Butterfly Network iQ

To Next Page

To Previous Page

The algorithm examines all of the frames in the cine and determines the maximum B-line count within a frame

across the cine loop. This maximum count is displayed to the user as the B-line count for the cine. (Note: It is

possible that multiple frames within the cine can have the maximum B-line count.)

The Auto B-line Counter has the capability of identifying cines that are not adequate for automatic B-line calculation

based on an internal quality check of the cine. The tool will return a count of “N/A'' if this occurs. This may occur, for

example, if the pleural line is off-center. In addition to image adequacy, B-line count accuracy can also be affected by

operator skill.

Performance Testing

Two validation studies were performed to evaluate whether the performance of Auto B-line Counter was non-inferior

when compared to clinician annotators (denoted as Study 1 and Study 2). The images collected for these studies

represent a broad and distributed cross-section of patients, including a diverse range of B-line counts, age, gender,

body mass index, ethnicity, and race

Study 1 Description: The objective of Study 1 was to demonstrate the Auto B-line Counter is non-inferior to

clinician annotators (Ground Truth). The primary endpoint was the inter-rater correlation coefficient (ICC) between

the B-line scores from the Auto B-line Counter tool and the B-line scores from the Ground Truth. The secondary

endpoint was the Dice Similarity Coefficient between the centroid-paired segmentation from the Auto B-line Counter

tool and the segmentation from the Ground Truth. Study 1 was a retrospective analysis of de-identified lung

ultrasound cines collected during the standard usage of the Butterfly iQ and Butterfly iQ+ products, uploaded to the

Butterfly Cloud. This data comes from the population of providers using Butterfly devices in concert with the Butterfly

Cloud application in the real world. The clinical validation dataset consists of 253 de-identified six-second cines from

109 clinical sites. The data was from patients aged 22 through 90 with balanced distribution across gender.

Study 2 Description: The Auto B-line Counter Algorithm Clinical Performance Evaluation was a supplemental

validation study designed to demonstrate the generalizability of the Auto B-line Counter across the relevant patient

demographic categories. The primary endpoint of this study was to demonstrate the Auto B-line counter algorithm

performance is non-inferior to consensus clinician interpretation (Ground Truth). The secondary objective of this

study was to evaluate the algorithm’s performance among diverse subgroups of age, gender, BMI/habitus, ethnicity,

and race. The primary endpoint was the inter-rater correlation coefficient (ICC) between the Auto B-line Counter tool

and the Ground Truth equal. Study 2 was a retrospective secondary data analysis of de-identified lung ultrasound

cines and subject demographic information collected from a single site during an IRB-approved study. Data was

collected from patients 22 years old or older that consented to participate in the study, and were included based

on their history of admission to a general care, telemetry, or moderate care unit with clinical concerns that included

pulmonary congestion. All patients enrolled in the study received lung ultrasound exams with the Butterfly iQ/ iQ+

Ultrasound system with the Lung preset. All cines were saved in the Butterfly Cloud. The data was curated to cines

from 97 unique subjects. The non-identifying subject demographic data collected included age, gender, height and

weight (for BMI), ethnicity, and race; these are summarized in the table below.

Table 8. Demographic breakdown of Study 2, n=97

Related product manuals