Fig. 3

Prediction performance of the early prediction models of LBW. A In the term birth group, the model showed the highest performance with health assessment data (F1-score of 0.95) and modest performance with both SNP array data and fetal ultrasonography data (F1-score of 0.84 for both). B In the preterm birth group, the model based on the SNP array data (F1-scores of 0.90) showed the highest performance. The models based on other data, including health assessment data and fetal ultrasonography data, showed insufficient performance (F1-scores of 0.71 and 0.69, respectively)