Automated Segmentation and Measurement of the Gestational Sac in Ultrasound Images
Author Information
Author(s): Danish Hafiz Muhammad, Suhail Zobia, Farooq Faiza
Primary Institution: University of the Punjab, Lahore, Pakistan
Hypothesis
Can deep learning improve the accuracy of gestational sac segmentation and gestational age estimation in ultrasound images?
Conclusion
The proposed pipeline offers a precise and reliable alternative to conventional manual measurements for gestational sac segmentation and gestational age estimation.
Supporting Evidence
- The ResUNet model achieved the best performance with an IoU of 0.946 and a Dice score of 0.978.
- The mean absolute error for gestational age estimation was 0.07 weeks.
- The study included a diverse dataset of both normal and abnormal pregnancy cases.
Takeaway
This study created a computer program that helps doctors measure the size of a pregnancy sac in ultrasound pictures, making it easier to check if everything is okay.
Methodology
The study used a dataset of 500 ultrasound scans and trained four deep learning models to segment the gestational sac and estimate gestational age.
Potential Biases
Potential bias due to reliance on manual selection of the largest view of the gestational sac by sonographers.
Limitations
The study was conducted at a single center, which may limit the generalizability of the findings, and the sample size was relatively small.
Participant Demographics
{"normal_cases":274,"abnormal_cases":226,"maternal_age":{"normal":"29 (19–40)","abnormal":"32 (22–46)"},"gravidity":{"normal":"2 (1–4)","abnormal":"3 (1–5)"},"history_of_RPL":{"normal":"17 (6.2%)","abnormal":"6 (2.7%)"}}
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website