Related Experiment Video
Updated: Jul 20, 2025
Human Fetal Blood Flow Quantification with Magnetic Resonance Imaging and Motion Compensation
Published on: January 7, 2021
A Computer-Aided Markov Random Field Segmentation Algorithm for Assessing Fetal Ventricular Chambers
Natarajan Sriraam1, T V Sushma2, S Suresh3
1Centre for Medical Electronics and Computing, MS Ramaiah Institute of Technology, Bangalore 560054, India.
Insights
This study evaluated automated segmentation for fetal heart chambers, crucial for detecting congenital heart disease (CHD). The method showed promising results comparable to expert annotations, aiding early diagnosis.
Area of Science:
- Medical Imaging
- Biomedical Engineering
- Fetal Cardiology
Background:
- Congenital heart disease (CHD) is the most common birth defect, affecting fetal heart development.
- Accurate segmentation of fetal cardiac chambers is vital for early detection and intervention.
- Ultrasound image speckle noise complicates fetal heart chamber segmentation.
Purpose of the Study:
- To evaluate the performance of an automated segmentation approach for fetal ventricular chambers.
- To compare automated segmentation results with manual annotations by clinical experts.
- To assess the effectiveness of probability-based segmentation and Markov Random Field (MRF) methods.
Main Methods:
- Utilized 837 ultrasonic biometry sequences from various gestations.
- Employed automated, probability-based segmentation and Markov Random Field (MRF) techniques.
- Validated segmentation efficiency using Dice coefficient, True Positive Ratio (TPR), Similarity Ratio (SIR), and Precision (PR).
- Incorporated ground truth validation through expert clinical annotation on 56% of the data.
Main Results:
- Automated segmentation achieved comparable results to manual annotations.
- Achieved an average Dice coefficient of 0.68.
- Reported an average TPR of 0.723, SIR of 0.604, and PR of 0.632.
- Demonstrated the potential of automated methods in challenging fetal cardiac imaging.
Conclusions:
- The automated segmentation technique provides a reliable method for analyzing fetal ventricular chambers.
- This approach can aid in the early and accurate diagnosis of congenital heart disease.
- Further development could enhance segmentation accuracy and clinical applicability.
Abstract:
Congenital heart disease (CHD) is the most widely occurring congenital defect and accounts to about 28% of the overall congenital defects. Analysis of the development of the fetal heart thus plays an important role for detection of abnormality in early stages and to take corrective measures. Cardiac chamber analysis is one of the important diagnosing methods. Segmentation of the cardiac chambers must be done appropriately to avoid false interpretations. Effective segmentation of fetal ventricular chambers is a challenging task as the speckle noise inherent in ultrasound images cause blurring of the boundaries of anatomical structures. Several segmentation techniques have been proposed for extracting the fetal cardiac chambers. This article discusses the performance evaluation of automated, probability based segmentation approach, and Markov random field (MRF) for segmenting the fetal ventricular chambers of ultrasonic cineloop sequences. 837 ultrasonic biometery sequences of various gestations were collected from local diagnostic center after due ethical clearance and used for the study. In order to assess the efficiency of the segmentation technique, four metrics such as dice coefficient, true positive ratio (TPR), false positive ratio (FPR), similarity ratio (SIR), and precision (PR) were used. In order to perform ground truth validation, 56% of the data used in this study were annotated by clinical experts. The automated segmentation yielded comparable results with manual annotation. The technique results in average value of 0.68 for Dice coefficient, 0.723 for TPR, 0.604 for SIR, and 0.632 for PR.
