Efficient Fetal Ultrasound Image Segmentation with MFP-Unet: A Deep Learning Approach

doi:10.26599/AUDT.2026.250052

Abstract

Abstract:

Background and purpose Fetal ultrasound image segmentation is a crucial component of prenatal diagnosis, enabling the accurate measurement of key biometric parameters and the early detection of abnormalities. Manual segmentation is prone to variability due to factors such as operator skill, image quality, and acquisition conditions. This study compares conventional and deep learning-based segmentation methods, proposing an enhanced Multi-Feature Pyramid U-Net (MFP-Unet) to improve accuracy, robustness, and clinical applicability.
Methods We utilised the HC18 Grand Challenge Dataset, which comprises 1,334 ultrasound images from 551 pregnancies without growth abnormalities. The proposed MFP-Unet integrates multi-scale feature pyramid fusion, dilated convolutions, and attention gates into the U-Net framework. Performance was compared with Dilated U-Net, U-Net, RU-Net, Attention U-Net, and R2U-Net using Dice Coefficient, Intersection over Union (IoU), and mean Average Precision (mAP). Preprocessing included Niblack’s adaptive thresholding and custom region extraction for femur length images.
Results MFP-Unet achieved the highest accuracy across all parameters: HC IoU = 90% (Dice = 0.95), FL IoU = 93% (Dice = 0.97), BPD IoU = 88% (Dice = 0.93), and overall mAP = 0.92. The architecture formed smoother anatomical boundaries, enhanced robustness to image variability, and compact measurement variations associated with other methods.
Conclusions The MFP-Unet approach shows strong promise for use in clinical fetal ultrasound analysis, offering reliable and precise biometric measurements that can aid in timely interventions and enhance prenatal care outcomes. Future research should validate performance on multi-centre datasets and explore real-time deployment.

Key words: Deep learning; Fetal ultrasound segmentation; Multi-Feature Pyramid U-Net; Biometric measurement; Prenatal care

Ruby A.Usha, Chandran.J George Chellin. Efficient Fetal Ultrasound Image Segmentation with MFP-Unet: A Deep Learning Approach. Advanced Ultrasound in Diagnosis and Therapy, 2026, 10(3): 216-222.

/ Recommend

Figures/Tables 9

Figure 1

Figure 2

Table 1

Figure 3

Figure 4

Table 2

Table 3

Figure 5

Figure 6

References 32

[1]	Liu S, Wang Yi, Yang X, Lei B, Liu Li, Li SX, et al. Deep learning in medical ultrasound analysis: a review. Engineering 2019; 5: 261-275.
[2]	Adineh-Vand A, Torabi M, Roshani GH, Taghipour M, Feghhi SAH, Rezaei M, et al. Application of adaptive neuro-fuzzy inference system for prediction of neutron yield of IR-IECF facility in high voltages. J Fusion Energy 2014; 33: 13-19.
[3]	Hearn-Stebbins B. Normal fetal growth assessment: a review of the literature and current practice. J Diagn Med Sonogr 1995; 11: 176-187. doi: 10.1177/875647939501100403
[4]	Loughna P, Chitty L, Evans T, Chudleigh T. Fetal size and dating: charts recommended for clinical obstetric practice. Ultrasound 2009; 17: 160-166. doi: 10.1179/174313409x448543
[5]	Bandeira Diniz P, Yin Y, Collins S. Deep learning strategies for ultrasound in pregnancy. Eur Med J Reprod Heal 2020; 6: 73-80.
[6]	Rueda S, Fathima S, Knight CL, Yaqub M, Papageorghiou AT, Rahmatullah B, et al. Evaluation and comparison of current fetal ultrasound image segmentation methods for biometric measurements: a grand challenge. IEEE Trans Med Imaging 2014; 33: 797-813. doi: 10.1109/TMI.2013.2276943
[7]	GE Versana Club-SonoBiometry n.d. https://www.versanaclub.net/emea/sono-biometry (accessed August 31, 2020).
[8]	Mohammadi R, Shokatian I, Salehi M, Arabi H, Shiri I, Zaidi H. Deep learning auto-segmentation of organs at risk in high-dose-rate brachytherapy of cervical cancer. Radiother Oncol 2021; 159: 231-240. doi: 10.1016/j.radonc.2021.03.030
[9]	Arabi H, AkhavanAllaf A, Sanaat A, Shiri I, Zaidi H. The promise of artificial intelligence and deep learning in PET and SPECT imaging. Phys Med 2021; 83: 122-137. doi: 10.1016/j.ejmp.2021.03.008
[10]	Shiri I, Arabi H, Salimi Y, Sanaat AH, Akhavanalaf A, Hajianfar G, et al. COLI-NET: Fully automated COVID-19 lung and infection pneumonia lesion detection and segmentation from chest CT Images. MedRxiv 2021.
[11]	Shiri I, Arabi H, Sanaat A, Janebi E, Becker M, Zaidi H. Fully automated gross tumour volume delineation from PET in head and neck cancer using deep learning algorithms. Clin Nucl Med 2021.
[12]	Ronneberger O, Fischer P, Brox T. U-Net: Convolutional networks for biomedical image segmentation. Int Conf Med Image Comput Comput Interv, Springer 2015:234-241.
[13]	Alom MZ, Hasan M, Yakopcic C, Taha TM, Asari VK. Recurrent residual convolutional neural network based on u-net (r2u-net) for medical image segmentation. ArXiv Prepr ArXiv180206955 2018.
[14]	Liang M, Hu X. Recurrent convolutional neural network for object recognition. Proc IEEE Conf Comput Vis pattern Recognit 2015:3367-3375.
[15]	Oktay O, Schlemper J, Le FL, Lee M, Heinrich M, Misawa K, et al. Attention u-net: Learning where to look for the pancreas. ArXiv Prepr ArXiv180403999 2018.
[16]	Lee J. Image_Segmentation 2018. https://github.com/LeeJunHyun/Image_Segmentation.
[17]	Moradi S, Oghli MG, Alizadehasl A, Shiri I, Oveisi N, Oveisi M, et al. MFP-Unet: A novel deep learning-based approach for left ventricle segmentation in echocardiography. Phys Med 2019; 67: 58-69. doi: 10.1016/j.ejmp.2019.10.001
[18]	Lin T-Y, Dollar ´ P, Girshick R, He K, Hariharan B, Belongie S. Feature pyramid networks for object detection. CVPR 2017:2117-2125.
[19]	van den Heuvel TLA, de Bruijn D, de Korte CL, van Ginneken B. Automated measurement of fetal head circumference. PLoS One 2018; 13: e0200412. doi: 10.1371/journal.pone.0200412
[20]	Kerbyson D, Atherton T. Circle detection using Hough transform filters. Image Process its Appl 1995:370-374.
[21]	Sobhaninia Z, Rafiei S, Emami A, Karimi N, Najarian K, Samavi S, et al. Fetal ultrasound image segmentation for measuring biometric parameters using multi-task deep learning. Annu Int Conf IEEE Eng Med Biol Soc 2019; 2019: 6545-6548. doi: 10.1109/embc.2019.8856981
[22]	Chaurasia A, Culurciello E. LinkNet: Exploiting encoder representations for efficient semantic segmentation. In 2017 IEEE Vis Commun Image Process VCIP 2017, vol. 2018- Janua, Institute of Electrical and Electronics Engineers Inc. 2018:1–4.
[23]	Sinclair M, Baumgartner CF, Matthew J, Bai W, Martinez JC, Li Y, et al. Human-level performance on automatic head biometrics in fetal ultrasound using fully convolutional neural networks. In Proc Annu Int Conf IEEE Eng Med Biol Soc EMBS, vol. 2018- July, Institute of Electrical and Electronics Engineers Inc. 2018:714-717.
[24]	Long J, Shelhamer E, Darrell T. Fully convolutional networks for semantic segmentation. IEEE Conf Comput Vis pattern Recognit 2015:3431-3440.
[25]	Irene K, Yudha P. A, Haidi H, Faza N, Chandra W. Fetal head and abdomen measurement using convolutional neural network, hough transform, and difference of gaussian revolved along elliptical path (Dogell) algorithm. ArXiv Prepr ArXiv191106298 2019.
[26]	Redmon J, Divvala S, Girshick R, Farhadi A. You only look once: unified, real-time object detection. IEEE Conf Comput Vis Pattern Recognit 2016:779-788.
[27]	Foi A, Maggioni M, Pepe A, Rueda S, Noble JA, Papageorghiou AT, et al. The difference of Gaussians revolved along elliptical paths for ultrasound fetal head segmentation. Comput Med Imaging Graph 2014; 38: 774-784. doi: 10.1016/j.compmedimag.2014.09.006
[28]	Carneiro G, Georgescu B, Good S, Comaniciu D. Detection and measurement of fetal anatomies from ultrasound images using a constrained probabilistic boosting tree. IEEE Trans Med Imaging 2008; 27: 1342-1355. doi: 10.1109/TMI.2008.928917
[29]	Tu Z. Probabilistic boosting-tree: Learning discriminative models for classification, recognition, and clustering. Proc IEEE Int Conf Comput Vis 2005; II: 1589-1596.
[30]	Usha Ruby, A., Chandran, J. G. C., Theerthagiri, P., Patil, R., Chaithanya, B. N., & Jain, T. S. (2024). Forecasting PM2. 5 Concentration Using Gradient-Boosted Regression Tree with CNN Learning Model. Optical Memory and Neural Networks 2024;33:86-96.
[31]	Rahmatullah R, Ma’Sum MA, Aprinaldi, Mursanto P, Wiweko B. Automatic fetal organs segmentation using multilayer superpixel and image moment feature. In Proc - ICACSIS 2014 2014 Int Conf Adv Comput Sci Inf Syst, Institute of Electrical and Electronics Engineers Inc. 2014:420-426.
[32]	Schlemper Jo, Oktay O, Schaap M, Heinrich M, Kainz B, Glocker B, et al. Attention gated networks: learning to leverage salient regions in medical images. Med Image Anal 2019; 53: 197-207. doi: 10.1016/j.media.2019.01.012

Metric	Formula	Role of the study
IoU	IoU = TP / (TP + FP + FN)	Assesses how closely predicted regions align with the reference segmentation.
Dice	Dice = 2TP / (2TP + FP + FN)	Measures segmentation boundary accuracy by balancing precision and recall.
mAP	Mean of precision values across IoU thresholds	Evaluates overall detection accuracy under varying overlap requirements.

Model	HC IoU (%)	HC Dice Coeff.	BPD IoU (mm)	BPD Dice Coeff.	FL IoU (mm)	FL Dice Coeff.	Overall Seg. mAP
Notably: HC: IoU = 90.0%; Dice = 0.95; BPD: IoU = 88.0%; Dice = 0.93; FL: IoU = 93.0%, Dice = 0.97; Overall mAP: 0.92
U-net	85.0	0.90	0.82	0.88	0.88	0.92	0.87
Dilated U-net	87.0	0.92	0.84	0.90	0.90	0.94	0.89
Attention U-net	89.0	0.94	0.86	0.92	0.92	0.96	0.91
RU-net	86.0	0.91	0.83	0.89	0.89	0.93	0.88
R2U-net	88.0	0.93	0.85	0.91	0.91	0.95	0.90
MFP-Unet	90.0	0.95	0.88	0.93	0.93	0.97	0.92

Set	Parameter	Correlation coefficien t (R)	Slope (a)	Intercept (b, mm)
Training	Biparietal Diameter (BPD)	0.98	1.02	−0.05
	Head Circumference (HC)	0.96	1.03	0.1
	Abdominal Circumference (AC)	0.97	1.00	−0.01
	Femur Length (FL)	0.99	1.01	−0.02
Testing	Biparietal Diameter (BPD)	0.97	1.01	−0.03
	Head Circumference (HC)	0.95	1.02	0.12
	Abdominal Circumference (AC)	0.96	0.99	0.02
	Femur Length (FL)	0.98	1.00	0.01