Method and system for automatic detection of congenital diaphragmatic hernia in fetus

By employing a two-stage cascaded detection framework and an improved FetalDH-YOLO model, combined with ultrasound and MRI images, the limitations of multimodal imaging collaboration and the challenge of small-target detection in prenatal fetal congenital diaphragmatic hernia detection have been addressed. This approach achieves efficient and accurate fetal diaphragmatic hernia detection, aligns with clinical diagnostic logic, and reduces the consumption of medical resources.

CN122244008APending Publication Date: 2026-06-19WUHAN UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
WUHAN UNIV
Filing Date
2026-05-11
Publication Date
2026-06-19

AI Technical Summary

Technical Problem

Existing technologies for prenatal detection of congenital diaphragmatic hernia in fetuses suffer from several problems, including insufficient multimodal imaging collaboration, incomplete data processing, poor model generalization ability, weak detection capability for small targets, and a disconnect between the detection framework and clinical diagnostic logic. These issues limit detection efficiency and accuracy, making it difficult to apply effectively in clinical settings.

Method used

A two-stage cascaded detection framework is adopted, combining ultrasound and MRI images. Multi-class target detection is performed through an improved FetalDH-YOLO detection model. Small target perception attention mechanism and multi-scale feature enhancement module are integrated. The bounding box regression is optimized using ShapeIoU loss function. Strictly following the clinical step-by-step confirmation diagnostic thinking, the system achieves accurate detection of diaphragmatic hernia.

Benefits of technology

It significantly improves the accuracy and efficiency of testing, reduces the consumption of medical resources, enhances the detection rate and positioning accuracy of small target organs, and has excellent prospects for clinical application.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122244008A_ABST
    Figure CN122244008A_ABST
Patent Text Reader

Abstract

This application discloses an automated method and system for detecting congenital diaphragmatic hernia in fetuses, belonging to the field of medical image processing technology. This application employs a multimodal, two-stage cascaded detection framework. First, it uses ultrasound four-chamber view images for initial screening; then, it uses an MRI pre-classification module to remove negative samples; finally, it uses the FetalDH-YOLO detection model for multi-class target detection, outputting the positive or negative result of the diaphragmatic hernia, the location of the herniated organ, and fine-grained categories of four herniated organs: stomach, intestine, liver, and kidney. The FetalDH-YOLO detection model integrates small-target perception dual-layer routing attention, an optimized multi-scale spatial pyramid pooling module, and a ShapeIoU loss function. This application significantly improves the accuracy and small-target recall rate of fetal congenital diaphragmatic hernia detection by leveraging the complementary advantages of ultrasound and MRI, cascaded filtering of invalid samples, and targeted network structure optimization. It aligns with the clinical diagnostic logic of step-by-step confirmation and is easy to promote and apply clinically.
Need to check novelty before this filing date? Find Prior Art