Unsupervised machine learning for risk stratification and identification of relevant subgroups of ascending aorta dimensions using cardiac CT and clinical data

Abstract

The potential of precision population health lies in its capacity to utilize robust patient data for customized prevention and care targeted at specific groups. Machine learning has the potential to automatically identify clinically relevant subgroups of individuals, considering heterogeneous data sources. This study aimed to assess whether unsupervised machine learning (UML) techniques could interpret different clinical data to uncover clinically significant subgroups of patients suspected of coronary artery disease and identify different ranges of aorta dimensions in the different identified subgroups. We employed a random forest-based cluster analysis, utilizing 14 variables from 1170 (717 men/453 women) participants. The unsupervised clustering approach successfully identified four distinct subgroups of individuals with specific clinical characteristics, and this allows us to interpret and assess different ranges of aorta dimensions for each cluster. By employing flexible UML algorithms, we can effectively process heterogeneous patient data and gain deeper insights into clinical interpretation and risk assessment.

Keywords: Aortic dimensions; Clusterization; Computed tomography coronary angiography; Coronary artery disease; Unsupervised learning.

Graphical abstract

Fig. 1

Cluster subgroups identified by unsupervised clustering. Plot showing 4 groups (Males A, Males B, Females A, Females B) of individuals identified by Random Forest and PAM (Partitioning Around Medoids) analysis.

Fig. 2

Comparison of variable importance in males (A) and Females (B) clusterization (from Random Forest model). We used 0.4 as the normalized weight threshold to select features used in the score definition. There is a difference on the x-axis between males and females because they have been analyzed separately and two different models (with two different feature importance) were built. AV, Aorta Variables; BV, Basic variables; RFV, Risk Factor Variables.

Fig. 3

Schematic representation of different ranges of aorta dimensions assigned to each cluster. A novel sample can be assigned to a cluster using their score compared to Th (threshold obtained from random forest feature importance analyses).

Figure S1

Comparison of Aorta dimension in males (A) and females (B) across the clusters. In all cases, there is a significant difference of values between cluster A and cluster B (p < 0.05)

Figure S2

Comparison of BMI (Body Mass Index), number of risk factor and age in males (A) and females (B) across the clusters. BMI was represented as Low (L) if < 25, High (H) if >=30 and Medium (M) if >= 24 and <30; Number of Risk Factors was represented as High (H) if >=3 and Low (L) if <3.