Classification of patients with early cardiac dysfunction among childhood cancer survivors using a machine learning algorithm with whole heart cycle strain data.
AI interpretation is pending for this paper.
Open original publication →What the AI sees
Not AI summarized yet.
Research significance
Pending deeper interpretation.
Source abstract
Predicting cardiac dysfunction due to anthracycline-induced cardiotoxicity in childhood cancer survivors (CCSs) remains challenging when relying on conventional echocardiographic assessments based on peak values. Therefore, in this study, we aimed to evaluate the utility of machine learning (ML)-based whole cardiac cycle strain analysis in the longitudinal assessment of cardiac function in CCSs. A total of 121 CCSs (aged 4-36 years) who had undergone anthracycline treatment and completed at least one echocardiographic assessment at least 1 year post-treatment were included. The echocardiographic evaluation included standard measurements, left ventricular longitudinal strain (LS), and circumferential strain at the basal (BCS), papillary muscle, and apical levels. Strain data from the entire cardiac cycle were normalized to 100 positional data points. Unsupervised ML clustering was applied to the CCS cohort, stratifying patients into normal-like and impaired groups for comparative analysis. The impaired group exhibited lower strain values, delayed peak strain timing, and a later onset of BCS compared with the normal-like group. At the initial evaluation, the impaired group demonstrated a significantly higher heart rate and lower E/A ratio and e' velocity; however, no significant differences were observed in ejection fraction (EF), LS, or BCS between the groups. Longitudinal assessment revealed that LS and EF declined significantly over time in the impaired group, whereas no significant changes were noted in the normal-like group. In conclusion, ML analysis of whole-cardiac-cycle strain data may facilitate the early detection of subclinical diastolic dysfunction; in addition, it may enable the classification of CCSs at potential risk of progressive cardiac dysfunction over time.