LASSO-Nomogram model for ultrasound-atherosclerosis correlation and diagnostic verification in anterior circulation

TL;DR

This study developed a LASSO-Nomogram model to predict the severity of anterior circulation cerebral atherosclerosis using ultrasound and Doppler parameters.

Contribution

A novel LASSO-Nomogram model was created for individualized risk assessment of cerebral atherosclerosis using ultrasound data.

Findings

The model showed excellent discrimination with a C-index of 0.850 in training and 0.796 in validation.

Key predictors included lipid profiles, carotid IMT, plaque stability, and MCA hemodynamics.

HDL and stable plaques were protective factors, while LDL, IMT, and MCA parameters were risk factors.

Abstract

This study aimed to investigate the relationship between cervical vascular ultrasound/transcranial Doppler ultrasound (TCD) parameters and anterior circulation cerebral atherosclerosis severity, and to develop a LASSO-Nomogram predictive model for clinical assessment.

We retrospectively analyzed 350 patients with anterior circulation atherosclerosis, randomly divided into training (n = 245) and validation (n = 105) sets. Collected data included: (1) demographics and medical history; (2) lipid profiles; (3) ultrasound parameters [carotid intima-media thickness (IMT), plaque stability, internal carotid artery stenosis rate, middle cerebral artery peak systolic velocity (MCA-PSV), end-diastolic velocity (MCA-EDV), pulsatility index (PI), resistance index (RI)]. Patients were stratified by atherosclerosis severity (mild-moderate vs. severe). LASSO regression identified key predictors for nomogram prediction model construction, with model performance rigorously evaluated.

Baseline characteristics were balanced between training set and validation set (p > 0.05). Univariate analysis identified 10 significant factors (all p < 0.05). LASSO regression selected 9 key predictors: age, high-density lipoprotein (HDL), low-density lipoprotein (LDL), carotid IMT, plaque stability, stenosis rate, and MCA hemodynamics (PSV, EDV, RI). Multivariate analysis showed HDL (OR = 7.410) and stable plaques (OR = 3.987) as protective factors of arteriosclerosis, while LDL (OR = 0.621), carotid IMT (OR = 0.038), MCA-PSV (OR = 0.978), MCA-EDV (OR = 0.960), and RI (OR = 0.010) were risk factors of arteriosclerosis (all p < 0.05). The model demonstrated excellent discrimination (training C-index = 0.850; validation = 0.796) with AUCs of 0.849 (95% CI: 0.792–0.907) and 0.801 (95% CI: 0.698–0.904), respectively. Decision curve analysis confirmed clinical utility across threshold probabilities of 10–80%.

Cervical vascular ultrasound and TCD parameters effectively reflect anterior circulation atherosclerosis severity. Our LASSO-Nomogram model provides clinicians with a reliable, visualized tool for individualized risk assessment, potentially improving patient management.

Figures