Automatic staging of placental maturity based on dense descriptor

Area of Science:

• Medical Imaging
• Computational Pathology
• Obstetrics and Gynecology

Background:

• Placental maturity staging is crucial for fetal well-being assessment.
• Current methods rely on subjective physician observation, leading to potential inconsistencies.

Purpose of the Study:

• To develop an automated placental maturity staging method using B-mode ultrasound images.
• To improve the objectivity and accuracy of placental maturity assessment.

Main Methods:

• Utilized dense sampled DAISY descriptors for scale and translation invariance.
• Employed Vector Locally Aggregated Descriptor (VLAD) for feature encoding and performance boosting.
• Applied the method to B-mode ultrasound images for placental maturity staging.

Main Results:

• Achieved an accuracy of 0.874 for placental maturity staging.
• Demonstrated high sensitivity (0.996) and specificity (0.874).
• Dense features significantly outperformed sparse features in this application.

Conclusions:

• The proposed automated method offers a promising, objective approach to placental maturity staging.
• Dense feature extraction and VLAD encoding are effective for analyzing subtle variations in ultrasound images.
• This technique has the potential to enhance prenatal care and fetal monitoring.