Fig. 2: The distribution of predict scores related to the duct-dependent CHD determined by the Transfer Learning Group and the General Group in retrospective datasets. | npj Digital Medicine

Fig. 2: The distribution of predict scores related to the duct-dependent CHD determined by the Transfer Learning Group and the General Group in retrospective datasets.

From: A multicenter study on two-stage transfer learning model for duct-dependent CHDs screening in fetal echocardiography

Fig. 2: The distribution of predict scores related to the duct-dependent CHD determined by the Transfer Learning Group and the General Group in retrospective datasets.

a Transfer Learning Group, b General Group. DDCHD-DenseNet uses the DensenNet-169 model architecture, and its performance is shown in (a) DenseNet-169. Risk scores (range 0–1) and confusion matrix predicted by the deep learning model for discerning fetal genetic diseases. Scores closer to 1 denote a higher probability of genetic diseases. The upper and lower bounds of the box refer to the 25th and 75th percentiles, and the line intersection in the box refers to the median. Whiskers refer to the full range of risk scores.

Back to article page