Table 2 Summary of aggregation methods and NaN removal used for each dataset.

From: Prediction of preterm birth from cervical length measurements in twin pregnancies using machine learning

Datasets

Aggregation Features

Gestational Age Bin: Number of CL NaNs

Method of handling NaNs

General

Minimum, maximum and average CL and GA for all exams

Differences between minimum and maximum CL and GA

Feature cross for CL and GA features

No Missing Values

None

Single

Average CL binned within the 18–20 week of GA window

Bin-18-20: 469

NaNs were removed from Dataset

GA-16-28

Average CL binned within the 18–20, 20–22, 22–24, 24–28 week of GA windows

Bin-16-18: 452

Bin-18-20: 339

Bin-20-22: 527

Bin-22-24: 336

Bin-24-28: 130

Single measurements of CL were dropped

NaNs within CL bins were interpolated (linear)

NaNs remaining were dropped

GA-18-24

Average CL binned within the 18–20, 20–22, 22–24 week of GA windows

Bin-18-20: 339

Bin-20-22: 527

Bin-22-24: 336

Single measurements of CL were dropped

NaNs within CL bins were interpolated (linear)

NaNs remaining were dropped