LUS assessment
LUS was performed using CX50; (Philips Healthcare, Eindhoven, Netherlands) The frequency of the linear array probe was 10 to 13 MHz.
Lung ultrasound was performed at one of the following timings after birth: 0.5 h, 1 h, 2 h, 4 h, and 6 h. The LUS was performed by a physician who received two months of formal training. This training includes a 2-day course and practice on no less than 100 cases under senior supervision14.
The scanning protocol was based on an adult adaptive method20 and adjusted for infants. LUS was performed in a total of ten regions, as shown in Figure S1. Scanning was performed continuously and quickly in each region to avoid missing images or images disturbed by the motion of the infants. A similar protocol has been used in other studies21, but we reduced the total number of regions from twelve (in the previous studies) to ten because the area of two lateral (left and right) regions was nearly as large as that of the other eight regions. The infants were positioned in a supine, lateral, or prone position, as needed. In every region, if any ”high-risk” patterns were detected, then the region was marked as ”high-risk”. By contrast, regions were only defined as “low-risk” when the whole region had no ”high-risk” patterns. The number of ”high-risk” regions and ”low-risk” regions were used to assess the accuracy of the prediction. The definitions of ”high-risk” patterns and ”low-risk” patterns were determined by a pilot experiment and confirmed in this study and are shown in Figure 2.
Results were recorded on a dedicated study form that was not included in patients’ files and was masked to other clinicians; this was the best way to mask the clinical conditions from the colleagues performing the LUS and the LUS results from other clinicians14. The images and their interpretations for each participant were recorded by the LUS examiner and were linked to a serial number (SN) from 1 to 310 (after excluding those without qualified data). The SNs then linked to the infant outcomes by the physician who offered NICU treatment or followed these infants. Finally, an independent data analyst who had no knowledge of the lung ultrasound results analyzed the data.