.png){kind=icon}
Return
Constructive learning: a high-performance framework for fetal head circumference estimation
.png){kind=icon}
PRE
AI
DOI：10.1007/s11227-026-08293-z.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Deep learning methods have been applied to fetal head circumference (HC) estimation from ultrasound images. However, accuracy remains limited in biomedical applications. This paper proposes a constructive learning approach to improve segmentation of fetal head regions. The method uses B-spline functions and pruning mechanisms to reduce overfitting and optimize network structure during training. The proposed approach requires high-performance computing for training large-scale models with repeated validation runs and parallel B-spline calculations across multiple network layers. Evaluations on ultrasound datasets show improved accuracy with DSC of 98.43% and reduced overfitting compared to baseline methods. The method achieves 2% improvement in HC detection and reduces overfitting risk by half.
Keywords:
High-performance computing (HPC)
Fetal head circumference
Ultrasonic image
Deep neural networks
Constructive learning
Overfitting
Kolmogorov-arnold network (KAN)
Journal
T
IF:
0
Papers:
647
Citations:
0
Organization
Cited Papers
ShareSaveShareSaveExcessive Enterprise Social Media Use Behavior at Work: Role of Communication Visibility and Perspective of Uses and Gratifications Theory.png)
IEEE ACCESS
OAAI
IEEE ACCESSIF3.6
OAAIShareSaveShareSaveShareSaveShareSave
View more