Biologically Inspired Deep Learning Approaches for Fetal Ultrasound Image Classification

2025-06-10

AI Summary

Fetal ultrasound image classification during the second trimester faces clinical challenges including low image quality, high intra-class variability, and severe class imbalance. Method: We propose a biologically inspired dual-path deep learning ensemble framework featuring a novel modular two-path architecture—shallow coarse-grained and deep fine-grained pathways—enabling end-to-end joint classification of 16 fetal anatomical structures within a lightweight model. The framework integrates EfficientNet-B0 and EfficientNet-B6 backbones, employs LDAM-Focal loss to mitigate class imbalance, and incorporates Dawid-Skene modeling for multi-expert annotation fusion, trained directly on real-world noisy clinical data (5,298 routine images). Results: Our method achieves >0.75 accuracy for 90% of anatomical structures and >0.85 for 75%, matching or exceeding the performance of more complex models on fewer-class tasks. This demonstrates robustness, scalability, and clinical applicability in realistic deployment scenarios.

Technology Category

• Machine Learning: Ensemble Methods
• Computer Vision: Multi-modal Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Web Mining and Content Analysis: Large pretrained models with web data
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

Accurate classification of second-trimester fetal ultrasound images remains challenging due to low image quality, high intra-class variability, and significant class imbalance. In this work, we introduce a simple yet powerful, biologically inspired deep learning ensemble framework that-unlike prior studies focused on only a handful of anatomical targets-simultaneously distinguishes 16 fetal structures. Drawing on the hierarchical, modular organization of biological vision systems, our model stacks two complementary branches (a"shallow"path for coarse, low-resolution cues and a"detailed"path for fine, high-resolution features), concatenating their outputs for final prediction. To our knowledge, no existing method has addressed such a large number of classes with a comparably lightweight architecture. We trained and evaluated on 5,298 routinely acquired clinical images (annotated by three experts and reconciled via Dawid-Skene), reflecting real-world noise and variability rather than a"cleaned"dataset. Despite this complexity, our ensemble (EfficientNet-B0 + EfficientNet-B6 with LDAM-Focal loss) identifies 90% of organs with accuracy>0.75 and 75% of organs with accuracy>0.85-performance competitive with more elaborate models applied to far fewer categories. These results demonstrate that biologically inspired modular stacking can yield robust, scalable fetal anatomy recognition in challenging clinical settings.

Problem

• Classifying diverse fetal ultrasound images accurately
• Addressing low quality and class imbalance in images
• Recognizing multiple fetal structures with lightweight model

Innovation

• Biologically inspired deep learning ensemble
• Hierarchical modular two-branch architecture
• Lightweight model for 16-class classification