Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

MiTU-Net: A fine-tuned U-Net with SegFormer backbone for segmenting pubic symphysis-fetal head

Fangyijie Wang, Guenole Silvestre, Kathleen Curran

TL;DR

This work targets automatic assessment of labor progression via the angle of progression (AoP) by segmenting fetal head and pubic symphysis in transperineal ultrasound. The authors introduce MiTU-Net, a fine-tuned U-Net that replaces the encoder with a pre-trained Mix Transformer encoder (MiT-B0) from SegFormer to achieve high-quality FH–PS segmentation with substantially fewer trainable parameters. Using a 3-class segmentation (background, PS, FH) on the JNU-IFM dataset, MiTU-Net achieves a final score of 0.9283 and a mean Dice of 0.9247, ranking 5th among state-of-the-art methods while enabling automatic AoP measurement with reduced computational cost. The approach combines detailed patch merging, self-attention, and a lightweight FFN to balance accuracy and efficiency, with code and models publicly available for reproducibility and broader clinical adoption.

Abstract

Ultrasound measurements have been examined as potential tools for predicting the likelihood of successful vaginal delivery. The angle of progression (AoP) is a measurable parameter that can be obtained during the initial stage of labor. The AoP is defined as the angle between a straight line along the longitudinal axis of the pubic symphysis (PS) and a line from the inferior edge of the PS to the leading edge of the fetal head (FH). However, the process of measuring AoP on ultrasound images is time consuming and prone to errors. To address this challenge, we propose the Mix Transformer U-Net (MiTU-Net) network, for automatic fetal head-pubic symphysis segmentation and AoP measurement. The MiTU-Net model is based on an encoder-decoder framework, utilizing a pre-trained efficient transformer to enhance feature representation. Within the efficient transformer encoder, the model significantly reduces the trainable parameters of the encoder-decoder model. The effectiveness of the proposed method is demonstrated through experiments conducted on a recent transperineal ultrasound dataset. Our model achieves competitive performance, ranking 5th compared to existing approaches. The MiTU-Net presents an efficient method for automatic segmentation and AoP measurement, reducing errors and assisting sonographers in clinical practice. Reproducibility: Framework implementation and models available on https://github.com/13204942/MiTU-Net.

MiTU-Net: A fine-tuned U-Net with SegFormer backbone for segmenting pubic symphysis-fetal head

TL;DR

This work targets automatic assessment of labor progression via the angle of progression (AoP) by segmenting fetal head and pubic symphysis in transperineal ultrasound. The authors introduce MiTU-Net, a fine-tuned U-Net that replaces the encoder with a pre-trained Mix Transformer encoder (MiT-B0) from SegFormer to achieve high-quality FH–PS segmentation with substantially fewer trainable parameters. Using a 3-class segmentation (background, PS, FH) on the JNU-IFM dataset, MiTU-Net achieves a final score of 0.9283 and a mean Dice of 0.9247, ranking 5th among state-of-the-art methods while enabling automatic AoP measurement with reduced computational cost. The approach combines detailed patch merging, self-attention, and a lightweight FFN to balance accuracy and efficiency, with code and models publicly available for reproducibility and broader clinical adoption.

Abstract

Ultrasound measurements have been examined as potential tools for predicting the likelihood of successful vaginal delivery. The angle of progression (AoP) is a measurable parameter that can be obtained during the initial stage of labor. The AoP is defined as the angle between a straight line along the longitudinal axis of the pubic symphysis (PS) and a line from the inferior edge of the PS to the leading edge of the fetal head (FH). However, the process of measuring AoP on ultrasound images is time consuming and prone to errors. To address this challenge, we propose the Mix Transformer U-Net (MiTU-Net) network, for automatic fetal head-pubic symphysis segmentation and AoP measurement. The MiTU-Net model is based on an encoder-decoder framework, utilizing a pre-trained efficient transformer to enhance feature representation. Within the efficient transformer encoder, the model significantly reduces the trainable parameters of the encoder-decoder model. The effectiveness of the proposed method is demonstrated through experiments conducted on a recent transperineal ultrasound dataset. Our model achieves competitive performance, ranking 5th compared to existing approaches. The MiTU-Net presents an efficient method for automatic segmentation and AoP measurement, reducing errors and assisting sonographers in clinical practice. Reproducibility: Framework implementation and models available on https://github.com/13204942/MiTU-Net.
Paper Structure (15 sections, 5 equations, 3 figures, 1 table)

This paper contains 15 sections, 5 equations, 3 figures, 1 table.

Table of Contents

  1. Introduction
  2. Methodology
  3. Model Achitecture
  4. Feature Representation
  5. Overlapped Patch Merging
  6. Self-Attention
  7. Feed-forward Network
  8. Decoder
  9. Data pre-processing and augmentation
  10. Fine-tuning
  11. Loss Function
  12. Evaluation Metrics
  13. Experiments Results
  14. Evaluation Metrics
  15. Compare with SOTA Methods

Figures (3)

  • Figure 1: The illustration of the process of measuring the angle of progression (AoP) using transperineal ultrasound. (A) Schematic diagram of calculating AoP. (B) An image showing the symphysis pubis and fetal head.
  • Figure 2: The illustration of the proposed MiTU-Net for automatic fetal head (FH)-pubic symphysis (PS) segmentation.
  • Figure 3: Visualization of segmentation results on the JNU-IFM dataset. GT: Ground Truth