Key result
Multimodal ultrasound machine learning accurately assesses carotid plaque vulnerability with an AUC of ~0.94.
Why the study?
To explore the application value of a machine learning diagnosis model based on multimodal ultrasound in assessing carotid plaque vulnerability.
Does a machine learning diagnosis model based on multimodal ultrasound accurately assess carotid plaque vulnerability in patients with carotid plaques?
Observational (n=100)
Randomly split (1:4 ratio)
No
Does a machine learning diagnosis model based on multimodal ultrasound accurately assess carotid plaque vulnerability in patients with carotid plaques?
Effect estimate: AUC 0.939 (95% CI 0.8824-0.9964)
A decision tree machine learning model based on multimodal ultrasound demonstrated high diagnostic accuracy (AUC 0.939) for identifying vulnerable carotid plaques.
May aid clinical identification of vulnerable carotid plaques; extends ultrasound methods with ML but should not yet change practice.
Objective To explore the application value of a machine learning diagnosis model based on multimodal ultrasound in assessing carotid plaque vulnerability. Methods A retrospective study was conducted, selecting 100 patients with carotid plaques admitted to our hospital from January 2022 to August 2023. The patients were divided into two groups based on the presence of vulnerable plaques: the vulnerable group ( n = 46) and the stable group ( n = 54). The patients were then randomly matched in a 1:4 ratio with data from 100 patients during the same period to establish a validation group ( n = 20) and a modeling group ( n = 80). Clinical data of the patients in the modeling and validation groups were compared. Univariate analysis was performed to identify factors influencing carotid plaque vulnerability. The machine learning diagnosis model based on multimodal ultrasound was developed, and the value of the DT algorithm model in assessing carotid plaque vulnerability was analyzed with ROC analysis. Results No significant differences were observed in clinical data, age, or gender between the modeling and validation groups ( p > 0.05). The vulnerable group had higher rates of diabetes, hypertension, smoking, and BMI than the stable group ( p < 0.05), while plaque thickness, length, and location were similar ( p > 0.05). However, the vulnerable group exhibited more ulcers, calcifications, low‐echogenic images, and greater enhancement, with lower Young’s modulus values in all segments ( p < 0.05). ROC analysis of the DT algorithm for plaque vulnerability yielded an AUC of 0.939 (95% CI: 0.8824–0.9964), with optimal sensitivity and specificity of 93.10% and 82.76%, respectively. Conclusion The DT algorithm model based on multimodal ultrasound has a high application value in assessing carotid plaque vulnerability.
No takes yet. Share an insight, caveat, or question.
Yan et al. (2026) conducted an observational in Carotid plaques (n=100). Machine learning diagnosis model (DT algorithm) based on multimodal ultrasound was evaluated on Carotid plaque vulnerability (AUC 0.939, 95% CI 0.8824-0.9964). A decision tree machine learning model based on multimodal ultrasound accurately assessed carotid plaque vulnerability, yielding an AUC of 0.939 (95% CI: 0.8824-0.9964).
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: