Tibbiyot, ta’lim va innovatsion tadqiqotlar, 2026, т. 2, № 11, с. 93-97
ARTIFICIAL INTELLIGENCE FOR ULTRASOUND ESTIMATION OF FETAL WEIGHT: OPPORTUNITIES, EVALUATION METRICS, AND COMPARISON WITH THE HADLOCK FORMULA
Аннотация · Abstract
Background: Knowing the estimated fetal weight before delivery is important for assessing fetal growth and planning obstetric care. The Hadlock formula has remained the principal method for ultrasound-based fetal weight estimation for decades; however, its individual prediction errors remain real and measurable. Machine-learning methods, in turn, offer the possibility of identifying nonlinear relationships between fetal biometric measurements and clinical context that fixed equations cannot capture. Main body: This narrative review examines established ultrasound-based formulas, analyses the findings of recent machine-learning studies, and discusses how estimated fetal weight should be compared with actual birth weight. The most informative evaluation metrics include absolute error, mean absolute percentage error (MAPE), the proportion of predictions within ±10% of actual birth weight, systematic bias and Bland–Altman limits of agreement, as well as predictive accuracy across clinically relevant weight categories. Particular attention should be paid to the interval between ultrasound examination and delivery, as the fetus continues to grow after the measurements have been taken. Conclusion: Artificial intelligence represents a genuinely promising approach to improving the accuracy of fetal weight estimation. However, better statistical performance does not necessarily translate into better medical care. Before any model is introduced into clinical practice, it must undergo external validation, calibration assessment, and evaluation of its effectiveness across clinically relevant subgroups.
Сведения · Details
- DOI 10.5281/zenodo.24052780
- Журнал · Journal Tibbiyot, ta’lim va innovatsion tadqiqotlar
- Язык · Language английский
- Цитирования (Crossref) · Citations 0