Fetal lung maturity analysis using ultrasound image features.

Area of Science:

• Medical Imaging
• Fetal Development
• Quantitative Ultrasound

Background:

• Assessing fetal lung maturity is crucial for predicting neonatal pulmonary risk.
• Current methods for evaluating fetal lung maturity can be invasive or limited.
• Ultrasound imaging offers a non-invasive approach to fetal assessment.

Purpose of the Study:

• To determine the feasibility of analyzing fetal lung maturity using ultrasound images.
• To investigate the efficacy of textural features from fetal lung and liver ultrasound images for maturity classification.
• To evaluate various machine learning classifiers for accurate fetal lung maturity assessment.

Main Methods:

• Collected ultrasound images from 24 to 38 weeks of gestation from normal pregnancies.
• Extracted textural features (fractal dimension, lacunarity, histogram-derived) from fetal lung and liver regions of interest.
• Computed ratios of fetal lung to liver features as potential maturity indicators.
• Employed multiple classifiers including k-nearest neighbor, radial basis function network, and support vector machines.

Main Results:

• Classification accuracies for fetal lung maturity ranged from 73% to 96% on the testing set.
• Ratios of specific textural features between fetal lung and liver demonstrated effectiveness in differentiating mature from immature lungs.
• The study successfully classified images into mature and immature fetal lung categories.

Conclusions:

• Ultrasound-based analysis of fetal lung and liver textural features is a feasible method for assessing fetal lung maturity.
• This non-invasive technique shows promise in identifying fetuses at risk for pulmonary complications.
• Further validation with larger datasets is warranted to establish clinical utility.