A deep learning framework for classification of fetal brain abnormalities in ultrasound images
- 1Department of Computer Engineering, Smt. Kashibai Navale College of Engineering, Savitribai Phule Pune University
- 2Department of Information Technology, Siddhant College of Engineering, Savitribai Phule Pune University
Abstract
Ultrasound imaging during pregnancy is commonly utilized to evaluate fetal growth and detect congenital anomalies. Effective classification of fetal brain abnormalities is vital for achieving early diagnosis and enhancing clinical decision-making in prenatal care. This work lies at the convergence of health sciences and engineering, leveraging medical imaging and sophisticated computational techniques to tackle challenges such as speckle noise, limited contrast, and anatomical variability in ultrasound imaging. At present, deep learning (DL) has emerged as a powerful approach, achieving substantial success in image processing. Nevertheless, fetal brain ultrasound image detection encounters several challenges, such as the accurate identification of fetal brain tissues and the issue of overfitting, which affect the overall detection performance. A parallel convolutional wide residual network (PCWRN) is proposed to overcome the existing limitations in fetal brain abnormality classification. The process begins by selecting a fetal brain ultrasound image from the dataset, followed by image enhancement using thresholding transformations. Afterwards, channel prior convolutional attention (CPCA) is utilized for region localization of the fetal brain Region of Interest (ROI) in ultrasound images. Thereafter, the image is augmented based on the augmentation techniques, and then, features are extracted for further classification process. Lastly, fetal brain abnormalities are classified utilizing PCWRN, which is modelled by integrating a parallel convolutional neural network (PCNN) with a wide residual network (WRN). Additionally, PCWRN obtained maximum values of accuracy about 91.461%, true negative rate (TNR) about 92.908%, true positive rate (TPR) about 90.609%, and precision about 91.179%. The experimental findings suggest that the proposed model is a reliable computer-aided diagnostic tool for the classification of fetal brain abnormalities. Furthermore, this research advances the fields of medical image analysis and biomedical engineering, with promising applications in prenatal diagnosis and healthcare decision support.
Related
- CERVEX: A Foundation-Model Framework With Built-In Explainable AI for Automated Cervical Cytology Classification From Pap Smear Images
- Enhancing prenatal diagnostic workflow: a deep learning-based system for multi-class fetal ultrasound classification across all gestational stages
- DualXBrNet: A Hybrid Dual-Path Cross-Level Fusion Framework for Multiclass Brain Tumor Classification Using MRI Images
- An explainable hybrid transformer–CNN framework for multi-class brain tumor classification from MRI images
- A unified deep learning framework for multi-lesion detection in diabetic retinopathy using fundus images
BibTeX
@article{navale2026deep,
title = {{A deep learning framework for classification of fetal brain abnormalities in ultrasound images}},
author = {Manisha Pandurang Navale and Brijendra P. Gupta},
journal = {Scientific Reports},
year = {2026},
month = sep,
doi = {10.1038/s41598-026-70379-8},
url = {https://doi.org/10.1038/s41598-026-70379-8},
note = {CC BY},
}
RIS
TY - JOUR
TI - A deep learning framework for classification of fetal brain abnormalities in ultrasound images
AU - Manisha Pandurang Navale
AU - Brijendra P. Gupta
PY - 2026
DA - 2026/09/01
JO - Scientific Reports
DO - 10.1038/s41598-026-70379-8
UR - https://doi.org/10.1038/s41598-026-70379-8
AB - Ultrasound imaging during pregnancy is commonly utilized to evaluate fetal growth and detect congenital anomalies. Effective classification of fetal brain abnormalities is vital for achieving early diagnosis and enhancing clinical decision-making in prenatal care. This work lies at the convergence of health sciences and engineering, leveraging medical imaging and sophisticated computational techniques to tackle challenges such as speckle noise, limited contrast, and anatomical variability in ultrasound imaging. At present, deep learning (DL) has emerged as a powerful approach, achieving substantial success in image processing. Nevertheless, fetal brain ultrasound image detection encounters several challenges, such as the accurate identification of fetal brain tissues and the issue of overfitting, which affect the overall detection performance. A parallel convolutional wide residual network (PCWRN) is proposed to overcome the existing limitations in fetal brain abnormality classification. The process begins by selecting a fetal brain ultrasound image from the dataset, followed by image enhancement using thresholding transformations. Afterwards, channel prior convolutional attention (CPCA) is utilized for region localization of the fetal brain Region of Interest (ROI) in ultrasound images. Thereafter, the image is augmented based on the augmentation techniques, and then, features are extracted for further classification process. Lastly, fetal brain abnormalities are classified utilizing PCWRN, which is modelled by integrating a parallel convolutional neural network (PCNN) with a wide residual network (WRN). Additionally, PCWRN obtained maximum values of accuracy about 91.461%, true negative rate (TNR) about 92.908%, true positive rate (TPR) about 90.609%, and precision about 91.179%. The experimental findings suggest that the proposed model is a reliable computer-aided diagnostic tool for the classification of fetal brain abnormalities. Furthermore, this research advances the fields of medical image analysis and biomedical engineering, with promising applications in prenatal diagnosis and healthcare decision support.
SN - 2045-2322
ER -