Computer-Aided Detection of Biparietal Diameter and Femur Length Measurements from Prenatal Ultrasound Images

Abstract Background Routine ultrasound exams for fetal biometry are hindered by the reliance on obstetricians' expertise, emphasizing the necessity for an automated computer-aided method to ensure accurate gestational age estimation. This paper describes a completely automatic method for identifying and accurately measuring two biometrics: biparietal diameter (BPD) and femur length (FL). Methods The method uses first- and second-order total variation denoising (FSTVD) to remove speckle noise and preserve edges in prenatal ultrasound images. Automatic estimation of FL and BPD was implemented by selecting and determining regions of interest based on their shape, size, location, and orientation in the filtered images. The method was tested on 431 ultrasound images, and the results were compared with manual measurements made by expert obstetricians. Pearson’s correlation and the Bland–Altman plot were used to assess the agreement between the automatic and manual measurements. Results The results showed a strong correlation (r > 0.99) between the automatic and manual measurements for both FL and BPD. A mean error of less than 2 mm was found for the difference between automatic and manual measurements for both FL and BPD. Additionally, the processing time for filtering a regular-sized image and determining FL or BPD using the presented method was less than 0.5 seconds. Conclusions The proposed method demonstrates high accuracy in estimating FL and BPD from ultrasound images. Additionally, its short processing time allows for real-time estimation of FL and BPD within ultrasound imaging systems.