An informative probability model enhancing real time echobiometry to improve fetal weight estimation accuracy.

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Fetal Medicine

Background:

• Fetal echobiometry is crucial for estimating fetal weight (EFW), but human measurement errors can impact accuracy.
• Existing methods for correcting these errors are limited, potentially affecting clinical decisions.
• Standardizing ultrasound procedures and improving data analysis are key challenges in fetal assessment.

Purpose of the Study:

• To develop and validate a multinormal probability model to correct human errors in fetal echobiometry.
• To enhance the accuracy of fetal weight estimation (EFW) using artificial neural networks (ANNs).
• To create a software tool for real-time error checking, staff training, and clinical decision support.

Main Methods:

• A multinormal probability model was designed with parameters dependent on pregnancy data.
• Feed-forward artificial neural networks (ANNs) were trained and tested using data from 4075 women.
• The model was numerically implemented to provide EFW and congruence probabilities, enabling interactive error correction.

Main Results:

• The developed model demonstrated real-time checking and interactive correction of ultrasound measurement errors.
• Software implementation proved useful for training medical staff and standardizing measurement procedures.
• Clinical testing with 61 women showed decisive improvements in EFW accuracy, validating the system's effectiveness.

Conclusions:

• The proposed multinormal probability model effectively corrects human errors in fetal echobiometry.
• Artificial neural networks (ANNs) are valuable tools for improving fetal weight estimation (EFW) accuracy.
• The developed software offers significant benefits for medical training, procedural standardization, and clinical decision-making in fetal assessment.