Abstract
BACKGROUND: Neonatal cerebral microbleeds (CMBs) occur infrequently, and during the initial phase, they often present without noticeable clinical symptoms, which can result in delays in both diagnosis and treatment. There has been relatively little research conducted on neonatal CMBs, with even less focus on their related risk factors. However, identifying risk factors and proactively preventing microbleeds is particularly crucial for effective treatment. Therefore, we aimed to develop and validate a nomogram based on clinical characteristics and to assess its efficacy in predicting neonatal CMBs. METHODS: This study included 230 neonates who were treated at The First Hospital of Tsinghua University and underwent a 1.5-T magnetic resonance imaging (MRI). There were 115 neonates with CMBs and 115 sex-matched healthy controls. The clinical and MRI data were collected, including gender, term or premature birth, mode of delivery, gestational age, days after birth, adjusted gestational age, birth weight, Apgar score, history of asphyxia, neonatal pneumonia, metabolic acidosis, mechanical ventilation, gestational hypertension and diabetes, intraventricular hemorrhage, subdural hemorrhage, ischemic infarction, with or without CMBs, and the number and grading of CMBs. All neonates were randomly divided into a training and validation cohort at a ratio of 7:3. Significant variables were selected to construct a nomogram based on multivariate logistic regression analysis results. The model's performance was assessed by receiver operating characteristic (ROC) curve, calibration curve, and decision curve analysis. RESULTS: Spontaneous delivery [odds ratio (OR) =7.88; 95% confidence interval (CI): 3.27-19.00; P<0.001], neonatal pneumonia (OR =2.63; 95% CI: 1.16-6.25; P=0.020), gestational hypertension (OR =4.69; 95% CI: 1.35-16.26; P=0.015), and gestational diabetes (OR =3.60; 95% CI: 1.24-10.40; P=0.018) were independent risk factors for neonatal CMBs. The models' area under the curve (AUC), corresponding optimal threshold, specificity, and sensitivity were 0.811 (95% CI: 0.746-0.877), 0.630, 0.872, and 0.627 in the training cohort and were 0.780 (95% CI: 0.667-0.892), 0.366, 0.649, and 0.875 in the validation cohort, respectively. The calibration and decision curve analysis showed that the model had high calibration and clinical application value. We also constructed a combined prediction model for moderate-to-severe CMBs based on clinical and MRI data. The results revealed that the presence of ischemic infarction (OR =5.00; 95% CI: 1.51-16.60; P=0.009) was an independent risk factor for moderate-to-severe CMBs; the models' AUC, optimal threshold, specificity, and sensitivity were 0.731 (95% CI: 0.574-0.888), 0.187, 0.786, and 0.706, respectively. CONCLUSIONS: The model based on these independent risk factors could effectively predict the occurrence of neonatal CMBs and may aid in early diagnosis and treatment.