. Shi R, Gao Y, Shen LL, Shi K, Wang J, Jiang L, et al. The effect of LDL-C status on the association between increased coronary artery calcium score and compositional plaque volume progression in statins-treated diabetic patients: evaluated using serial coronary CTAs. Cardiovasc Diabetol. 2022;21(1):121.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  15. Chen S, Shen Y, Liu YH, Dai Y, Wu ZM, Wang XQ, et al. Impact of glycemic control on the association of endothelial dysfunction and coronary artery disease in patients with type 2 diabetes mellitus. Cardiovasc Diabetol. 2021;20(1):64.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  16. Johnson KM, Johnson HE, Zhao Y, Dowe DA, Staib LH. Scoring of coronary artery disease characteristics on coronary CT angiograms by using machine learning. Radiology. 2019;292(2):354–62.

    Article  PubMed  Google Scholar 

  17. Flores AM, Schuler A, Eberhard AV, Olin JW, Cooke JP, Leeper NJ, et al. Unsupervised learning for automated detection of coronary artery disease subgroups. J Am Heart Assoc. 2021;10(23): e021976.

    Article  PubMed  PubMed Central  Google Scholar 

  18. Hadamitzky M, Hein F, Meyer T, Bischoff B, Martinoff S, Schomig A, et al. Prognostic value of coronary computed tomographic angiography in diabetic patients without known coronary artery disease. Diabetes Care. 2010;33(6):1358–63.

    Article  PubMed  PubMed Central  Google Scholar 

  19. Cury RC, Abbara S, Achenbach S, Agatston A, Berman DS, Budoff MJ, et al. CAD-RADS(TM) coronary artery disease—reporting and data system. an expert consensus document of the society of cardiovascular computed tomography (SCCT), the American college of radiology (ACR) and the North American society for cardiovascular imaging (NASCI). endorsed by the American college of cardiology. J Cardiovasc Comput Tomogr. 2016;10(4):269–81.

    Article  PubMed  Google Scholar 

  20. Austen WG, Edwards JE, Frye RL, Gensini GG, Gott VL, Griffith LS, et al. A reporting system on patients evaluated for coronary artery disease report of the ad hoc committee for grading of coronary artery disease, council on cardiovascular surgery american heart association. Circulation. 1975;51(4):5–40.

    Article  CAS  PubMed  Google Scholar 

  21. Huang Z. Extensions to the k-means algorithm for clustering large data sets with categorical values. Data Min Knowl Disc. 1998;2(3):283–304.

    Article  Google Scholar 

  22. Reel PS, Reel S, Pearson E, Trucco E, Jefferson E. Using machine learning approaches for multi-omics data analysis: a review. Biotechnol Adv. 2021;49: 107739.

    Article  CAS  PubMed  Google Scholar 

  23. Virani SS, Alonso A, Aparicio HJ, Benjamin EJ, Bittencourt MS, Callaway CW, et al. Heart disease and stroke statistics-2021 update: a report from the american heart association. Circulation. 2021;143(8):e254–743.

    Article  PubMed  Google Scholar 

  24. Huang AL, Leipsic JA, Zekry SB, Sellers S, Ahmadi AA, Blanke P, et al. Effects of chronic kidney disease and declining renal function on coronary atherosclerotic plaque progression: a paradigm substudy. Eur Heart J Cardiovasc Imaging. 2021;22(9):1072–82.

    Article  PubMed  Google Scholar 

  25. Wang Z, Yang T, Fu H. Prevalence of diabetes and hypertension and their interaction effects on cardio-cerebrovascular diseases: a cross-sectional study. BMC Public Health. 2021;21(1):1224.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  26. Shigemoto E, Iwata A, Futami M, Kato Y, Yamashita M, Imaizumi S, et al. Influence of chronic kidney disease on coronary plaque components in coronary artery disease patients with both diabetes mellitus and hypertension. Heart Vessels. 2019;34(7):1065–75.

    Article  PubMed  Google Scholar 

  27. Venuraju SM, Lahiri A, Jeevarethinam A, Cohen M, Darko D, Nair D, et al. Duration of type 2 diabetes mellitus and systolic blood pressure as determinants of severity of coronary stenosis and adverse events in an asymptomatic diabetic population: proceed study. Cardiovasc Diabetol. 2019;18(1):51.

    Article  PubMed  PubMed Central  Google Scholar 

  28. Cheezum MK, Kim A, Bittencourt MS, Kassop D, Nissen A, Thomas DM, et al. Association of tobacco use and cessation with coronary atherosclerosis. Atherosclerosis. 2017;257:201–7.

    Article  CAS  PubMed  Google Scholar 

  29. McClelland RL, Bild DE, Burke GL, Mukamal KJ, Lima JA, Kronmal RA, et al. Alcohol and coronary artery calcium prevalence, incidence, and progression: results from the multi-ethnic study of atherosclerosis (MESA). Am J Clin Nutr. 2008;88(6):1593–601.

    Article  CAS  PubMed  Google Scholar 

  30. Rayner KJ, Sheedy FJ, Esau CC, Hussain FN, Temel RE, Parathath S, et al. Antagonism of miR-33 in mice promotes reverse cholesterol transport and regression of atherosclerosis. J Clin Invest. 2011;121(7):2921–31.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  31. Yoon SH, Kim E, Jeon Y, Yi SY, Bae HJ, Jang IK, et al. Prognostic value of coronary CT angiography for predicting poor cardiac outcome in stroke patients without known cardiac disease or chest pain: the assessment of coronary artery disease in stroke patients study. Korean J Radiol. 2020;21(9):1055–64.

    Article  PubMed  PubMed Central  Google Scholar 

  32. Lee SE, Sung JM, Andreini D, Al-Mallah MH, Budoff MJ, Cademartiri F, et al. Sex differences in compositional plaque volume progression in patients with coronary artery disease. JACC Cardiovasc Imaging. 2020;13(11):2386–96.

    Article  PubMed  Google Scholar 

  33. Schulman-Marcus J, Hartaigh BO, Gransar H, Lin F, Valenti V, Cho I, et al. Sex-specific associations between coronary artery plaque extent and risk of major adverse cardiovascular events: the confirm long-term registry. JACC Cardiovasc Imaging. 2016;9(4):364–72.

    Article  PubMed  PubMed Central  Google Scholar 

  34. Halon DA, Lavi I, Barnett-Griness O, Rubinshtein R, Zafrir B, Azencot M, et al. Plaque morphology as predictor of late plaque events in patients with asymptomatic type 2 diabetes: a long-term observational study. JACC Cardiovasc Imaging. 2019;12(7):1353–63.

    Article  PubMed  Google Scholar 

  35. Andreini D, Pontone G, Mushtaq S, Bertella E, Conte E, Baggiano A, et al. Prognostic value of multidetector computed tomography coronary angiography in diabetes: excellent long-term prognosis in patients with normal coronary arteries. Diabetes Care. 2013;36(7):1834–41.

    Article  PubMed  PubMed Central  Google Scholar 

  36. Rana JS, Dunning A, Achenbach S, Al-Mallah M, Budoff MJ, Cademartiri F, et al. Differences in prevalence, extent, severity, and prognosis of coronary artery disease among patients with and without diabetes undergoing coronary computed tomography angiography: results from 10,110 individuals from the CONFIRM (COronary CT angiography evaluation for clinical outcomes): an international multicenter registry. Diabetes Care. 2012;35(8):1787–94.

    Article  PubMed  PubMed Central  Google Scholar 

  37. Stoekenbroek RM, Boekholdt SM, Luben R, Hovingh GK, Zwinderman AH, Wareham NJ, et al. Heterogeneous impact of classic atherosclerotic risk factors on different arterial territories: the EPIC-Norfolk prospective population study. Eur Heart J. 2016;37(11):880–9.

    Article  PubMed  Google Scholar 

  38. Thai PV, Tien HA, Van Minh H, Valensi P. Triglyceride glucose index for the detection of asymptomatic coronary artery stenosis in patients with type 2 diabetes. Cardiovasc Diabetol. 2020;19(1):137.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

Download references

Acknowledgements

Not applicable.

Funding

This work was supported by the Sichuan Province Science and Technology Support Program (2022NSFSC0828), the 1·3·5 project for disciplines of excellence, West China Hospital, Sichuan University (ZYGD18013), and the National Natural Science Foundation of China (81471722).

Author information

Authors and Affiliations

Authors

Contributions

LY and JY designed the study. SR, QWL and YWF were responsible for collecting data. JY and HPL conducted the data analyses. JY drafted the manuscript. LY, YZG, and WJ critically revised the manuscript for important intellectual content. LY is the guarantor. The corresponding author attests that all listed authors meet the authorship criteria and that no others meeting the criteria have been omitted. All authors read and approved the final manuscript.

Corresponding author

Correspondence to Yuan Li.

Ethics declarations

Ethics approval and consent to participate

The institutional review board approved this study, and the written informed consent was waived due to the retrospective nature of this study.

Consent for publication

Not applicable.

Competing interests

The authors declare that there are no conflicts of interest.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Jiang, Y., Yang, ZG., Wang, J. et al. Unsupervised machine learning based on clinical factors for the detection of coronary artery atherosclerosis in type 2 diabetes mellitus. Cardiovasc Diabetol 21, 259 (2022). https://doi.org/10.1186/s12933-022-01700-8

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s12933-022-01700-8

Keywords

  • Machine learning
  • Coronary atherosclerosis
  • Diabetes mellitus
  • Coronary computed tomography angiography